guest editorial - telenor

Mobile Virtual Network Operators

1 Guest Editorial; Do Van Thanh

3 The Mobile Virtual Network Operator concept: Truth and Myths; Do Van Thanh

7 An Analysis of the MVNO Business Model; Anders Lillehagen, Lars Armyr,Terje Hauger, Vegard Masdal and Kari-Ann Skow

15 Why Should You Invest in MVNO?; Timothy Staley, Loc H Khuong and Do Van Thanh

23 Access to Mobile Networks; Eli Solvang

26 Service Opportunities for Realistic MVNO Models; Eskil Dahlen, Stein Troneng, Jan-Tore Deilkås and Steinar Låg

34 How do the Price Agreements Affect the Business Case of an MVNO?; Kjetil Andersson, Hilde Halvorsen and Ole Christian Wasenden

39 An Entrant Strategy for an MVNO Based on Communities; Kenth Engø, Eirik Befring and Kenneth Tilley

46 Modelling Market Uncertainty and Competition in TelecommunicationEnvironment: Network Providers and Virtual Operators;Jan A Audestad, Alexei A Gaivoronski and Adrian Werner

65 The Future of Virtual Network Operation: Lessons from the ICT-Industries;Svein Ulset

82 MORO – Move-On Real Options; Kenth Engø and Steinar Låg

98 The Open Service Access and Opportunities for MVNOs; Do Van Thanh and Gunvald Martin Grødem

106 Option Pricing of Mobile Virtual Network Operators; Jan A Audestad and Alexei A Gaivoronski

115 Status: Introduction; Per Hjalmar Lehne

116 IETF and Internet Standards; Zaw-Sing Su

120 Towards a new Internet Architecture? – Or how IPv6 Addresses the Future;Paal Engelstad

Contents

Telektronikk

Volume 97 No. 4 – 2001

ISSN 0085-7130

Editor:

Ola Espvik

Tel: (+47) 913 14 507

email: [email protected]

Status section editor:

Per Hjalmar Lehne

Tel: (+47) 916 94 909


Editorial assistant:

Gunhild Luke

Tel: (+47) 415 14 125


Editorial office:

Telenor Communication AS

Telenor R&D

N-1331 Fornebu

Norway

Tel: (+47) 67 89 00 00


Editorial board:

Ole P. Håkonsen,

Senior Executive Vice President.

Oddvar Hesjedal,

Vice President, R&D.

Bjørn Løken,

Director.

Graphic design:

Design Consult AS, Oslo

Layout and illustrations:

Gunhild Luke, Britt Kjus (Telenor R&D)

Prepress and printing:

Optimal as, Oslo

Circulation:

3,750

1

There is no business like ... Mobile business.Although being involved with the mobile busi-ness for a while now I never stop being amazedand fascinated by the way things happen. Onemay, of course, imagine an unorganized devel-opment since wireless communication is indeed“untethered” and “unguided” by nature, but no-body can ever dream that so many unpredictablethings can happen and so many predictable onesnever do. Although difficult, one could foreseethe success of GSM since it brings a real valueto the users, namely freedom in communica-tions. However, few or perhaps no one couldhave anticipated that SMS actually became thekiller application. In fact, how can anybodyimagine that typing messages on tiny telephonekeypads could gain such popularity amongyoungsters? Looking into the future – who cansay what will be the killer application?

An interesting and strange phenomenonoccurred in 1998. As if it is not enough with thedisappearance of the wires, the Mobile NetworkOperator in his turn tends to vanish as well bybecoming virtual. By launching the concept ofthe Mobile Virtual Network Operator (MVNO),Sense Communications (“Sense”) in Norwaycreated such turmoil in the mobile business. Itwanted to get interconnection to operator’s mo-bile networks in Scandinavia. Although obtain-ing a lot of press coverage and supports from theregulators Sense succumbed due to the lack ofcommercial agreements with operators who con-sidered MVNO as a threat.

However, the MVNO concept did not die. Infact, Sense came back from bankruptcy in late1999 as a “thinner” MVNO or Service Provideras many may say, and did not fight against butrather allied themselves with mobile networkoperators. The MVNO concept may rise strongerwith 3G. On one side, mobile network operatorshaving paid high prices for 3G licenses, at least,may need MVNOs who buy available network

capacity and help financing their debt. Manyoperators also realise that they cannot addressevery sector in the market and need the expertisefrom MVNOs. On the other hand, companieswishing to offer mobile services may considerMVNO as a good entrance way to the mobilebusiness. The MVNO concept provides also thepossibilities for operators to extend their cover-age. 2G operators may extend their coverageby having MVNO agreement with 3G licenseholder and vice versa. Operators may also pro-vide 3G coverage in countries where they donot hold licenses.

Briefly, many things, known and unknown, pre-dictable and unpredictable, will continue to hap-pen in a rather confusing way for the MVNOconcept. Therefore, this Telektronikk issue isaiming at shedding light on the Mobile VirtualNetwork Operator concept by considering thetechnological, financial, economic aspects.

Within Telenor there have been several activitiesrelated to MVNO. Telenor Corporate Universityhas launched the Master of TelecommunicationStrategy Program in cooperation with Norwe-gian University of Science and Technology in2000. This program has at is core a commonproject where the students are developing a busi-ness plan. The first time this program was run,MVNO was chosen as core project.

Through a workshop on “MVNO, access tomobile networks and the juridical and economicaspects regarding regulation and competition”organised by Telenor Brussels on the 19 March2001 in Brussels, experts and regulators had theopportunity to meet and extend their view onMVNO. Many authors in this issue have alsoparticipated in the session on Mobile VirtualNetwork Operators: Technologies, Services,Markets and regulations organised by ProfessorJan A. Audestad and myself at the World Multi-conference on Systemics, Cybernetics and Infor-

Guest Editorial

Do Van Thanh

Front cover: New actorsemerge and vanish, competeand co-operate according tomarket demands.

Traditionally a network opera-tor possessed his own physi-cal network. The operatoralso provided the servicesoffered in that network. How-ever, the physical networksbasically constitute a “road”system enabling an ever-increasing variety of servicesolutions. Therefore, “virtual”providers/operators need nophysical facilities. Insteadthey rent facilities from others.To make things work properlythe virtual operator mustextend its service provisionmanagement into and throughthe domains of others. Theartist Odd Andersen indicatesthis situation by showingmobile providers – hereabstracted to waves extend-ing outwards – expandingtheir influence over availablenetwork resources – hereabstracted to “the band” – beit narrow, medium or broad innature. The managerial andcontrol mechanisms are indi-cated by downward beamsof hierarchical and collateralinfluence centred at the coreof the virtual operators.

The artist’s generic message:Matching each operator’sinfluence to the proper rangeand hierarchical level.

Ola Espvik, Editor in Chief

Telektronikk 4.2001

2 Telektronikk 4.2001

matics (SCI 2001), July 22–25, 2001 in Orlando,Florida.

As mentioned, this issue covers different aspectsof MVNO from technologies, business, regula-tion and opportunities. It starts with an elucida-tion of the MVNO concept. Since MVNO is abusiness concept several articles in this issuesfocuses on the business aspects covering every-thing from the financing, market, pricing busi-ness models and service opportunities. The regu-

latory aspects are also covered. The issue con-cludes with a look on the future with an article onfuture network architecture and one on pricing.

I hope that after reading this issue the readerswill either have a good understanding about theMVNO concept or get confused and provokedenough to start doing research on MVNO ...

Enjoy your reading.

3

IntroductionThe first idea concerning Virtual Network Oper-ator (VNO), that is, operators not owning theirown communication infrastructure, goes back toearly nineties. The early VNOs were concernedwith resale of traffic and marginal enhancementof services because value-added services re-quired Intelligent Network functionality. TheInternet combined with third generation mobileservices are likely changing this picture, givingrise to new ways of implementing services.

The concept of Mobile Virtual Network Opera-tor (MVNO) arose three years ago with SenseCommunication, called Netsystem International.The company fought for interconnection tomobile operators’ networks in Scandinavia.Although having the support of regulators it didnot succeed in reaching commercial agreementswith mobile operators that consider MVNO asa threat. Sense went bankruptcy but came backagain more as a service provider than an MVNOand did not attempt to extend regulatory frontiers.

An MVNO is an operator not owning radioinfrastructure. The MVNO carries out its busi-

ness by roaming, where the customers roambetween the radio infrastructures of ordinarymobile operators. The MVNO must, as a mini-mum, own a subscription database. It may alsoown additional platforms for executing services(for example Internet services). The MVNO atthe same time competes and cooperates withordinary mobile operators. It competes for thesame customers as the ordinary operator and itcooperates by buying infrastructure services (air-time) from that operator, and sometimes, by cre-ating teletraffic that otherwise would not exist.Examples of the latter are offering connection topublic information services and supporting fleetmanagement. The role of the MVNO is thereforecomplex. It is likely that regulators of the tele-com market will require that ordinary mobileoperators must cooperate with MVNOs in orderto enhance competition in the marketplace andstimulate evolution of mobile services.

The definition of the MVNO is vague. The rea-son is that there will be many ways in whichsuch operator are formed, and that several typesof MVNO may compete in the same market.Note also that the concept of MVNO also may

The Mobile Virtual Network OperatorConcept: Truth and MythsD O V A N T H A N H

Do Van Thanh (44) obtained hisMSc in Electronic and ComputerSciences from the NorwegianUniv. of Science and Technology(NTNU) in 1984 and his PhD inInformatics from the Universityof Oslo in 1997. In 1991 hejoined Ericsson R&D Depart-ment in Oslo after 7 years ofR&D at Norsk Data, a minicom-puter manufacturer in Oslo. In2000 he joined Telenor R&D andis now in charge of PANDA (Per-sonal Area Network & DataApplications) research activitieswith a focus on SIP, XML andnext generation mobile applica-tions. He also holds a professorposition at the Department ofTelematics at NTNU in Trond-heim. He is author of numerouspublications and inventer of adozen patents.

[email protected]

IN VLR AuC

BSC MSCEIR

BSC MSC AuC

EIRIN VLR

HLR

PSTN / ISDN

Home network

Visiting network

Network SubsystemBase Station SubsystemMobilephone

SM

MEBTS

BTS

BTS

Figure 1 Roaming registration of mobile phone

Telektronikk 4.2001


comprise mobility between fixed network termi-nations, mobility of sessions and mobility ofapplications. We are not only referring to tradi-tional services in traditional networks. In thisarticle, we will try to elucidate the MVNO con-cept and to make difference between truth andmyths.

MVNO versus MNOIn order to understand how a mobile networkoperator (MNO) can be “virtual” let us startrecalling what a mobile network operator is andwhat infrastructure and equipment it must have.

When subscribing to an MNO the subscriberreceives a SIM (Subscriber Identity Module),which is a smart card containing the Interna-tional Mobile Subscriber Identity (IMSI), whichallows the Mobile Network Operator to uniquelyidentify the subscriber. Indeed, The IMSI con-sists of three fields: the Mobile Country Code,The Mobile Network Code and the Customer ID.In addition, each base station broadcasts a loca-tion code consisting of information that makes itpossible for the mobile terminal to determine thearea in which it is operating: cell, VLR, network,country. The mobile phone, also called MobileEquipment (ME) in GSM terminology cannotoperate normally without a SIM and only emer-gency calls are allowed in this case. The mobilephone has its own identity called InternationalMobile Equipment Identity (IMEI).

The Mobile Switching Center (MSC) is organ-ised under the Network Subsystem. Other ele-ments in the subsystem are the Home LocationRegister (HLR), the Visitor Location Register(VLR), the Authentication Center (AuC) and theEquipment Identity Register (EIR). The Subsys-tem may be strengthened with an Intelligent net-work Node (IN) that enables enhanced servicessuch as Call forwarding, Voice mail, etc. TheMSC is a switching node providing connectionto PSTN/ISDN or other network subsystemsbelonged to other MNOs.

In order to understand how the MVNO mayoperate the following describes how the GSMnetwork manages location registration, calls tothe mobile terminal (in-calls), and call initiatedby the mobile terminal (out-calls). Note that theidentity and number referred to below are relatedto the mobile subscriber and not the physical ter-minal. The physical terminal is identified withan international equipment identity not used forordinary communications.

Location UpdatingWhen the mobile terminal accesses an area witha new VLR code, it initiates a location updatingtoward the new VLR. The procedure is the samewhether the VLR is in the home network or in a

foreign network (visited network). The VLR up-dates the HLR of the MS so that the HLR knowsin which VLR the MS (mobile subscriber) iscurrently registered. The updating procedureconsists of several operations not important forthe context of this paper (authentication, alloca-tion of temporary identity).

In-callSince the ISDN number (or IP number) of themobile user is non-geographic, the only informa-tion an exchange in the ISDN/PSTN (or router inInternet) can derive from the called number isthe identity of the HLR of the MS. In order toroute the call to the current location of the MSthe network (PSTN/ISDN or Internet) requiresa network address identifying this access pointwhere the MS is found. The HLR then, by initi-ating a complex procedure toward the VLR,derives the addressing information the networkrequires for routing the call.

However, the capability of accessing the HLRis not generally implemented in the network.Therefore, the network routes the call to an MSC(or router) of the home network of the MS. ThisMSC then obtains the routing information andforwards the call to the MS. Sometimes thisrequires unnecessarily long connections. Hence,a network needs an MSC (or router) in order toserve in-calls.

Out-callThis case is simple. The MS just places the callvia the serving MSC (or router) that forwards thecall to the called user in the normal way.

The infrastructure that an MNO must have con-sists of both the Base Station Subsystem and theNetwork Subsystem. It has an own Mobile Net-work Code (MNC). In addition, it must have thenecessary equipment to issue and personalizeSIM cards.

An MVNO as its name tells, should not havethe complete infrastructure since it is supposedto be “virtual”. Indeed, an MVNO does not havethe Base Station Subsystem. It does not have alicense to use radio spectrum but has access toone or more. There is unanimous agreementabout what an MVNO does not have but there isno consensus about what an MVNO must have.There are different views, which go from“plump” MVNO to “skinny” MVNO [1].

Ovum [2] and ODTR, the Irish regulator [3]define an MVNO as an organisation that:

• offers mobile services to customers;

• has it own mobile network code;

5Telektronikk 4.2001

• issues its own SIM card;

• operates its own MSC and HLR;

• does not have its own radio frequency spec-trum.

The reason why this configuration is proposed isto provide an access for incoming calls as de-scribed above. The MSC then acts as the accesspoint to the MVNO defined by the internationalISDN/PSTN (or IP) number of the MSs of sub-scribing to the services of the MVNO. The con-figuration of an MVNO operating together withan MNO is shown in Figure 2.

The purpose of the “other platform” is to offerservices. This may be a LAN, an Internet, anintranet or any other platform. The platformmay, for instance, offer seamless internationalmobile extension to business LANs.

Note that the MSs of the MVNO is alwaysroaming in the network of other operators. Thismay lead to difficulties concerning the businessagreement between MVNOs and MNOs asdescribed in [5].

However, companies known as MVNOs such asVirgin Mobile in UK, Australia and Singaporedo have neither their own MSC nor their ownMobile Network Code but relies heavily on theMNO’s facilities. As stated in [1], a skinnyMVNO may rely almost totally on the MNO’sfacilities, and calls to and from its subscriberswould be treated as if there were calls to theMNO’s own customers.

We believe that the key issue lies on the percep-tion of the customers. An MVNO should be per-ceived by end users as a real MNO. OFTEL, theBritish regulator [4], states that subscribers’ con-tract with the MVNOs would cover all services;there would be no contract with the mobile oper-ator (hence the term “virtual” mobile networkoperator).

RevenuesTo survive an MVNO needs to be profitable.The main source of income derives naturallyfrom the customers. The customers pay for callconnection and services. The second source ofincome is the interconnect payment from otheroperators. In order to succeed an MVNO mustbe able to provide innovative services since thisis unique source of income that it controls tot-ally. Such services may be international exten-sions of LANs. Other services may be associatedwith software in the SIM offering capabilitiesbeyond mobility.

On the cost side, the MVNO has to pay to theMNO for both outgoing and incoming calls. Inaddition, it has to pay interconnect payments toother operators for completing outgoing calls.Last but not least, it has also investment, market-ing and operational costs. An MVNO needs toreach commercial agreements with an MNOconcerning the charging for using the radio spec-trum. There are currently two possible schemes.In the cost plus charging, the payment is derivedfrom the MNO’s actual costs. In the retail minuscharging, the payment is the MNO’s retail priceminus some discount. Usually, the retail minuscharging is higher and less advantageous for theMVNO than cost plus charging.

Figure 2 MVNO with MSC

IN

BSC MSC

MSC

EIR

VLR

HLR

Home network

Visiting network

MS

PSTN /ISDN

Internet

Other platform


References1 Gilbert & Tobin. The Emergence of the

Mobile Virtual Network Operator – VirtuallyHere. February 27, 2001 [online] – URL:http://www.gtlaw.com.au/pubs/morethanjustabigbrand.html

2 Matthews, J, Sweet, S. 2000. Virtual MobileServices: Strategies for Fixed and MobileOperators. Ovum.

3 ODTR Response to Consultation into Open-ing the Market for Third Generation MobileServices. December 2000.

4 OFTEL. 1999. Mobile Virtual NetworkOperators: Oftel inquiry into what MVNOscould offer consumers. Director General ofTelecommunications. http://www.oftel.gov.uk/publications/1999/competition/mvno06999.htm

5 Audestad, J A, Gaivoronski, A. 2001. Optionpricing of virtual network mobile operators.Telektronikk, 97 (4), 107–112 (this issue).

7

1 IntroductionThe purpose of this paper is to analyse differentMobile Virtual Network Operator (MVNO)business models in order to validate their sus-tainability.

The mobile telecommunication industry isfor the time being characterised by a lack oflicences, high investments in network building,and deregulation of old monopolies. There is anincreased complexity both in the market and inthe industry, and there is a search for new busi-ness models for expansion.

The mobile market is a market still in stronggrowth with many players that want to get aslice of the pie. Each country has a limited num-bers of available frequencies, and, therefore, canonly award a limited number of licenses. Asthere are a lot of players that want to be mobileoperators and only a few of them are able to geta license, both because of the limited number oflicenses and the huge cost of building a mobilenetwork, new types of players, MVNOs, haveemerged. The regulators have welcomed the newplayers because they want to enhance competi-tion for the benefit of the users.

An MVNO is a new business model, and facesseveral challenges, such as the obviously asym-metric agreement with an MNO. The strategicposition for an MVNO will depend on the originof the company. Possible origins are divided intothree groups: retailers, expanders and integra-tors. Each position has different strategic optionsand possible strategic advantages, which are dis-cussed in this paper.

2 DefinitionIn a high level description, a service provided bya telecommunication network is the result of anexecution of a function. Some functions go handin hand and cannot be separated either from abusiness or technology perspective. The smallestset of functions that cannot be further dividedconstitutes a role.

The “eco-system” of the telecommunicationsindustry consists of several roles and interfacesbetween them. The providers of access, switch-

ing, transport, service platform, billing, customerhandling, etc. each constitute a specific role. Aplayer – or a firm – on the other hand is a collec-tion of several roles.

It is important to make the distinction betweenplayers and roles. The traditional telecommuni-cation operators will often cover many or allroles, mostly due to their history with monopo-lies and vertically integrated value systems.Breaking up these systems has given rise tonew business players like resellers and serviceproviders.

Rather than defining the MVNO as one newrole, it makes sense to view an MVNO as a col-lection of roles. An MVNO could in principleundertake any role, but by definition it must,from a customer perspective, appear to be anMNO, having the same interfaces towards thecustomer as an MNO. In addition, the MVNOcannot hold its own licensed radio spectrum.

The question for an MVNO is how “deep” (withcustomer facing being the highest level, andradio access provisioning being the lowest) intothe value system it should go. The MVNO has toconsider which facilities to own and run, whichto outsource and which to lease from the MNOand what type of agreements it should seek withpartners in order to appear as an attractive choiceto the customers.

3 The Mobile BusinessLandscape – an AnalyticalApproach

A business model may be defined as an individ-ual company’s set of choices about what to doand how to do it. In this section, some frame-works used to evaluate sustainability of MVNObusiness models are described. Keywords arenetwork externalities, complementary services /products and value configuration.

Many various environmental aspects may be ofgreat importance. Focus of approach is businessstrategy, industrial organization and competitivepower, but other aspects such as underlyingneeds of the customer and government policy(regulation) may be crucial.

An Analysis of the MVNO Business Model*A N D E R S L I L L E H A G E N , L A R S A R M Y R , T E R J E H A U G E R ,V E G A R D M A S D A L A N D K A R I - A N N S K O W

Lars Armyr

[email protected]

Anders Lillehagen (44) is Advi-sor at Telenor ASA’s corporateRegulatory Affairs unit where heis engaged in all kinds of strat-egy and policy regulatory issuesincluding international regula-tions. He received hisCand.Oecon. degree from theUniversity of Oslo in 1986, hisMaster of Business Administra-tion from Manchester BusinessSchool, UK, in 1992, and hisMaster of Technology Manage-ment in Telecom Strategy fromthe Norwegian University of Sci-ence and Technology in 2001.He joined Telenor in 1997 asregulatory advisor in CorporateAudit and became manager ofTransport and Communicationsand the Norwegian CompetitionAuthority.

[email protected]

* The article is based on a paper by Lars Armyr, Terje Hauger, Anders Lillehagen, Vegard Masdal and Kari-AnnSkow.

Telektronikk 4.2001

Telektronikk 4.2001

The Value Chain, Value Net and Value Net-works frameworks are discussed. In addition, thedifference between effectiveness and strategyand increasing return are discussed as well.

Value ChainThe standard approach to the analysis of indus-try attractiveness is Michael Porter’s Five Forcesframework [1]. The attractiveness of an industry,such as the telecommunication industry, dependson the state of competition. Competition in anindustry is rooted in the underlying economicstructure of the industry. The state of competi-tion in an industry depends on five basic com-petitive forces. The figure below gives a pictureof the Porter’s “Five Forces” framework: Thepower of the five forces – Suppliers, Buyers,Potential Entrants, Substitutes and Rivalryamong existing firms – depends on some majorfactors and characteristics listed in the work ofPorter.

In his book ”Competitive Advantage” [4]Michael Porter suggested analyzing the “costleadership” and “differentiation” strategies bymeans of the value chain model, which hasbecome the standard approach to these analyses.An effective competitive strategy according tothis approach takes offensive or defensive actionin order to create a defendable position againstthe five competitive forces.

Value NetThe Value Net can be seen as a generalization ofthe Five Forces framework. Complementors areadded as a new dimension. The Value Netemphasizes that the value to the customers candepend on a package of complementary servicesand/or products. Logically, complementors willinfluence the attractiveness of an industry. Fig-ure 2 gives a picture of the Value Net of a com-pany.

Vegard Masdal (33) is ResearchScientist at Telenor R&D. Heholds a Master of Science(1991) in Electronics from theNorwegian University of Scienceand Technology and received aMaster of Technology Manage-ment in Telecom Strategy fromthe same institution in 2001.Since joining Telenor in 1994Vegard Masdal has worked withlarge-scale deployment of tele-com and computer networks.Main current interests are busi-ness models and business strat-egy.

[email protected]

Terje Hauger (52) is Advisor atTelenor Business Consultingwhere he is engaged in strategyissues and organisation devel-opment. He graduated from theNorwegian School of Economicsand Business Administration in1975 and received a Master ofTechnology Management inTelecom Strategy from the Nor-wegian University of Scienceand Technology in 2001. Hejoined Telenor in 1995 as Direc-tor of Quality and Security inTelenor Business Solutions. Pre-viously he has been working asmanagement consultant withCoopers & Lybrand and RailoInternational, and as researcherat the Institute for SocialResearch.

[email protected]

Suppliers

Entrants

CustomersExisting

competitors

Substitutes

Figure 1 Porter’s Five Forces

Competitors

Suppliers

ComplementsBusiness

Customers

Figure 2 The Value Net

8

9

A player is a complementor if customers valueyour product more when they have that player’sproduct than when they have your product alone.A player is a competitor if customers value yourproduct less when they have that player’s prod-uct than when they have your product alone.

To draw the map of the whole Value Net of acompany (function) of the telecommunicationindustry is more and more challenging due toincreasing complexity of relations between com-panies (functions) within the ICT business land-scape. The number of companies and types ofunbundled functions are increasing. Companiescan to various degrees be both horizontally andvertically integrated. In addition, companies canbe present in several countries (local markets).Further, the number of different technologies(e.g. access technologies), the number of prod-ucts and applications etc. is increasing. Varioussources of payment, e.g. from subscription, ad-vertising, and transaction, can also increase thelevel of complexity and the relations betweenroles and companies.

Nalebuff and Brandenburger [2] apply the ValueNet model to illuminate competitive and cooper-ative business strategies for the digital economyand see the relations between the companies as agame. A game has players and the added valuethat each player brings. A game has rules thatstructure the interaction between players. Thegame is also affected by perceptions, what theplayers believe. A game can only be changed bychanging one of the components. The term co-opetition, coined by Noorda, is introduced todescribe strategy in a business landscape of bothcompetitors and complementors. A company’sstrategy towards different types of players has tobe different. An example based on pricing andvalue to customers describe the importance ofdifferent strategies. If companies price theircompeting products independently, price willcome down as companies try to compensate forlower margin by gaining profit from increasedmarket share. The reverse situation is true forcomplements. The natural incentive is to setprice too high, which makes the total packagetoo expensive. Complementors should agreeto lower prices. This is why we see so muchbundling. It’s a way of bringing the price of acomplementary package down. According toNalebuff and Brandenburger, there are threeapproaches to the problem; Do it yourself, Forman alliance or Set up a proprietary business.

Value NetworksStabell and Fjeldstad [6] suggest that Porter’sValue Chain framework is only one of threegeneric value configurations – the other twobeing “value shop” – which describes a firmwhere value is created by mobilising resources

and activities to resolve a particular customerproblem (e.g. hospitals, professional servicesand educational institutions) – and “value net-work” – which may be used to model firms thatcreate value by facilitating a network relation-ship between their customers using a mediatingtechnology (e.g. telephone companies, trans-portation companies, insurance companies andbanks).

Fjeldstad [7] claims that the Value Chain model(and the related Five Forces model) do notreflect the strategic logic of mediators, and inparticular fail to incorporate positive networkexternalities both as a source of competitive ad-vantage and as a barrier to entry and innovation.

Positive network externalities introduce uniquestrategic challenges. A new service has relativelow value to its first customers; whereas cost istypically highest in the introductory phase. How-ever, because the value of the service is depen-dent on who else adopts it, it may be difficult totarget customers on an individual basis. Conse-quently, in many cases it is impossible to levy arealistic charge for the service or equipment inthis initial phase, leading to “give away strate-gies” seen in areas such as mobile telephony orInternet browsers.

Like the value chain, the value configurationmodel distinguishes between primary and sup-port activities: Primary activities are those thatcreate value for the customers, which in the caseof telecommunications are those involved inselling and providing a service. Support activi-ties are engaged to perform effectively the pri-mary activities of the firm.

According to Fjeldstad [7] the primary activitiesof a telecom operator’s value network are:

• Marketing and contract management, whichconsists of activities associated with invitingpotential customers to join the network, theselection of customers allowed to join, end theinitialisation, management and termination ofcontracts governing provisioning and charging.

• Service provisioning, which consists of activi-ties associated with establishing, maintainingand terminating links between customers, andbilling for value received. Billing requiresmeasuring customers’ use of the networkcapacity both in volume and time.

• Network infrastructure operation, which con-sists of activities associated with maintainingand running a physical and information infras-tructure. These activities keep the network inan alert state, ready to service customer re-quests.

Kari-Ann Skow

[email protected]

Telektronikk 4.2001


Support activities include product and processdevelopment (development and implementationof new services, reconfiguration of networkinfrastructure, development of new technologyand implementation of standards), humanresource management and procurement

Fjeldstad [7] denotes multiple firms contributingto the creation of a service as a co-productivevalue system. In such industries co-productiontakes place both between horizontally intercon-nected firms and between vertically layeredfirms. Unique structural properties of the tele-communication industry following the patternsof horizontal and vertical co-production are criti-cal to understanding the nature of competition.

Mediators co-produce value in a layered struc-ture in which the network service of one media-tor serves as a platform for a higher-level ser-vice. Several companies may serve the samecustomer, but at different levels. Fjeldstad [7]uses electronic payments over the Internet as anexample: Currently, the telecommunications net-work serves as the platform for the Internet ser-vice, which in turn serves as a platform for theelectronic payment service. The customer typi-cally subscribes to all three and they are usedconcurrently to produce the desired service, i.e.electronic payment. The banks mediate pay-ments by credibly exchanging transactions be-tween accounts, the Internet service providersmediate standardised packets between IP-add-resses, and the telecommunication operatorsmediate bit streams between phone numbers, sothat, together the firms co-produce the paymenttransfer service in a layered business system.In the simplest case, with only one firm at eachlevel, the firms supply the same customer withdifferent components of the service. The firmshave a strong impact on the value of each other’srespective services, but they are not, in the con-ventional sense, suppliers or customers of eachother. It is the common customer of all three net-works that typically pays each contributor sepa-rately.

The layered structure of co-production in media-tion renders the traditional concepts of supplierand customer relationships insufficient for de-scribing and analysing relationships betweenmediators.

Value configuration analysis is based on theassumption that competitive advantage is foundin the activities that a firm performs. Such firmsoffer value to their customers both through theaccess option and the actual use of services.Hence, cost and value must be associated withboth.

Fjeldstad [7] argues that mediation servicesoffered by value networks represent an extremecase of network externalities because the depen-dency among customers is the main productdelivered. The service of a value network mainlydelivers the customers’ opportunities to exercisethose dependencies. Size and composition of thecustomer base is therefore the critical driver ofvalue in the value network, and the value of theservice offered is affected by the characteristicsof the customers that join the network. Capacityutilization is closely related to scale and is both acost and a value driver, for while it may reduceunit cost, high capacity utilisation also mayreduce service levels.

Increasing ReturnThe term “increasing return” denotes a situationwhere mechanisms of positive feedback makethe leading player maintain his competitiveadvantage over the trailing player. Main charac-teristics of a situation of increasing returns arehigh upfront cost, network externalities and highswitching costs for customers. Decreasing returndenotes the assumption that products or compa-nies that get ahead in a market eventually runinto limitations.

The cost-structure of telecommunication net-works is normally characterized by high upfrontcost due to license fees and cost of equipmentsuch as switches, antennas, etc. However, whenthe network is built, the cost of producing oneadditional call is very low. Telecommunicationsnetworks have this cost structure of high fixedcost and low variable cost in common with otherproducts and services of the so-called “neweconomy” [3].

“Network externalities” denotes the fact that themain asset of a product lies not in the productitself, but in external factors such as how manyother people are using it, i.e. demand-side econ-omies of scale. One more customer to the net-work not only benefits that customer, but alsocustomers already connected since this impliesanother person to call to.

Positive feedback is a more potent force in thenetwork economy than ever before [3]. Positivefeedback is not only related to the demand side,but also to the supply side. The source of posi-tive feedback from the supply side is known aseconomies of scale in production: large firmstend to have lower unit cost. Both demand-sideeconomies of scale and supply-side economiesof scale have been around for a long time, butthe combination of the two that has arisen inmany information technology industries is new.The two sides feed back to each other makinga multiplication effect. The result is especially


strong positive feedback and a popular para-phrasing is “the winner takes it all”.

Switching costs encompass all the costs incurredby a customer in changing to a new supplier.One example of switching cost is cost relatedto changing telephone number. Switching costcontributes to lock-in situations.

Effectiveness and StrategyStrategy should not be confused with operationaleffectiveness. Strategic positioning means per-forming different activities from those of rivalsor performing similar activities in different ways.Porter [5] argues that robust strategies involvetrade-offs. A company must abandon or forgosome product features, services, or activities inorder to be unique at others. Further, strategydefines how all the elements of what a companydoes, fit together. The large numbers of MVNOs,which appear to be similar, most likely lack adistinct strategy.

4 A Generic Business ModelFigure 3 illustrates all roles necessary to providemobile communication services and the relation-ships between them. The model reflects the ver-tical layering in Fjeldstad’s [7] model describedin the preceding section.

The figure shows four different interfaces be-tween the customer and supplier roles: The salesrole takes care of subscriptions and service initi-ation (corresponding to Fjeldstad’s “marketingand contract management”), Service provision-ing comprises the continuous activities neces-sary to establish connectivity between cus-tomers, and corresponds to Fjeldstad’s category– apart from the fact that our model treats occa-sional interaction with the customer like invoic-ing etc. as a separate role (subscription manage-ment).

The roles in the lower half of the model corre-spond to Fjeldstad’s “network infrastructureoperation”. In order to describe a business modelfor an MVNO it is necessary to separate accessnetwork operation from the operation of thetransport network. To give room for differenttechnical configurations that an MVNO maychoose, we have also separated the switching/routing role. The service platform operator pro-vides functionality that is common for manyapplications (authentication, accounting, autho-rization, catalogue services, presence manage-ment). Service provisioning may include tele-phony services, messaging services, Internet ser-vices etc. Each service represents a distinct role,and different players may supply each of the ser-vices. The sales and subscription roles in ourmodel comprise administrative activities.

In a 2.5G and 3G world content will be ofgreater importance in mobile communication.Content provisioning is therefore included as aseparate role in the model. Content provisioningrelates to hosting of the content as well as to theservices provided.

The last role included in the model is that of sys-tem integrator. This role comprises activities thatintegrate content and services into a whole inorder to facilitate the customers’ use of the ser-vices.

According to our definition, a player acting as anMVNO must at least fill the sales and subscrip-tion management roles. At the same time, anMVNO cannot fill the access network operatorrole based on a licensed radio network (but mayoperate unlicensed radio access networks – e.g.WLAN). All other roles in the model may ormay not be filled by the MVNO.

Which combination of roles a particular playerchooses, depends on his present business and onwhich assets he brings into the mobile communi-cation business.

Figure 3 A generic businessmodel

Sales

SubscriptionManagement

ServiceProvisioning

Service PlatformOperation/

Hosting

ContentProvisioning

SystemIntegration

Switching/Routing

TransportNetworkOperatio

AccessNetwork

Operation


5 Analysis of SustainabilityOur hypothesis is that sustainability of theMVNO business models will depend on wherethe company setting up the MVNO business,originates from. We have grouped the compa-nies into three groups:

a) Companies originating from outside the Infor-mation and Communication Technology (ICT)Industry;

b)Companies originating from inside the tradi-tional telecommunication industry; and

c) Companies originating from inside the ICTindustry, but not as a telecommunication net-work operator.

The three groups are called Retailer, Expanderand Integrator respectively.

The RetailerCompanies grouped as “Retailer” providesmobile services to its own customers and canperform various roles such as sale, distribution,billing, and customer management. But the char-acteristics that differentiate the Retailer are thatthe Retailer does not enter the mobile industryfrom a position as a telecommunication com-pany or from industries of the communicationeco-system such as application provider or ISP.The motivation to enter the mobile industry canvary from obtaining a new “own brand” productin its portfolio of products to becoming a tele-communication company with own network ele-ments.

The Retail business will tend to be a low profitindustry due to the following main characteris-tics:

• The suppliers of the mobile network, theMobile Network operator (MNOs), have largenegotiation power due to the necessity of theinput, concentration (small number of poten-tial suppliers) and the fact that MNOs havetheir own business of mobile service provi-sioning (the MNO and the MVNO compete inthe downstream market). The MNO will mostlikely use the power to set condition that willlimit the MVNO’s flexibility of pricing – forexample through “retail price minus” condi-tion – i.e. the price structure and the margin.In addition, the MVNO will not be suppliedwith higher quality of the network servicesand degree of functionality than the MNOitself. The types of agreement (set of condi-tions) offered are limited, often only to achoice of elements from a standard agreement.

• Low entry barriers. If the MNOs give accessto their networks, they seem to be willing or

obliged by regulation to give access to every-body accepting their standard agreement andfulfilling a set of minimum conditions.

• High degree of rivalry due to low degree ofdifferentiation among the retailers, limitedvalue added and a relatively high number ofRetailers due to low entry barriers. Differenti-ation can be difficult to achieve due to thesame underlying quality of network and lackof complementary activities in the “eco-sys-tem” of the ICT, e.g. provision of applicationor content. The option left to the Retailer isprice as the mean of competition.

• The Retailer will hardly be in a position toenjoy increasing returns. Both supply-side anddemand-side positive feedback is linked to thenetwork and the MNO. Even if the retailerinvests in network elements and/or manages tobuild a club of customers faced with switchingcost, increasing returns benefits the largeroperator.

The key success factors for the Retailer to besustainable (earning a positive profit relativelyhigher than the rivals) seems to be:

• Differentiation through better value proposalto the targeted customer group given that thetargeted group is large enough. There aresome ways of differentiation such as productfeatures, service, product mix and brand-namereputation. Brand-name reputation is oftenmentioned related to MVNOs. The best wayto add value is building on the idea of com-mon interest among members of a communitywith common communications needs andrequirements. However, it is not easy to finda good example of an MVNO without otherengagement of the ICT eco-system, such asapplication or content. Virgin of the UK is anexample of company with a brand and with aproduct mix, but Virgin is a content provideras well. Uniqueness through trade-offs is diffi-cult to copy.

• Capitalising on owned facilities such as billingsystems, distribution chain, customer care sys-tem, group of customers etc. If the cost ofadding a mobile service product-line is mar-ginal, the MVNO can sustain low margin. Butthe margin will be smaller with the downwardtrend of end-user prices and of traditionallymobile services (voice) and of average rev-enue per user. Tesco of UK may be an exam-ple of a company with low added cost ofintroducing own brand mobile services.

• The agreement with MNO – value proposal tothe MNO. Both the MVNO and MNO wouldlike negotiated conditions that enhance its


individual profit possibilities and power in thevalue system. Logically, the MNO should pre-fer MVNOs able to increase number of usersand usage, complementary in market segmentsand user value to the MNOs own service pro-vider and with potential to cannibalise themarket share of the competing MNOs. TheMNO’s trade off is between loss in its serviceprovider business and gain in other businesslevels. What matters to the Retailer should bethe fit of negotiated conditions with its serviceconcept, not necessarily the dept of access intothe mobile network. To build and run a net-work with mobile switching centre, homelocation register and other network facilitiesrequires high capital expenditure – high trafficvolume to earn a profit – and higher compe-tence level which may be a source of disad-vantage compared to an experienced MNO.The MVNO’s freedom to pursue individualstrategies, such as pricing strategies, should beof high importance to the MVNO. The MNOand the MVNO may have common interest.

• Regulation. Without an obligation forced onthe MNO by the regulator, the MNO may notwish to give access to its network. However,the MVNO may be better off with an agree-ment based on free will of the MNO.

The ExpanderCompanies grouped as “Expander” enter newsegments or markets as a telecommunicationsnetwork provider, e.g. fixed network operatorexpanding into mobile, mobile network operatorexpanding into a new geographical market.

The purpose of the Expander will be to get abetter grip on the customer, to capitalize onresources such as competence and network facil-ities or to seek opportunities in new markets.The benefits and advantages to the expanderdepend on many factors, perhaps first of alladded value to the customers, change of powerand relationships within the Value Net and thecompany’s uniqueness and defend ability.Synergies is a key word, because setting up aMVNO in a new marked with none or limitedsynergies, the new MVNO business will tendto be a stand alone “Retailer”.

The key success factors for the Expander to besustainable (earning a positive profit relativelyhigher than the rivals) seems to be:

• Increased returns. The move of the companyshould increase positive feedback, preferablyboth from the demand and supply side. Inaddition, increased functionality and pricingshould increase switching cost locking in thecustomer.

• Increased value added to the customer makingthe customer pay more and/or increase num-ber of customers and/or usage.

• Influence the game of business within theValue Net in light of the value network of lay-ers of business landscape. The strategy of theExpander should be to build a defendablecompetitive position. The player should usethe MVNO business as a strategic move todifferentiate itself from the competitors –companies on the same layer performing thesame or substituting roles. The MVNO busi-ness can be used as a tactical move as aresponse to competitors with a broader port-folio of substituting roles, e.g. bundling ofvarious access technologies, or a move tobecome the broadest provider.

• Attitude of the MNO and regulation. TheExpander would most likely prefer an agree-ment giving him access to the unbundledradio-network and make him able to controlthe routing of traffic, control the location ofcustomer, provide technology roaming etc.The probability that the MVNO and the MNOwill have different interest is greater than inthe Retailer case. But Tele2 is an example of atelecommunications company with a MVNObusiness in Denmark, with access deep intoSonofon’s (the MNO) network. The relationto the MNO can be a barrier to the Expander’sbusiness model. See section above regardingthe Retailer case for discussion.

• Capitalizing on in-house resources such asnetwork elements and competence of buildingand running a telecommunications operation.

• Other advantages such as brand, ref successfactors of the Retailer case.

The IntegratorCompanies grouped as “Integrator” enters otherlayers of the value network, e.g. an applicationprovider or an Internet portal provider move intothe mobile service layer.

The aim of the Integrator will be to get a bettergrip on the customer, enlarge the market /increase the total value to customers, influencethe game between the companies of the value netto become relatively more powerful or to seekopportunities in new layers/markets. The bene-fits and advantages to the Integrator depend onmany factors, perhaps first of all the actual posi-tion in the business landscape (relations andpower within the Value Net), total added valueto the customers, and the company’s uniquenessand defend ability. Positioning through compe-tence is the key for an integrator, because thiswill give a unique position, and might change


the relations between the players, hence avoid-ing the new MVNO business to be just a standalone “Retailer”. The key success factors of theIntegrator to be sustainable (earning a relativelyhigh share of the revenue stream) seems to be:

• Secure and increase the grip on the users andsecure the channel to the users and the rev-enues. The grip on and value added to the cus-tomer can depend on roles of various layers,e.g. functionality of the Internet portal servicecan depend on information from the mobilenetwork and mobile terminal. A company ora partnership between companies performingroles of several layer of the Value Network(vertically integrated) can lock out other com-panies. The MVNO business can also be usedto reduce the power of the company(s) of themobile business layer, and in that way in-crease the power of its own core activity/layer.

• Increased returns. The move of the companyshould increase positive feedback, preferablyboth from the demand and supply side. Inaddition, increased functionality and pricingshould increase switching cost locking in thecustomer. The MVNO should not be evalu-ated independent of the whole operation ofthe company or corporate group.

• Increased value added to the customer makingthe customer pay more and/or increase num-ber of customers and/or usage.

• Other advantages such as brand, re successfactors of the Retailer and the Expander.

6 ConclusionThe value of telecommunication services andproducts to the customers depend on comple-mentary services and activities. The businessmodel of a company have to give the company(or the corporate group) a defendable positionin the net of various complementary and supp-lementary services provided by companies per-forming different sets of roles. In addition, thenet can also be viewed as a complex set of rela-tions between the playing companies and inter-ests. Further, the position of the company/corpo-rate group has to enable the company to defend– preferably increase – its share of revenue gen-erated by the whole Value Net.

The main key success factors are:

• Take advantages of increasing returns andcustomer lock-in.

• Stimulate high total value of the whole ValueNet. Aim to achieve a win-win situation withthe MNO and other partners.

• Powerful position in the whole Value Net dif-ficult to copy or bypass.

• Be in or manure into a position where a rela-tively large share of total revenues can betapped.

The analysis show that there may be a largenumber of sustainable business models depend-ing on where the company/corporate group camefrom, the relations to other companies in theValue Net and the characteristics of the influenc-ing business landscape. But the winner(s) in theICT business landscape will tend to be the com-panies and corporate groups that have businessmodels building on the above success factors.The business model of the MVNO business unitshould be assessed in relation to other businessunits of the company or corporate group.

In general, the Retailer tends to be a low profitbusiness. The possible business models of theExpander and the Integrator are difficult to gen-eralize, but high profit companies will tend tobuild on the above success factors.

7 References1 Porter, M E. Competitive Strategy – Tech-

niques for Analyzing Industries and Com-petitors. New York, The Free Press, 1980.

2 Nalebuff, B J, Brandenburger, A M. Co-ope-tition : Competitive and Cooperative Busi-ness Strategies for the Digital Economy.Strategy & Management, November-Decem-ber, 1997.

3 Shapiro, C, Varian, H R. Information Rules.Boston, Harvard Business School Press,1999.

4 Porter, M E. Competitive Advantage. NewYork, Macmillan, 1985.

5 Porter, M E. What is strategy? HarvardBusiness Review, November-December,1999.

6 Stabell, C B, Fjeldstad, Ø D. ConfiguringValue for Competitive Advantage: OnChains, Shops and Networks. Strategic Man-agement Journal, 19, 1998, 413–437.

7 Fjeldstad, Ø D. The Value System inTelecommunications. In: Eliassen, K A, Sjø-vaag, M (ed.). European Telecommunica-tions Liberalization. London, Routledge,1999.

15

1 IntroductionWith the strong and rapid growth in the mobilecommunications market reflected in the explo-sion of the number of mobile phones and thenumber of subscriptions, companies both insideand outside the telecommunication branch havebeen attracted to the mobile sector. Besides play-ing a passive role, i.e. investing in the mobilesector by acquiring stocks of mobile operatorcompanies, a more aggressive and perhaps moreprofitable strategy would be to participateactively by becoming a mobile operator. Thethreshold of becoming a mobile operator wasdramatically reduced with the introduction ofthe Mobile Virtual Network Operator (MVNO)

concept. Indeed, it is possible to be an operatorwithout having to invest in frequency spectrumlicense or in wireless network infrastructure andwithout having to build up competence in wire-less technologies. In this paper a financial studyof the MVNO concept is carried out. The goalsare to identify and to create the values for theshareholders.

2 What is a Mobile VirtualNetwork Operator?

Although there is no consensus about the defini-tion of MVNO [1,2] we choose to adopt the fol-lowing definition:

Why Should You Invest in MVNO?T I M O T H Y S T A L E Y , L O C H . K H U O N G A N D D O V A N T H A N H

In this paper we present an evaluation of the Discounted Cash Flow on Return on Equity of the Mobile

Virtual Network Operator (MVNO) concept, which is an emerging business concept in the telecommuni-

cation sector. An MVNO is an operator offering mobile communications services without owning a radio

infrastructure. The preliminary evaluation of Business Strategies, Designs of an MVNO, and Synergies

between an MVNO and a Mobile Network Operator has been examined. Various Critical Success

Factors have been examined within the context of a Business Case advanced by Ovum, and the

financial analyses conducted in this paper show sound financial results confirming that the MVNO

concept is deemed attractive from a shareholder’s return on equity investment standpoint.

Figure 1 MVNO in GSM architecture

Loc H. Khuong (44) has 15years experience in a SeniorManagement position at theVice-President of Finance andDirector levels. He is currentlyserving as assistant to the CEOof Chemical Lime Company. Heis a resident of Fort Worth,Texas, and an MBA and doc-toral candidate at Nova South-eastern University (Doctor ofBusiness Administration) – ABDscheduled for graduation in fall2002.

[email protected]

Tim Staley is a doctoral studentin the Business Administrationprogram at the Nova Southeast-ern University in Ft. Lauderdale,Florida. He also teaches tech-nology-based technology andbusiness programs at a localcollege in Dallas, Texas. He hasserved in various academicadministrative positions over thepast fifteen years. His currentresearch interests include state-gic management, organizationalchange and electronic commu-nications.

[email protected]

MSCAu C

HLRBR

IN

MSC

Au CHLR

BR

IN

BSC

VLR

SIM

ME

BTS

PSTN/ISDN

Base Station Subsystem Network SubsystemMobilephone

Mobile Network Operator

SIM Subscriber Identity ModuleME Mobile EquipmentBTS Base Transceiver StationBSC Base Station ControllerMSC Mobile Switching CenterIN Intelligent NodeVLR Visitor Location RegisterHLR Home Location RegisterAuC Authentication CenterEIR Equipement Identity Register

Mobile Virtual Network Operator

BTS

Telektronikk 4.2001

Telektronikk 4.2001

An MVNO is an organization that is perceivedby the customers as a regular mobile networkoperator, i.e. offering mobile communicationservices, but does not own a radio infrastruc-ture.

In order to understand how this could be possi-ble let us examine how the GSM system [9]works. As shown in Figure 1, the GSM systemcomprises three subsystems: the Mobile Station,the Base Station Subsystem and the NetworkSubsystem.

The Mobile Station (mobile phone) consistsof the Mobile Equipment (the terminal) and asmart card called the Subscriber Identity Module(SIM). The SIM contains the InternationalMobile Subscriber Identity (IMSI), which allowsthe Mobile Network Operator to uniquely iden-tify the subscriber. Indeed, The IMSI consists ofthree fields: the Mobile Country Code, TheMobile Network Code and the Customer ID.

The Base Station Subsystem is composed ofBase Transceiver Stations (BTS) and Base Sta-tion Controllers (BSC) and is responsible for theradio link.

The Network Subsystem consists of the Mobileservice Switching Center (MSC), the HomeLocation Register (HLR), the Visitor LocationRegister (VLR), the Authentication Center(AuC) and the Equipment Identity Register(EIR). Lately, the Subsystem may be strength-ened with an Intelligent network Node (IN) thatenables enhanced services such as Call forward-ing, Voice mail, etc. The MSC is a switchingnode providing connection to PSTN/ISDN orother networks. The HLR contains subscriberinformation along with the current location ofthe mobile, i.e. the current VLR. The VLR con-tains selected information for each mobile sta-tion located on its service area. The AuC assiststhe HLR in the authentication of the mobile sta-tion. The EIR is a database containing a list ofall valid mobile equipment (terminal).

In a mobile environment where the mobile sta-tions (and the users) are moving constantly, it iscrucial to determine their location in order todeliver services to them. The GSM network isorganised hierarchically. At the lowest level, agroup of cells corresponding to a BSC forms alocation area in which the mobile phone canmove without location updating.

At the next level, several location areas underthe same MSC/VLR form a service area. Whenthe mobile phone changes location area, it mustregister the new location to the current MSC/VLR but the HLR need not to be informed as

long as the mobile phone is still in the same ser-vice area.

A registration towards HLR is only necessarywhen the mobile phone moves to a service areabelonging to a new MSC/VLR. A locationupdate message is sent to the new MSC/VLR,which records the location area information andthen sends it to the subscriber’s HLR. The iden-tity of the HLR is resolved based on the IMSI,which contains the Mobile Network Code. Aftersuccessful authentication, the HLR sends a sub-set of the subscriber information needed for callcontrol to the new MSC/VLR and sends a de-registration to the old MSC/VLR. The HLR alsorecords the identity of the new MSC/VLR.

A Mobile Network Operator (MNO) must ownat least a Base Station Subsystem, a NetworkSubsystem and a SIM personification system(for issuing SIM cards to customers) in order tobe able to offer communication services to theusers (both its customers and visiting users)when they are within its service area. When thecustomer visits another MNO’s service area,there must be a roaming agreement between thevisited MNO and the home MNO that ensuresservice delivery to the customer. In this case thehome MNO pays the visited MNO for outgoingcalls and incoming calls initiated by or add-ressed to its customers (call origination and ter-mination cost). In return, it will get paid forincoming calls from other networks and will alsocharge its customers extra, both for call deliveryand call initiation at visited MNO.

According to our definition an MVNO does notown any radio infrastructure. Hence, it does notown any Base Station Subsystem. Consequently,it does not need to have a VLR since it is notcapable of receiving any visitor. However, itmay or may not own an HLR, an AuC, an EIR,an IN and a SIM personification system sinceoutsourcing is always a possibility. An MVNO’scustomer is always on visit to another MNO’sservice area. Some sort of agreement is neces-sary. Such an agreement could not be a roamingagreement since an MVNO can never return thefavour to an MNO that has offered services to itscustomers. It has to pay the MNO for both out-going and incoming calls. It will get paid forincoming calls from other networks but cannotget paid extra from its customers for call deliv-ery and initiation at visited MNO since they arealways on visit to some MNO’s service area. Atfirst glance, it may appear that an MVNO has avery weak business case. However, with a goodagreement with one or more MNO in one ormore countries, an MVNO may be able to offersubscriptions with competitive prices and qualityto its customers.


[email protected]

16


3 Business Environment ofthe Decade 2000

One of the most profound changes in businessis that there is a new definition of success. Busi-ness is intrinsically different in the new millen-nium from what it has been in the past. Compe-tition is becoming more intense as marketsbecome global [3]. The widespread use of com-puter technology is altering ways in which oneconducts their business. Competitive advantagesare becoming more difficult to maintain whilethe international community has reached ourbackyard and competition has reached an alltime high [10]. MVNO is one of the majordrivers of these changes. The 1990s saw anexplosion of telecommunication and the utilisa-tion of cellular phones. The first decade of thenew millennium will see a convergence of fixedand mobile services, and operators are keen tosee a full range of telecom services. Sincemobile spectrum is limited to spectrum con-straints, Mobile Network Operators may belooking to negotiate commercial agreementswith MVNO. MNO and MVNO can exploresynergies as well as differentiate their services toservice their customers [1]. The emergence androle of MVNOs will be essential in developingthe market where the MNO’s market share isweak and marginal, especially for an MNO dur-ing a market penetration phase.

4 Changing Nature andIntensity of Competition,hence Strategic Alliance

The envelope of competition is changing. Firmstoday should not deny the need to improve one’sperformance through greater performance andeffectiveness. Business competition from acrossthe globe; change brought to an organizationthrough technological advances, and price com-petition are all examples of just some of thechallenges facing businesses today [5]. Out ofthe five hundred (500) companies on Fortunemagazine’s 1990 list of the largest industrialcompanies in the United States, only 300appeared on the 1999 list. Some were displacedby other companies with higher total revenues.Others disappeared in mergers, acquisitions orbankruptcies [4]. Drucker (1990) states that theonly firms that survive are on the two extremes;those that set the standard and those that serve ina specialty area. The in-between position is nei-ther desirable nor viable [6]. There are also newand fewer leaders in telecommunications due toindustry consolidation. There is no longernational boundary for competition. The marriageof computing with telecommunications hasmade competition more intense than ever. Forexample, a know-how, specialty or scientificknowledge of an individual living in Russia canbe made available via Telecommunication andInformation Technology. The most obvious

change is the intensity of competition in almostevery facet of business [7]. The telecommunica-tions industry is using a technique referred to as‘scenario planning’ to explore possible futureenvironments and determine whether or not theirstrategic plans will be appropriate in the comingyears.

The very nature of competition is changing aswe move into a service economy and into theinformation age. The emergence of MVNOrepresents that change. The global village hasarrived for virtually everyone who sells goods orservices. Advances in manufacturing, changingpatterns of transportation, and social and politi-cal changes all contribute to changes in organi-sations. However, Mobile Network Operators’growth in telecommunication is not the onlydriver of change. It is the catalyst that underliesmany other drivers. The dramatic growth rateof the number of mobile phone users in the lastdecade cannot continue indefinitely in the nextdecades. As a result, Mobile Network Operatorswill be competing with each other for existing

Box 1 Valuation

A project, a company or an investment can be valued using different method-ologies. Assets can be measured by looking at its fair market value determinedat an arm’s length transaction. This process is however not applicable for anexpansion project like an MVNO whereby shareholders are required to invest alarge sum of money for capital expenditures plus incurring continuing expensesin a long run at a significant risk. The most common methodology used is a NetPresent Value of Cash Flows. This requires development of pro forma projectionusing realistic assumptions to arrive at Statements of Earnings and Cash FlowsStatements for future years (5 to 10 years). These streams of cash flows (Box 4)would be discounted at a cost of capital (defined in Box 2) to arrive at a NetPresent Value (defined in Box 3). This Net Present Value would represent thevaluation of the project, in this case, the valuation of the MVNO project. A pro-ject that yields a positive Net Present Value is deemed attractive, the higher,the better.

Box 2 Cost of Capital

Cost of capital is defined as the average cost that a company incurs in orderto raise funds. Typically a company would raise funds through debt or throughinfusion of shareholder’s equity, or most often through a combination of debtand equity. Cost of capital is defined as the after tax weighted average cost ofdebt (business interest expenses are tax deductible) plus the rate of return onequity desired by shareholders adjusted for risks.

Example: MVNO company A has a debt to equity ratio of 2 to 1. What thismeans is, for every 2 dollars of debt, there will be $1 of equity. Assuming that acompany can borrow long-term interest rate at 7 % per year and the corporateincome tax rate is 20 %. In addition, for shareholders to invest in MVNO com-pany A, they would require an after tax return rate of 15 %; otherwise they wouldinvest in a less risky project or a project of equal risk but with higher return.

In a simplified way cost of capital would be computed as follows:

Cost of capital = (75 % x (1–20 %) x 7 %) + (25 % x 15 %) = 7.95 % or roughly 8 %.


customers rather than for new customers in amature market environment and for marketdevelopment in third world countries [1]. MobileNetwork Operators may soon realize that astrategic alliance with an MVNO may be thenew industry norm to increase and to retain itsmarket share in a global market.

5 New Paradigm for BusinessIn addition to the new global competitive naturefor producers and the demands of customers andultimate consumers, competition is furtherchanged by the increasing demands of stake-holders [13]. Business success means meetingthe needs of all stakeholders: workers, suppliers,customers, the community at large, the share-holders and the regulators. The need for rethink-ing how one views the organization and how itachieves goals and objectives is currently under-way in the literature of business and manage-ment journals. Witness a new theory of organi-zational knowledge, or corporate epistemologyprovides a new way of interpreting that is gain-ing interest in journal articles [8]. Successfulexecution of strategies in today’s highly compet-itive environment requires a sharp focus on cus-tomer needs and on new ways of managing tal-ents and energies of all members in the organisa-tion [5]. MVNO could offer a competitive edgebeyond the short-term revenue, cost and profit-

ability issues for a Mobile Network Operator. Intoday’s increasing global competition, a constantfocus on the ever changing needs of the cus-tomers with fast paced technological changescould require an MNO to team up with anMVNO to attain a sustainable competitiveadvantage. Above all, it means managing forflexibility to meet changing conditions, flexibil-ity to take risks in order to serve the ever chang-ing needs of the customers who may be spreadout in a remote geographical areas.

6 Synergies versus ThreatsNew entrant MNOs would see the potential ofsynergies by partnering with an MVNO in enter-ing a new market where initial market entry costwould be high. In a typical medium size Euro-pean country, one third of total cost would berequired for establishing the Base Station Sub-system. Utilizing the current excess capacity ofan MNO would require the MVNO to pay foreach year usage fee without incurring the largeinitial outlay of cash and thereby reducing thefinancial risk exposure. This would be veryimportant for public companies whose debt toequity rating is very important in driving stockprices and shareholder’s value. Reducing theoutlay of capital investment would improve thecash flow position for a company in an aggres-sive growth position; sources of cash could bebetter used in more strategic areas. On the otherhand, being an MVNO requires significant out-lays of cash as operating expenses for marketingand business development purposes. Significantexpenses would be required to be incurred formarketing and advertising expenses; free airtimeand waiver of connection charges are standardcosts that are incurred with delayed benefitsbefore revenues can be generated. That is an areathat would be of interest to an MNO to rely onthe expertise of an MVNO. For example in UK,mobile operators spent $160 million in advertis-ing costs. Orange operator spent an average of$240 to $480 per customer while the investmentrequired would be $320 per customer. Thesedata indicate a long-term payback for an MNOunless synergies can be explored between anMNO and an MVNO [1].

The partnering spirit between a new entrant ormarginal MNO and that of an MVNO is quitedifferent than that of a market leader who maylook at an MVNO as a threat to their marketshare. A market leader would be unwilling tonegotiate a commercial agreement with anMVNO and resist vigorously any attempt fromregulators to regulate access to its market share[1].

Box 3 Net Present Value

Net Present Value is defined as the net sum of negative and positive cash flowstreams occurred during current and projected into future periods, discountedback into today’s value at the cost of capital (defined in Box 2). This is neces-sary to recognize the value that a dollar today is worth more than a dollar to-morrow.

For example, at a cost of capital rate of 8 % per year, the following would be thediscounting factor used to calculate a net present value:

Cost of Capital Year 1 Year 2 Year 3

8 % 1.000 .857 .794

$1000 in today’s dollar value, at 8 % cost of capital, would be worth $794 in year3 or ($1000 x .794).

Box 4 Cash Flows

Cash Flows are determined as the sources of funds (cash in-flow or cashreceipts) and uses of funds (cash outflow or cash expenditures). It is differentfrom Net Income or Net Earnings in many aspects, one of which, cash flows donot take depreciation expense or accruals of revenue or expenses into account.The reason that cash flows are used in the valuation of a project instead of NetIncome because cash flows represent a project’s ability to meet its obligationtoward its debtors and shareholders, regardless of depreciation rules or con-vention which may vary from company to company.


7 Definition of Shareholder’sValue and MeasurementYardsticks for MVNO(Present, Future)

Shareholder’s value is often defined as theReturn on Equity in terms of appreciation ofvalue of capital invested to compensate theshareholders for risks and opportunity cost ofmoney taking into account the inflationaryimpact on the value of money. The concept ofshareholder’s value creation is often associatedwith the long term building of residual value ofthe capital employed or the stock performance ofthe enterprise. Traditional accounting methodsof measuring the benefits of an investment in atelecommunication in an MVNO in creatinglong term shareholder’s value are often incom-plete and short term in nature.

Accounting measurement of an investment in aMobile Network Operator or an MVNO usingcommon measurement yardsticks such as returnon assets and return on investment are often mis-leading due to inconsistency and lack of unifor-mity varying from company to company. Differ-ent depreciation methods and depreciation lifeoften lead to misleading results. To compensatefor the shortcomings of the accounting yardstickof measurement, other methods such as presentvalue of money using the cash flow derived fromthe project are discounted at the hurdle rate. Inthe case of an MVNO, discounting the presentvalue of cash flow would not be sufficient dueto the long payback. It would require a carefulevaluation of the residual value beyond theexplicit forecast using a perpetuity method tovalue a MVNO company.

8 Success when it MeansSustainable CompetitiveAdvantage and Bottom Line

Various techniques exist to measure the benefitsof an MVNO including:

Discounted Cash Flow and Net Present Valuemethod. This method is widely accepted in theindustry practice. The synergies that can be har-vested between a fixed operator and a mobilenetwork operator need to be identified and quan-tified.

Critical Success Factors method devised by JohnRockart. Computer systems are developed tosupport a company’s business objectives. Criti-cal Success Factors need to be defined withinthe context of the MVNO’s business objectives.

Valuation methods often vary from one businesscase to another and depend largely on data avail-able. The most common methods used involve aweighted average on the three methodologies:

1 Discounted free cash flows at a cost of capitalto arrive at a Net Present value;

2 Comparable data by using data of currentpublic companies whose various multiple ofEBITDA, earnings and sales can be readilyavailable and computed;

3 Past transactions by relying on past transac-tions of public companies that were traded,acquired or merged with multiple of EBITDA,sales and net earnings.

Unfortunately, since MVNO is a new conceptwith no past or current history, the valuationmethod using the Comparable multiple methodor multiple of Past transactions cannot be used.The only method available that can be used isthe Discounted Cash Flow Present Value method

Pro Forma Income Statement/Cash Flow: Assumptions: Initial cash outlay investment: ($4) Million

In 000’s Year 1 Year 3 Year 6

Customers 15 38 93

Revenues $5496 $19,312 $43,079

Cost ($7,441) $19,182 $33,962

EBIT ($1945) $130 $3208

EBITDA ($1545) $530 $3608

EBIT: Earnings before Interest and Taxes; EBITDA: Earnings before Interest, Taxes and Depreciation

In 000’s Year 1 Year 3 Year 6

Customers 15 38 93

Revenues/customer $366 $508 $463

Cost/customer $496 $504 $465

EBIT/customer ($130) $3.4 $35

EBITDA/customer ($103) 14 $39

Table 1 Pro Forma IncomeStatement and Cash flow

Financial Results/Sensitivity Study:

After TaxCost of Capital 10% 12%

NPV $15 Mil. $11 Mil.

PAYBACK 5.2 YRS 5.2 YRS

DCF ROE 60 % 60 %

Financial assumptions: $4 million capital: 3 to 1 Debt to Equity ratio.Shareholder’s equity investment: $1 million USD; $3 million USD loanamortization over 6-year period at 7 % per annum.

DCF ROE: Discounted Cash Flow Return on Equity

NPV: Net Present Value

Table 2 Financial results


using a Business Case approach with assump-tions as advanced by the work done in Ovum –Virtual Mobile Services Strategies for Fixed andMobile Operators by John Matthews and SusanSweet [1]. These data and assumptions will beused in this paper with the largest unknown fac-tor as delineated as the amount that an MVNOwould have to pay an MNO. Since this is themost major cost for an MVNO and it varies fromarrangement to arrangement, the profitability ofan MVNO remains an uncertainty until an indus-try standard can be established. If the paymentrequired were the operator’s retail price minus adiscount, the MVNO’s cost would be signifi-cantly higher than if it is based on the cost plusbasis. It would be extremely hard for an MVNOto generate a profit if it is charged at retail priceminus a discount unless there is a very signifi-cant economy of scale. A reasonable assumptionin the business case is that retail price minus dis-count is roughly 50 % higher than cost plus. It istherefore important for an MVNO to be able toconvince that it can generate additional traffic,value-added services and market share for aphysical operator in a competitive environmentdue to its marketing skills.

The following are data and summaries takenfrom the Business Case in Ovum-Virtual MobileServices Strategies for Fixed and Mobile Opera-tors by John Matthews and Susan Sweet [1]:

Residual Value: The calculation of ResidualValue is essential to an MVNO valuation. TheBusiness Case assumption in the Ovum studyassumed a 10 % of Original Investment asResidual value in year 7. This assumption isarbitrary and does not follow a going concernentity concept from a valuation standpoint. Themodified valuation method as proposed in thispaper treats the Residual Value under a perpetu-ity method. This is extremely important andrequired given the infrastructure and nature of anMVNO. With an initial capital outlay assumed at$4 million dollars, once successful in acquiringfavorable airtime rate from an MNO, the growthrate of an MVNO can be tremendous and is lim-ited only to competition and market acceptancewithout requiring additional capital outlay. Forthis reason, in valuing an MVNO, a growth per-petuity method needs to be utilized which in this

case would be calculated as follow: Last yearCash Flow divided by the Cost of Capital [12].

Another factor to consider is the sensitivitystudy using a range of Cost of Capital assumed.A strategic buyer or a larger firm with strongborrowing capacity would assume a lower costof capital (for discounting Cash Flow purpose)while a non-strategic buyer or a smaller firmwould have a higher cost of capital. A range ofafter tax cost of capital (10 %, 12 %,) would beused in this paper for valuation purposes insteadof a fixed number as assumed by the Ovumstudy.

Discussion of Financial ResultsBased on the above results, the investment in anMVNO is deemed financially very attractivewith high risk since most of the value is realizedin the later years. It would take 5.2 years pay-back on a present value of cash flow; most of thenet present value is realized in the Residualvalue as the MVNO reaches its high potentialafter year 4 starting with 60,000 customers andreaching 93,000 customers at the outpost of theexplicit forecast in year 6.

A Discounted Cash Flow After-Tax Return onEquity of 60 % is considered attractive given therisks associated with an MVNO. A Net PresentValue varying from $11 Million to $15 million(net of Initial Capital Cash Outlay of $1 millionin equity, debt of $3 million fully paid off after5.2 years) is considered very attractive from afinancial standpoint, creating significant share-holder’s value.

Upside potential exists if customer growth canbe acquired faster than assumed during the ini-tial first three years or operating costs can be re-duced and value added service revenue growthcan be realized above the Business Caseassumption levels:

Major Revenue Assumptions:The starting point for these forecasts are theexisting Ovum forecasts (assumed to follow anS curve) for overall subscribers and revenues.

• Assumed 1/3 market of MVNO customers &traffic-country size of Portugal. This would be1.5 % of total mobile market;

• Call Revenues 65 % of Total Revenues;

• Interconnect Revenues 26 %;

• Value Added Revenues 6.5 %;

• Connection Revenues 3 %;

• Value added services 10 % of Call Revenue;

Box 5 Payback

Payback is defined as a measure of risk with the number of years ormonths required for a project to generate sufficient cash flow to equalthe original investment in capital expenditures plus any additional work-ing capital. The quicker the payback period is, the smaller the risk ofthe project to the shareholders or investors.


• Interconnection 40 % of Call Revenue.

Major Cost Assumptions [1]:• Payments to MNOs for call termination equal

to Interconnect Revenues;

• Payment to MNOs for call origination is 35 %of Call Revenues;

• Interconnect Payments equal to InterconnectRevenues;

• Churn rate 20 % – Effective corporate tax rate20 %;

• Operating Cost decline as a proportion ofRevenues as system grows, i.e. $1 millionplus 5 % Revenues;

• Customer acquisition cost: $150/customer.

9 Supporting BusinessStrategy and Vision

In order for an MVNO to be successful, basedon the Business Case assumptions above, it isessential that additional revenues can be derivedeither by expanding in niche markets, offeringadditional value added revenues and by reducingoperating costs and capital costs. Brand manage-ment is essential in offering additional channelsto market existing products and services [1]:

The following are Key Business Drivers of asuccessful MVNO:

• Implicit or explicit support from the publicand regulators;

• Demand from users for more services andcompetition to keep prices low;

• Support from MNOs who will benefit fromadditional channels to market airtime;

• Ability to acquire mobile market penetrationwith lower risk and lower capital;

• Strong marketing, brand management skills;

• Value added services;

• Exclusive Cost plus airtime purchasing agree-ment;

• Manage to eliminate resistance from MNOsseeing MVNO as a threat to their business.

In order for various business strategies to worksuccessfully, an MVNO has to be organized witha holistic business design and strategic approachtaking into account the interests of all the stake-holders.

10 Holistic Business Designand Strategic Approach

Business requirements change as competitionintensifies and as business and markets increasetheir geographic coverage. Business allianceschange the structure of business processes andthe information needed to support them [11].

Effective and efficient business processes arecentral to the success of every enterprise. Whena company grows significantly or when its busi-ness or its markets change, the field of telecom-munication must facilitate the changes requiredby business processes. It needs to consider thecriteria of maximum local control flexibility inoperations and ability to accommodate changes.Revolutionary change would require a completeredesign of a business process, in this case thecreation MVNO in partnering spirit with anMNO as a catalyst for growth of market share.The new design process is called holistic design.Some of the important aspects of the holisticdesign include:

• Instead of looking at one aspect at a time,design all facets of the process concurrentlyinvolving all aspects of telecommunication.

• Needs of all stakeholders must be taken intoaccount during process design.

• Support of business vision and strategy is aparamount design goal.

Overall, MVNO’s holistic design should reflectall the aspects of the business working in part-nering spirit with MNOs. It must support thebusiness strategy and make communicationchannels more accessible. As the telecommuni-cation global market consolidates into fewer andlarger players, market leaders form an oligopolymarket structure. The natural mergers and strate-gic alliance of physical operators and MVNO,the partitioning and rationalisation of marketshare from one country to another, a holisticbusiness design of the MVNO are tantamount tolong term and sustainable success. The synergis-tic of a holistic design is merely more than thesum of its parts [14].

12 Critical Success FactorsThe following are summaries of critical successfactors for an MVNO organization [1]:

• Understand brand management and consumermarketing.

• Mobile Network Operators with infrastructurein one country could explore a niche market asan MVNO in another market.


• MVNO’s agreement with an MNO needs tofind a win-win situation: Lower cost and addi-tional revenues.

• MVNO obtains support from regulators orlobby for regulations.

• Obtain support from smaller MNOs to findcommon synergies which may benefit fromadditional channels to market airtime.

• MVNO negotiates for cost plus pricing versusmarket price negotiation with MNO.

• Use existing investment to reduce capitalcosts.

• Leverage on existing front or back systems toreduce costs.

• Reduce cost with lower churn rate.

• Explore value-added services not reachable bytraditional MNOs such as providing channelsto market additional products and services.

13 ConclusionThe preliminary evaluation of Business Strate-gies, Designs of an MVNO and Synergiesbetween an MVNO and a Network Operator hasbeen examined. Given a 60 % Return on Equitywith 5.2 years payback on a $4 million USDcapital with 3:1 debt to equity ratio, the MVNOconcept as proposed appears to be financiallyattractive. Various Critical Success Factors havebeen examined within the context of a BusinessCase advanced by Ovum and the financial analy-ses have been conducted in this paper showingsound financial results and are deemed attractivefrom a financial investment standpoint. CriticalSuccess factors remain in the regulatory aspectsand/or a successful cost plus commercial agree-ment yet to be reached between an MNO andan MVNO. In summary, the creation of share-holder’s value is possible with low initial capitaloutlay but it has to be carried out in a holisticapproach taking into account the interests ofall the stakeholders. Upside potentials exist toreduce the operating costs and capital costs aswell as common synergies between an MVNOand an MNO in order to create substantial share-holder’s value. Future studies may focus ona more detailed level of cost assumptions andrevenues breakdowns of an MVNO operation.An update of a country specific of regulatorychanges would also be useful as a platform foran initial MVNO launch. Studies of cases wherean MNO in one country can operate as anMVNO in another country would also be possi-ble with certain non-compete agreement. Futurestudies involving cross equity ownership be-

tween an MVNO and an MNO can also be ex-plored to strengthen the strategic alliance ofthese two potentially complementary operations.

References1 Matthews, J, Sweet, S. 2000. Virtual Mobile

Services: Strategies for Fixed and MobileOperators. Ovum Ltd. (ISBN 1 9022566-37-8.)

2 Knott, P, Wilkins, D. 3G, MVNOs & Acqui-sitions: Opportunities for entering new mar-kets. Analysis Research Limited. (ISBN 1903648 00 9.)

3 Bettis, R A, Hitt, M A. 1995. The new com-petitive landscape. Strategic ManagementJournal, 16, 7–19.

4 Buzzell, R. 1987. The PIMS Principles:Linking Strategy to Performance. New York,The Free Press.

5 Camillus, J C. 1997. Shifting the strategicmanagement paradigm. European Manage-ment Journal, 1 (15), 1–7.

6 Drucker, P. 1990. The emerging theory ofmanufacturing. HBR, 68 (3), 94–102

7 Grant, R M. 1996. Toward a knowledge-based theory of the firm. Strategic Manage-ment Journal, 17.

8 Krogh et al. 1994. An essay on corporateepistemology. Strategic Management Jour-nal, 15, 53–71.

9 Mouly, M, Pautet, M-B. 1992. The GSM sys-tem for mobile communications. Palaiseau,France, Mouly & Pautet.

10 Porter, M E. 1980. Competitive strategy:Techniques for analyzing industries com-petitors. New York, Free Press.

11 Prahalad, C K, Hamel, G. 1994. Strategyas a field of study. Why search for a newparadigm? Strategic Management Journal,15 (5).

12 Rappaport, A. 1986. Creating Shareholder’sValue. New York, The Free Press.

13 Scott Morton, M S. 1991. The Corporationof the 1990’s. New York, Oxford UniversityPress.

14 Stewart, G B. 1991. The Quest For Value.New York, Harper Business.

23

The scope of the study included

• Access to service providers/airtime resellers;• Mobile call-by-call selection;• Mobile carrier pre-selection;• Mobile number portability;• Mobile call termination rates;• National roaming; and finally;• Access for MVNOs.

The current EU telecom directives do not clearlyaddress access to mobile networks. Of the aboveissues, only mobile termination rates are regu-lated at EU level. The ONP InterconnectionDirective imposes an obligation on mobile oper-ators with SMP on the national interconnectionmarket to have cost oriented interconnectioncharges. The other issues have so far been leftto the discretion of national regulators.

A new legislative framework for the electroniccommunications market is currently being dis-cussed in Brussels. Many issues are yet to besolved not least and the consequences the newframework will have for the mobile industryremain uncertain. For instance, whether or notthe new legislation would impose heavy cost ori-entation obligations on all mobile operators orregulation of international roaming prices. Inaddition, with the new, very broad definition ofaccess, national regulatory authorities would beallowed, where they find a specific market notto be competitive, to impose a whole range ofaccess obligations on operators with significantmarket power (SMP), be that on the fixed or onthe mobile market. It is also more than likelythat mobile number portability will becomemandatory. Finally, it cannot be completelyruled out that NRAs under the new framework,at least in theory, would have the possibility toregulate retail prices.

But pending the adoption and the subsequententry into force of the new framework, nationalregulatory authorities to a large extent remainfree to regulate the mobile market.

Going back to the findings of the study, wefound that out of the nine Member States cov-

ered, five had imposed obligations on mobilenetwork operators to deal with service providers/airtime resellers. First of all, high barriers toentry at the network level caused by the scarcityof radio spectrum, limiting the number of play-ers, have triggered regulatory action. The pres-ence of service providers is viewed as increasingcompetition in the provision of mobile servicesto end-users. And it is not good, from the pointof view of competition, that the possibility toprovide services should be restricted to thosefew operators with network infrastructure. Theregulator thus supports a split between the provi-sion of network capacity and services. Sweden isperhaps the country which has gone farthest inthis direction. The four UMTS licenses, whichwere handed out by PTS at the end of 2000, arefor the provision of UMTS network capacity“only”. The PTS does not in fact require thelicense holders to provide services over thosenetworks. In addition, mobile operators in Swe-den have since July 2000 (following an amend-ment to the Telecommunications Act) beenobliged to sell excess network capacity to ser-vice providers without their own networks. Arecent report of the PTS shows that the Swedishmarket has seen the emergence of a number ofnew players during the last 12-18 months. ThePTS says however that it is too early to saywhether this change in market structure is dueto regulatory changes.

In those countries where mobile serviceproviders are left to the market, the regulatorshave either

• not seen any reasons to intervene since com-mercial agreements are already in place; or

• fear the effects mandatory access for serviceproviders would have on network operators’incentives to invest in new mobile infrastruc-ture. And less investment means less innova-tion.

What about access to mobile networks throughcarrier selection? (either on a call-by-call basisor through pre-selection). In six of the countriescovered, regulators had imposed obligations on

Access to Mobile NetworksE L I S O L V A N G

A few months ago, Cullen International conducted a short study on national regulation of access to

mobile networks. The countries covered in the study were Sweden, Finland, the UK, Denmark, Ireland,

Netherlands, Germany, Italy and Spain. The study looked at a number issues on access to mobile net-

works and how regulators in those Member States covered had dealt with them.

Eli Solvang

[email protected]

Telektronikk 4.2001


mobile operators with SMP to offer call-by-callselection (CS) while carrier pre-selection (CPS)had been mandated in four countries. In Spainand Finland, the options are available for inter-national calls only. However, looking at thepractical implementation, the picture changes.With the exception of the UK (check), neithercall-by-call selection nor pre-selection has reallytaken off anywhere.

The regulators which have intervened argue thatmobile CS and CPS would increase the choicefor end users by allowing them to choose alter-native carriers for outgoing calls. CS and CPScould also encourage new packaging of fixed/mobile services. On the other hand, regulatorswho have, at least so far, refrained, say that CSand CPS are more suitable tools for breaking upthe fixed market where there normally is onlyone access network available. Also these regula-tors would consider regulating mobile CS andCPS if other types of intervention fail to deliver.

It is probably also fair to say that the incentivefor regulators to impose obligations on indirectaccess to mobile networks has declined sincethere are no signs that the new EU legislationwill cover it.

Mobile number portability is not a form ofaccess since it only can take place between oper-ators already established on the market. Still, themajority of regulators have mandated mobilenumber portability. Statistics show however thatwhere it has been implemented, the effects so farhave been insignificant even in those countrieswhere the churn (the customer changes operator)is high. Mobile number portability is more thanlikely to be mandated in the future EU legisla-tion.

As was noted in the beginning, the Interconnec-tion Directive imposes an obligation on mobileoperators with SMP on the national interconnec-tion market to establish cost oriented intercon-nection charges. During the last couple of years,many regulators have been keen to scrutinise themobile termination charges for fixed to mobilecalls. This has in many cases led the regulatorto set a maximum call termination charge. Thisrather intense activity shows that regulatorstoday see mobile termination as a real bottleneckputting off competition, hence the inclination tointervene.

Recently, Oftel, the UK regulator, published aspectacular determination to strengthen the regu-lation of mobile termination rates. At present,

BT Cellnet’s and Vodafone’s termination ratesare subject to a price cap. But Oftel is proposingthat for the period March 2002 to March 2006,the termination rates of all four mobile operators(i.e. also Orange and One2One) should be sub-ject to a price cap. It is only in Austria where ithas happened before that also non-SMP opera-tors are being subjected to the cost orientationrequirement.

Mandatory roaming between two 2G networksin the same country is relatively rare in the EU.Only a few countries have imposed compulsory2G national roaming. One of the main reasonsbehind this hands-off approach is that the 2Gnetworks in many countries have roughly thesame geographical coverage. But the story willbe a different one now that mobile operators arestarting to roll out their 3G networks. Of thecountries covered in the study, 3G-2G roaminghas been mandated everywhere except in Ger-many.

The rationale behind this by and large parallelthinking is the idea of creating a level playingfield for 3G new entrants and existing operatorsfrom the beginning. To achieve this, the newentrant must be given the right to use the net-work by the 2G operator while it is building outits own network. In most countries, thresholdshave been set, for example that the 3G newentrant will only have the right to roam on exist-ing networks after having covered a certain geo-graphical percentage. In addition, 3G-2G roam-ing is generally required as a temporary, or “sun-set” measure, to alleviate the initial inability ofa new entrant to provide nationwide mobile ser-vices.

The last issue we looked at was the right ofaccess for mobile virtual network operators(MVNOs). The basic principle behind theMVNO concept is that the MVNO has not beenallocated spectrum on its own and so must usethe radio access network1) of one or more exist-ing mobile networks to offer services to its owncustomers. The MVNO may choose to provideall other network elements itself2) or maychoose to rent nearly all network elements froman existing network operator. In the second sce-nario the MVNO’s ability to differentiate itsown services from the network operator’s ser-vices is reduced and the MVNO resembles atraditional service provider.

It is probably right to say that the lack of a clear-cut definition of MVNOs has complicated regu-lators’ task on how to deal with MVNOs. It was

1) Radio spectrum and mobile base stations.2) Including the long-distance routing of calls, running its own user location register, customer billing etc).


the type of access that Sense Communicationsrequested from the Scandinavian mobile opera-tors’ networks back in 1998 which launched theconcept. Although Sense did not succeed backthen, the attempt was useful because it forcedthose regulators to think in new terms.

Out of the nine countries we looked at, regula-tors in five countries have attempted to clarifythe rights of MNVOs. The approach has takenvarious forms. In Denmark, for example,MVNOs have rights of access through thenational roaming legislation. In Sweden, mobileoperators have an obligation to sell excess net-work capacity, which could enable MVNO ser-vices. In Spanish regulation, the MVNO hasexplicitly been introduced as a new category ofoperator. In the UK however, where OFTELchose not to regulate rights of MVNOs, severalcommercially negotiated MVNO agreements arein existence.

Two opposing arguments are repeatedly put for-ward in the MVNO debate. It is possible that oneeffect of mandating access for MVNOs could beto hamper mobile operators’ incentive to investin new infrastructure. It is also possible that theexistence of more market players could generatemore traffic over the mobile networks. Lowerbarriers to market entry for MVNOs could have

larger benefits for consumers than traditionalservice providers because they are in a betterposition to differentiate their services from thoseof the mobile operators’, also in terms of prices.But regulators may also ask themselves, howmany players are necessary in the end to guaran-tee effective competition? Two or three is notenough, but what about four or five? And per-haps 3G new entrants will have an incentive tonegotiate MNVO type agreements on a commer-cial basis to maximise traffic on their networks?The regulator may also sit back and wait for theeffects of other measures before taking steps tomandate MVNOs.

The situation will inevitably differ from countryto country, depending on how the regulator hasdealt with other forms of access, the number ofmobile licences, licence fees, availability of fre-quencies, degree of competition in the wholesaleand retail markets etc. It is tempting to say thatthe safest way for mobile operators to preventMVNO obligations is to negotiate commercialdeals. But then the problem of non-discrimina-tion could arise. Why would a mobile operatewant to negotiate an innovative deal with a ser-vice provider if it would have to, subsequently,to offer the same terms and conditions to all theothers?

Telektronikk 4.2001

1 IntroductionThe intention of this paper is to define differentMVNO models and discuss the related serviceopportunities. Ovum has analysed the termMVNO in a separate report [1]. From the man-agement summary the following description ofan MVNO is used:

• Offers mobile services to customers;

• Has its own mobile network code (MNC);

• Issues its own SIM card;

• Operates its own home location register(including mobile switching centre as anoption);

• Does not have its own radio frequency (spec-trum) allocation.

This interpretation of an MVNO is somewhatlimited to GSM technology entities. We will inthe following use a looser definition:

“An MVNO is an entity which provides mobilecommunication services to customer under itsown brand, without a radio license.”

Based on this definition, further referred to asour MVNO definition, we have identified vari-ous shades of organisations, which might also becalled MVNOs and investigated the possible ser-vice opportunities for the different models.

1.1 MVNO ModelsFigure 1 shows to the left Resellers, Mobile Ser-vice Providers and the Ovum defined MVNOs(SIM/HLR with MSC option) that exist in theGSM-sphere. To the right we have introducedthe GPRS extension, Mobile IP and WLAN IPzones together with an IP backbone to get abroader picture. The seven alternatives based onthis figure all need to interconnect with one ormore Mobile Network Operator (MNO) and/orIP backbone provider to produce mobile servicesto the customers. Further the MVNO will possi-bly have a relationship with content providers,application providers, Internet service providers,etc.

2 MVNO Possibilities andConstraints

As for all market players, it is important for anMVNO to be able to provide added value rela-tive to what is already offered in the market.This is important in relation to the customer, butalso in relation to the MNO (mobile networkoperator), in order to achieve the best possibleMNO agreement. In this paragraph, four differ-ent aspects for creating value-adding opportuni-ties are suggested. The value adding opportuni-ties for a Mobile Service Provider is discussedseparately in paragraph 3.

Sub-paragraph 2.1 deals with the extent of inte-gration with the MNO’s network (i.e. verticalintegration), sub-paragraph 2.2 discusses possi-ble WLAN deployment, sub-paragraph 2.3 dis-cusses possible service opportunities with theGPRS extension and the scope of sub-paragraph2.4 is to present possible technology roamingconcepts.

2.1 Integration with Mobile NetworkOperator

Our MVNO definition, as presented in para-graph 1 above, focuses on the customer percep-tion rather than the degree of equipment owner-ship. All the four roles indicated in Figure 1(Reseller, MSP, MVNO1 and MVNO2) complywith such an MVNO definition, but the degreeof equipment ownership (i.e. technical integra-tion with the MNO) still governs which servicecapabilities the MVNO can offer.

In the following we have chosen to look at twoof these cases, the Reseller and MVNO2 (con-trolling both MSC and HLR), because these arethe extremes in terms of technical integrationwith the MNO. In the table below, some MVNOfeatures and their dependability of the choice ofMVNO model have been indicated. The tablehas been limited to the features that differentiatethe models.

The two models and their associated characteris-tics are described in somewhat more detail in thefollowing:

Service Opportunities for RealisticMVNO ModelsE S K I L D A H L E N , S T E I N T R O N E N G , J A N - T O R E D E I L K Å SA N D S T E I N A R L Å G

Stein Troneng is assigned asManager of a Switching Plan-ning department in Telenor Net-works. His formal educationcomprises Master of Sciencefrom NTNU (1993) and Masterof technology management intelecom strategy from NTNU(2001). Stein Troneng has heldvarious positions within TelenorNetworks and Telenor Interna-tional. Areas of experience com-prise SDH systems, IntelligentNetworks, switching and gen-eral project management.

[email protected]

Eskil Dahlen (37) graduatedfrom the Norwegian Universityof Science and Technology in1988 and holds a Mastersdegree in Technology Manage-ment, Telecommunication Strat-egy (2001). He is now workingas a Senior Project Manager inTelenor Plus. Activities includedevelopment of content andcommunication services forbroadband access and thirdparty access to the telecommu-nication platform.

[email protected]

26

27

2.1.1 ResellerThis model requires a “turn-key” traffic contractwith an MNO. Without control over any networkelements, the service capabilities are governedby that MNO. As shown in the table, in mostrespects, this can be considered a limiting con-straint for the reseller. However, there are obvi-ous advantages of sitting on the MNO’s back aswell; all roaming and interconnect agreementsare already in place, all the MNO’s services areavailable etc. For this reason, the reseller roleshould not require any significant amount ofmanpower on operational nor technical matters.

Also, as discussed in paragraph 3 below, thereare many other ways for the Reseller to addvalue. The reseller role could suite major playersin different (non-telecom) markets with an estab-lished brand, already populated customer data-base and existing customer relations. This couldgive a good negotiating position towards theMNO, which again may result in a cost advan-tage. The Reseller model is seen as a possibleway into the mobile market for e.g. fixed net-work operators and ISPs.

2.1.2 Owning/Controlling MSC and HLRIn GSM this would equate to owing your ownHLR and MSC. Having control over HLR,makes it possible to provide value added ser-vices based on HLR contents such as locationdata, logon status etc. Also, the options arewider (at least technically) with regards to spe-cial roaming arrangements, whether it is nationalor international roaming or roaming between dif-ferent access technologies.

An MVNO of this type may build and operate itsown backbone network in order to reduce costson international traffic. This makes the modelattractive for players that already control fixednetwork resources. On the other side, this modelalso requires more from the MVNO, in that itmust make roaming/interconnect agreements onits own. To own the MSC makes it possible toprovide value added services based on the callitself, e.g. IN solutions w/voice-prompt-menus,call routing, etc.

However, it is important to note that an alterna-tive approach to own and operate the MSC andHLR is to negotiate access to and use of the rele-vant HLR-info and MSC functions with theMNO. This could provide a cost-efficient alter-native, and potentially make the reseller modelalmost equally powerful as the MSC+HLRmodel, also when it comes to service capabilities.

2.1.3 ConclusionIt is more useful to use a loose rather than strin-gent MVNO definition with regard to equipmentownership. What is important is whether and

how a particular player without a radio licence,by providing mobile communications, can addvalue for customers and partners. The Resellermodel allows for a lot of value adding opportu-nities, even though service capabilities to a greatextent are governed by the MNO. TheMSC+HLR model allows for more freedomthan the Reseller model with regards to servicecapabilities and more flexibility with regards toroaming and routing arrangements

2.2 GSM-GPRS ExtensionWith the introduction of GPRS (General PacketRadio Service) in the GSM network it will bepossible to have always-on access to Internetand other data/IP services. The main drivers forthe introduction of GPRS is:

• Efficient use of radio resources;

• Dynamic allocation of radio resources, (allo-cates channels only when data are transmitted);

• GPRS is designed for data and WEB/Internetapplications – always on;

• Extend data-rates up to 40 kb/s in the firstphase and up to 80 kb/s in later implementa-tions;

• Offers billing on volume;

• Quality of service is specified in profiles todifferentiate on customer-services.

Three different classes of mobile stations aredefined:

• Class A: Supports both GPRS and circuitswitched services, independent from eachother. This is the most complex version.

• Class B: Supports both GPRS and circuitswitched services, but not at the same time.It shall be possible to shift between the twoservices automatically. A packet transactioncan be put on hold while an incoming speechtransmission is handled and then continuedwhen the speech is finalised.

• Class C: Supports both GPRS and circuitswitched services, but not at the same time.A C-terminal could be realised as a packetdata terminal only.

2.2.1 MVNO OpportunitiesThe equipment needed to operate a GPRS ser-vice will be more appropriate for introducingmobile Internet services. It is also believed thatthe interconnection towards the MNO’s GPRSsystem will be more open and flexible than theMSC interface.

Steinar Låg (33) was, at the timeof writing this article, student inNTNU and Telenor CorporateUniversity’s Master of TelecomStrategy Program 2000–2001.Since receiving his Master ofScience in Electronic Engineer-ing from NTNU in 1993, he hasbeen with Nera, Telenor andInmarsat (London), working withsystem and network aspects ofmobile satellite communicationssystems, with focus on systemdesign and development ofInmarsat services. Låg is cur-rently employed by TelenorSatellite Services, working onsystem and product develop-ment as well as future strate-gies.

[email protected]

Jan-Tore Deilkås (41) obtainedhis Master degree from the Nor-wegian University of Scienceand Technology in 1985 and hisMaster of telecom strategy in2001. His main work experienceis within Telenor Networks, theinfrastructure division, with afocus on introduction of newtechnology in the acccess- andtransport network. His currentwork is within the developmentof the access network and inparticular ADSL and VDSL sys-tems.

[email protected]

Telektronikk 4.2001


The GPRS subscription will be an add-on in theMNO’s services provisioning portfolio. For anMVNO this can either be seen as an extensionfor the services offered in a vertical MVNOagreement (Reseller or MSC+HLR) or viewedas a possible stand-alone entrant strategy intothe mobile business.

As a Reseller the customers will normally usethe HLR of the MNO to get access to the GPRSservice. The first optional move for an MVNOcould be to negotiate a separate GPRS roamingagreement with other operators. You would thenneed an IP interconnection point, a MNC and aHLR. Further it should be possible to negotiate aseparate GPRS interconnection agreement withyour MNO, either to the GGSN (Gateway GPRSSupport Node) or if possible deeper into theMNO’s GPRS network with introduction of your

own SGSN (Serving GPRS Support Node). Thisleave you as an MVNO, having a Reseller con-tract for the GSM part and an interconnectagreement for the GPRS part.

Initially the GPRS terminals will be of Class B.This means that the customers will not be ableto handle circuit-switched services (voice) andpacket transactions (e-mail, browsing) at thesame time anyway. This split in operation willprobably mean that the customer will feel nodegradation on the service delivered from suchan MVNO.

This suggestion will be very interesting whenthe focus of your operation is mobile Internetservices bundled with commodity services likevoice and SMS from the GSM world. Furtherthis will probably give you a better position if

Notes:

1 “No freedom” in the meaning “same as MNOM.

2 In the MSC+HLR case, the opportunity to control/own the backbone networks also exists.

3 It is possible for the reseller to support technology roaming if assuming a proprietary mobility management system(i.e. external to GSM network). However, the MSC+HLR model is more suitable because of the option to modify theHLR.

4 Value added services using HLR info such as logon status, position information, etc.; whereas value added servicesbased on MSC could be IN-solutions, voice prompt menus, call routing logic, etc.

Feature / Model Reseller MSC+HLR

HLR and MSC Does NOT own Does own

GSM roaming No freedom1 Possible

Technology roaming Possible3 Possible

Call routing / interconnect No freedom1 Flexibility2

Figure 1 The networkenvironment and possibleMVNO alternatives

MVNOalternatives

Con-tent

HLR

Service andcontent provider

SIM cards, branding,price differentiation

MNO 1

IN (VAS)MVNO 1

groove, banking,news, etc.

Content provider

MVNO 2

MSCGPRS GW Radius (AAA)

GPRS extension

InternetService Provider

Mobile IPprovider

Homeagent

IP zonesintegrator

IP backbone(part of MNO or

stand-alone)

IP zone

GGSNSGSN

Laptop

Fixednetwork

Data networks(ATM, FR, Intranet, LAN)

PDA (Personal Data Assistant)

Satellite

Mobile phone

BTS

BSC

Mobile ServiceProvider

Mobile operator

Network and Accessprovider

Terminal

Reseller

IN (VAS)Op. and maint.

MSC

IN (VAS)

HLR

HLR

HLR


you want to negotiate a complete interconnec-tion for all your services with the MNO. AGPRS interconnection will also be importantfor technology roaming solutions.

2.3 IP Zones IntegratorA business that wants to operate as an MVNOhas an option in offering IP zone services inaddition to the GSM/GPRS services. The fol-lowing sections describe what an IP zone is andwhy IP zone offering is an interesting option.Challenges are how to quickly obtain sufficientcoverage (to attract customers and to achievea minimum system capacity), how to make aviable business model and how to defend againstcompeting players. An IP zone is a term used fordescribing a geographical area that is covered bya Wireless Local Area Network (WLAN) com-bined with a connection to an IP network, mostoften the Internet.

2.3.1 WLAN as IP ZoneThere are several technologies that can supportwireless Internet access. A promising technologyis referred to as WLAN. One could argue thatWLAN is a generic term, but it is very closelyassociated with the IEEE-standard 802.11.WLAN is a semi-mature technology that headsUMTS by several years when it comes to de-ployment. In addition the WLAN operates inan unlicensed radio spectrum.

The beauty of WLAN technologies is that onecan leverage on years of research and develop-ment in the fixed LAN-environment. Thus pro-cedures for user Authorisation, and Authentica-tion are already established. From a functionalpoint of view there are very small differencesbetween a wired LAN and a wireless LAN. TheLAN environment has no tradition for chargingthe end user. The cost associated with operatinga LAN has been a part of the company’s over-head. For this reason there is no establishedbusiness model for WLANs that can be lever-aged on. Thus the major challenges for a com-pany that tries to offer public WLAN-services isto build the right business model to support theoperation.

2.3.2 Relevant Players in the IP ZoneContext

A complete IP zone service delivery involvesmany roles, technically and business-vice. Thebelow mentioned roles are a minimum of whatis necessary for a complete service:

• Site owner: The site owner has property rightsover the infrastructure (houses, buildings, are-nas ...) in which the equipment is to be placed.

• Equipment owner: Somebody has to providethe necessary technical equipment (radio inter-faces, antennas, hubs, routers, etc.).

• Access provider: To connect the IP zones weneed a an access line from the site to the back-bone network.

• IP backbone provider: To interconnect the IPzones and to provide Internet access we are inneed of an Internet backbone provider.

• ISP: An ISP will deliver some Internet ser-vices, but most of all he is the business roleresponsible for giving out IP addresses to con-nected terminals.

• In addition to the above examples there is aneed for a lot of functionality (security,accounting, billing, authorisation).

To make all this work, one can either take on allthe roles oneself, or one can act as a clearingfunction for all the parties involved. Our chal-lenge is therefore to find a role as an integratorand/or clearing house that has a long-terminvolvement.

2.3.3 Value for an MVNOFor IP zones operation to make sense to anMVNO there has to be value in it not only forthe customer, but also for the MVNO and allother relevant players.

Negotiation Power – Entrance StrategyBy definition an MVNO does not hold its ownlicensed radio spectrum. Thus an MVNO needsan agreement with the (MNO) to be able to offerservices. The MNOs have traditionally beenreluctant to offer access into their networks. Oneobvious reason is that there is a large asymmetrybetween the MVNO and the MNO. The MNO’scustomers will under no circumstances send traf-fic into the MVNO’s network, it is always theother way around. If an MVNO can offer IP con-nectivity in selected areas to the MNO’s cus-tomer, the MNO is more likely to be willing totrade accesses. Thus an entrance strategy fora company could be to start out as an IP zoneoperator. After building an initial customer basethe company could expand it’s service portfolioto become an MVNO. The advantage of thisincreased negotiation power should not beunderestimated. The biggest obstacle forMVNOs so far has been to negotiate fair dealswith the MNOs. To bring this kind of negotia-tion power could be a significant competitiveadvantage compared to other MVNOs.

Value for the CustomerAn MVNO is more attractive to the customerwhen complementing GSM/GPRS connectivity


with a high-speed WLAN service. This is espe-cially relevant when the user is moving within asmall space (offices, airports, rests, ferry-portsand other mandatory waiting areas). WLANplug-in cards are currently available, and can beused with laptops as well as PDAs, which areterminal types more suitable for business workand data applications than GPRS telephones.Following from low investment costs, a lowprice for the end user is anticipated, probablymuch lower than GPRS and UMTS.

Value as a Stand-alone OperationRunning an IP zone operation could very well bea good business in its’ own right regardless ofthe increased negotiation power. This businessopportunity of IP zones should be investigatedas stand alone operation as well as value add toan MVNO-operation.

2.3.4 ConclusionIntroducing IP zones as a value added service inthe portfolio of a MVNO, could be useful.

• It will increase the power in negotiations withthe MNO and hence lead to a better deal.

• It will give the customer the experience of abetter service, both on price and bandwidth,in the IP zone areas.

• It may be the feature of an MVNO that makescertain target customer segments willing totake on a MVNO subscription, due to specificneeds.

2.4 Technology Roaming

2.4.1 Roaming between Different AccessTechnologies

Roaming is in telecommunication most oftenassociated with enabling user access, serviceprofile and charging capabilities across homoge-neous networks. The GSM system is an exampleon a well performing roaming capability on arather high level. Subject to roaming agreementsbetween peer mobile operators, the users canroam between the mobile access networks withmost service capabilities intact.

The roaming logic is implemented in the net-work entities and the users do not have to beaware of which underlying mobile operator isenabling the services. The handset is frequencycompatible in the visited network and access isgranted through authorisation and authenticationprocedures based on the SIM data provided.However, today’s pricing regime in Europeoften charge the end users specially roaming feesand premium calling rates when roaming in aforeign network.

Roaming between different access technologiescould be somewhat more complicated than in theGSM case. First of all the basic user functional-ity has still to be catered for:

• Authorisation;• Authentication;• Service provisioning;• Accounting.

But maybe even more basic when consideringdifferent networks is the physical interfacebetween the user handset and the access net-work. In the GSM case it is implicitly given thatthe user is roaming with the same handset.

One roaming alternative is user roaming (typicalSIM card roaming) where the user is able to in-sert his/ hers user profile in a different handsetwhen roaming in a foreign network. This is thecase when Europeans travel to US (or viceversa) and use their personal SIM card in a ter-minal compatible with the local frequency stan-dard.

A more sophisticated roaming is when the hand-set supports two or more frequency bands (dualband, three band etc.) in combination with SIMauthentication. With this solution available theroaming from one network technology to an-other could be implemented transparently as inthe GSM case with homogenous networks. Thisis the case for roaming between the GSM-900and GSM-1800 radio access in Europe today.

There are several challenges involved when con-sidering roaming between different networks. Atpresent there are several initiatives developingnetwork solutions to allow mobility and enabletransparency between various technologies. Oneof them is mobile IP, which is a standard devel-oped by IETF [2]. The concept of Mobile IP isbased around three network entities: The MobileNode (MN), the Foreign Agent and the HomeAgent (HA). The basic idea is that the MN getsaccess to visited networks (facilitated by the FA)and gets connected to its home data and applica-tions. HAs are entities located on the MN’shome network that are capable of tunnelling themobile node’s datagrams to it while it is away.This Mobile IP concept can also be applied towireless network, and networks of different type.

While the Mobile IP concept is well defined,some challenges remain in order to make viableimplementations for e.g. technology roamingconcepts. One challenge is where to implementthe FA; at the mobile or in a router in the visitednetwork. Another challenge is how to implementthe MN software on the mobile device, becauserelevant terminals such as laptops, PDAs and


mobile phones are rather closed platforms, re-quiring proprietary applications.

2.4.2 MVNO OpportunitiesFrom an MVNO point of view roaming bothwithin homogeneous networks and heteroge-neous networks is considered important. TheGSM roaming capability implies that an MVNOneeds a service agreement with only one mobileoperator and still provide large service coverage.

Roaming in between heterogeneous networkmight introduce new opportunities for a virtualoperator. There are several scenarios, whichmight be promising:

1 Fixed net operators might extend their busi-ness with mobile services by enabling roam-ing, thus cater for a complete product offering.

2 Mobile services combined with WLAN hotspots might attract a mobile customer segment

with large capacity needs to communicate outof office.

3 Mobile services supporting several localmobile standards, thus enabling real interre-gional coverage.

3 Service ProviderTo summarise the possible MVNO alternativesand the related service opportunities so far, wehave:

• Reseller;• MSC+HLR;• GPRS extension;• IP zones integrator;• Technology roaming.

All these alternatives share one common charac-teristic – they are MNO or technology centric.The two first seek to negotiate an agreementwith the MNO to get a suitable vertical integra-

Figure 3 MVNO room ofoptions (pink background)with indication of “pure”

MVNO vertical networkintegration, GPRS & WLAN

horizontal network integration(lower arrow) and Service

Provider horizontal serviceintegration (upper arrow)

Figure 2 Roaming betweenGPRS/UMTS and IP zones

with use of Mobile IP

Hot spot WLAN areas

GPRS / 3GRoaming

Radio Base StationRadio Base Station

Radio NetworkController

GPRS / 3G Core

Network

Internet

MN

Customers & Users

Reseller

Mobile ServiceProvider

Infrastructure

Mobile NetworkOperator

WLANIP zones

incl.GPRS

Techn.Roaming

GPRSextension

GPRS & WLANHorisontal network

integration

Service providerHorisontal network

integration

“pure” MVNOVertical network

integration


tion point for making business. The three lastseek to find some alternative business and/ormarket entrance strategies horizontal to theMNO via alternative technical solutions. Look-ing at the MVNO definition in paragraph 2.1.1,we recognise the customer approach. The gluefor combining these two will be the ServiceProvider role, shown as the upper horizontalarrow in Figure 3.

There are many service opportunities that areavailable to an MVNO regardless of the choiceof one of the models indicated by Figure 1.Some such opportunities are:

• Branding;

• Billing;

• Customer support;

• Exclusive application/features for terminal/SIM;

• Exclusive service/features in the network;

• Collaboration with partners such as contentproviders, application providers, Internet ser-vices providers, etc.

The possibilities for an MVNO when merging aReseller role with a Service Provider will be tocontinue to handle the three first ones in a properway and search for some uniqueness in a combi-nation of the three last.

3.1.1 Whole ProductsActors in the telecommunication market striveto fulfil their user needs by offering completeservices, thus being the only communicationprovider towards the customer. A Service Pro-vider approach would make sense to a lot ofactors that today only provides partially services.Typical candidates could be fixed net providers,CATV providers, ISP’s. These actors wouldcomplement their existing service portfolio byintroducing resold mobile services or IP zones,dependant on their customers needs and owncapabilities.

3.1.2 Customer Mobility NeedsCustomers have different service needs withrespect to mobility. Many segments are very pre-dictable and mobile within a limited area: Home,work, local community, commuting in betweenand from time to time a holiday or a businesstrip outside the primary roaming space. Suchservices needs might be well served by a ServiceProvider capable of identifying the customermobility and combing access technologies acc-ordingly. Work and home is suitable for IPzones whereas GSM/GPRS comprises the com-muting and ad hoc coverage.

3.1.3 Time to MarketTime to market is in many cases vital to a com-pany’s success. The Service Provider approachis on several areas a relatively fast way to launchservices since little or no physical build out isneeded. However, this might only be the case ifthe strategy is based on “me too” thinking. Most

Figure 4 Horizontal serviceintegration – Access andnetwork independence

WLANbase

Service=

content&

servers&

applications

IPback-bone

GSM / GPRSGPRS

services

MobileIP

IP zonesintegration

INTRANET

INTERNET

ADSL/VDSL


companies tries to stand out from the crowd, cre-ate a competitive edge, by offering somethingunique. Each unique entity or service feature hasto be developed and here there are no shortcutsfor a Service Provider. The benefit will then beto focus on just this uniqueness and not the net-work related developments and trade-offs.

3.1.4 MVNO OpportunitiesThe idea behind an MVNO is to deliver mobileservices to the customer without building yourown mobile access network. This means lowercost and higher flexibility with regard to busi-ness focus. As a Service Provider this idea istaken even further – trying to be independent ofnot only the MNO but the network and radioaccess over all, see Figure 4.

The strategy will then be to decide where tonegotiate the interconnection and reseller dealsand what services and application to developand/or bundle in the product offer to the cus-tomer.

4 ConclusionMVNO was initially defined as sitting on topof the MNO, serving the customers with mobileservices without its own mobile access licence.The experience from several markets is that theReseller model makes it handier for the MNOto control the vertical integration.

The focus for MVNOs as we see it will shiftfrom this vertical integration view, to investigatealternative MVNO models. One approach is tolook for alternative technologies to deliver thesame services or at least the interesting art of theservices for your business. WLAN IP zones andGPRS extension seems to be the candidates withthe highest potential for the moment. When wecombine them, we also build the basis for ser-vice provisioning across access technologies –technology roaming.

Further it has been recognised that one of themain drivers for the MVNO, the mobile serviceoffering to the customer, points in the directionof focus on the Service Provider role. This hori-zontal integration providing mobile services tothe customer, possible also in combination witha WLAN or GPRS service, can give an MVNOthe needed strength when dealing with differentnetwork operators.

5 References1 Virtual Mobile Services: Strategies for Fixed

and Mobile Operators. Ovum Ltd., 2000.

2 IETF. IP Mobility Support. (RFC 2002.)

Telektronikk 4.2001

Figure 1 A mobile virtual network operator’s customer – making and receiving calls

IN

SIM

INSIM

Otheroperator

MNO

MVNO

SIM - Subscriber indentification module

IN - Intelligent network

HLR - Home location register

VLR - Visitor location register

MSC - Switch

Outgoing call

Incoming call

Signaling/database

HLR

VLR

HLR MSCMSC

1 IntroductionMobile communications markets contain severalkinds of players who offer mobile services tocustomers. Some of them do not own the entireinfrastructure necessary to deliver the services,including a license to utilize the radio spectrum.A useful overview of these kinds of operators isgiven in Ovum 2000.

In August 2000 we witnessed the appearance ofthe first mobile virtual network operator (here-after MVNO or virtual operator) in the worldwhen Tele2 signed an agreement with themobile network operator Sonofon in Denmark.A year later Tele2 signed another MVNO agree-ment,2) this time with Telfort in the Netherlands.Mobile network operators (hereafter MNO ornetwork operator) in the Netherlands do nothave an obligation to give access to MVNOs.In Denmark mobile network operators are ob-liged to provide access to their networks, includ-ing MVNO access, but merely on the basis

of commercially negotiated agreements. Appar-ently, both the mobile network operators andTele2 have found the MVNO concept commer-cially interesting.

In this paper we discuss what might influencethe parties’ incentives to voluntary enter into anMVNO agreement. We emphasize particularlythe effect of prices and price structures betweenthe host MNO and the MVNO. A comparisonwill also be made between accessing the MNOnetwork as a service provider and as an MVNO.3)

The rest of the paper is organised as follows: Insection 2 we give a brief overview of the rela-tionship between a network operator and amobile virtual network operator. In section 3 wetake a closer look into the MNOs’ choice of howto market their infrastructure-based products.We then go on to discuss the effects of differentprice structures offered to potential third partymarket organisations, like service providers (SP)

How do the Price Agreements Affectthe Business Case of an MVNO?K J E T I L A N D E R S S O N , H I L D E H A L V O R S E N A N D O L E C H R I S T I A N W A S E N D E N 1 )

Hilde Halvorsen (31) is Advisorat Telenor Mobile, where she isworking with business strategy.She holds an MSc in Economics(cand.oecon) from the Depart-ment of Economics, Universityof Oslo, 1997. She has previ-ously worked as an advisor inthe Ministry of Transport andCommunications, Postal Ser-vices and TelecommunicationsDepartment, with special focuson key regulatory issues in tele-coms such as allocation of tele-com licenses. She has alsobeen on a temporary stay as acivil servant at the EuropeanCommission.

[email protected]

Kjetil Andersson (37) isResearch Scientist at TelenorR&D. He holds a B.Com. (1990)from the Norwegian School ofEconomics and BusinessAdministration (NHH), and he iscand.polit. from the University ofOslo (1992). Before he joinedTelenor R&D he was a ResearchAssociate at the Department ofEconomics, the Norwegian Uni-versity of Science and Technol-ogy (NTNU). His main currentresearch interests are appliedeconometrics, telecommunica-tion market analysis and busi-ness strategy.

[email protected]

1) We thank Petter Bliksrud, Thor Gunnar Eskedal, Øystein Foros, Bjørn Hansen, Steinar Sandbu and LoneSemmingsen for helpful comments and discussion. None of these are responsible for remaining shortcomingsand possible errors in the paper.

2) See press release http://www.bit.se/bitonline/2001/08/24/20010824BIT00230/08240023.htm orhttp://www.Tele2.com

3) In the following, the term Service Provider (SP) will be used to describe all commercial entities that offermobile services to its customers by means of an MNO’s infrastructure but do not issue their own SIM cards.

34

35

and virtual operators in sections 4, 5 and 6. Sec-tion 7 offers some concluding remarks.

2 The MVNO – MNORelationship

A mobile network operator controls the wholechain of values needed to produce mobile tele-com services. It holds a radio spectrum license,issues its own SIM cards, owns and operatesthe infrastructure including the base stations,switches and the gateway with the home locationregister (HLR). It is in full control of marketing,branding and pricing, as long as it stays withinthe regulatory framework. A mobile virtual net-work operator offers mobile services to cus-tomers, but does not hold a spectrum license andhas to make an agreement to access an MNO’sradio network. However, the MVNO issues itsown SIM cards, has its own switch, HLR andservice platforms, and has a full customer-careand billing system.

Figure 1 illustrates the structure of these twooperators and demonstrates the routing of a callmade by an MVNO customer to a customer inanother network (black arrows) and vice versa.For the outgoing call, the nearest base station ofthe host MNO first accepts the MVNO cus-tomer’s SIM card, and the call is routed to theMNO switch and then on to the MVNO’s switch.This operation will be referred to as ‘roamingout’ in the following. Finally, the call is routedto the network of the other operator. The dottedarrows depict the routing of an incoming call.The part from the MVNO’s switch to the termi-nal (SIM) of the MVNO customer is referred toas ‘roaming in’. A call from one MVNO cus-tomer to another (not illustrated in the figure) isthe combination of roaming out and roaming in.

3 The Network Operator’sChoice of Market Channel

An MNO faces several options on how to markettheir infrastructure-based products. It mightchoose to let its own market organisation sellthe final products; it might choose to use a thirdparty such as a virtual operator or a serviceprovider, or it may choose a combination ofthese.

As pointed out by e.g. Foros and Hansen (2000),the strategy of the MNO towards such third par-ties depends on the business stealing and the vol-ume effects: Does the network operator simplyface a new competitor who steals customersfrom their own core business, or will the third

party contribute positively to the revenues bymaking the old customers buy a larger quantityor by bringing in new customers.

Even if the third party does not add valuethrough the volume effect, the network operatormay choose to use the third party as a marketchannel. If the third party can run the marketoperation more effectively than the MNO,5) totalprofits will be larger if the former runs the mar-ket operation. If the two parties are able to agreeupon a contract that splits total profits such thatboth parties are better off, an outsourcing of themarket organisation could be realised.

A network operator with a small customer basewill probably have higher incentives to giveaccess to a third party than one with a large cus-tomer base. The reason is twofold: Firstly, thesmaller network is more likely to have idlecapacity than the larger, and hence is more likelyto avoid congestion when connecting the thirdparty customers. Secondly, the business-stealingeffect for the smaller network operator will beless because the third party will capture morecustomers from the larger network. The factthat neither Sonofon in the Danish market, norTelfort in the Dutch market are the largest opera-tors in their respective markets lends support toour assertion that small players relative to largeplayers, indeed have larger incentives to giveaccess to third parties.

Ole Christian Wasenden (30) isResearch Scientist at TelenorR&D. He holds a Cand.Polit.degree in economics from theUniversity of Oslo (1997).Before he joined Telenor R&Dhe worked at the Ministry ofIndustry and Trade, Departmentof Industrial Policy. His maincurrent research interests aretelecommunication marketanalysis, business strategyand regulation.

[email protected]

4) The figure is from Foros and Hansen (2000).5) Normally this will not be the case because most existing operators will have a larger customer base than the

third party and hence lower cost through enjoying economies of scale in the network operation and/or themarket operation.

Figure 2 The choice of the MNO on how tomarket their infrastructure-based products4)

Infra-structuresegment

MNO

The market

MVNOSP

(Thirdparty)

Market-organisation

Telektronikk 4.2001


As described in section 2, an MVNO issues itsown SIM cards and controls its own switch.This puts it in a far better position to introduceadvanced services on its own service platformcompared to a service provider. Thus, it may beeasier for an MVNO to add value by introducingnew services, complementary or independentto the MNO’s services. On the other hand, theMVNO may choose to offer services in directcompetition to the network operator’s serviceportfolio. In the next section we take a closerlook at the price structure between the host MNOand potential third parties and discuss how thenetwork operator may set prices to balance thebusiness stealing effect and the volume effect.

Payments from (+) to (–)

Call: SP customer MNO Other

From SP to MNO s1 –r1 -

From SP to SP s2 –r1 -

From SP to Other s3 –r2 -

From MNO to SP - - -

From Other to SP - - -

Payments from (+) to (–)

Call: Customer ofthe MVNO MNO Other8)

From MVNO to MNO t1 –(p1 + p3) -

From MVNO to MVNO t2 –(p1 + p2) -

From MVNO to Other t3 –p1 –m1

From MNO to MVNO - q1 – p2 -

From Other to MVNO - –p2 q2

Table 1 Income and costs from calls. MVNO7)

Table 2 Income and costsfrom calls. SP

4 Third Party Price StructuresTables 1 and 2, together with Figure 3, give anoverview of the structure of the end-user prices,and the roaming and interconnect price compo-nents that need to be specified for the call port-folio of an MVNO and a service provider (SP).6)

The structure is partly based on the observedMVNO agreement beween Tele2 and Sonofon,and on the general agreements the Norwegiannetwork operators Telenor Mobil AS and Net-Com GSM have made with their serviceproviders, see Sonofon and Tele2 (2000),Telenor Mobil (2001), and NetCom (2001)respectively. Nevertheless, the structures arequite general, and provide a fairly wide-rangedescription of the core business models of anMVNO and an SP.

A positive symbol denotes an income for theMVNO/SP, and a negative symbol a cost.Hence, if all price elements were scalars and thesubscribers of the MVNO make x calls to sub-scribers of the MNO, the MVNO makes a sur-plus of (t1–(p1 + p3))x from these calls. How-ever, each element will in general comprise avector of prices. Thus, the end-user prices, ti andsi, i = 1,2,3, will contain all the elements in theprice menu the MVNO/SP offers its customers.Normally this would include monthly subscrip-tions, connect charges, and airtime tariffs. Thesame goes for the elements in the last twocolumns. At least, most MNOs would requiredifferent airtime tariffs for peak and off-peakperiods. Furthermore, the prices may or may notcontain connect charges and subscription.

For a call made from one MVNO customer toanother MVNO customer, the MVNO wouldpay the MNO p1 for transmitting the signal fromthe call-originator’s terminal to the MVNO’sMSC (roaming-out), and p2 for transmitting itfrom the MNO’s MSC to the receiving MVNOcustomer (roaming-in). The price p3 is theMNO’s termination price. For a player with sig-nificant market power, the regulatory authoritymay determine the level of this. Thus, althoughthe roaming in and termination services are tech-nically identical, the prices p2 and p3 may differfor a network operator that is free to set its ownMVNO access prices. For calls made from the

6) For simplicity, the tables do not contain the tariffs when an MVNO/SP customer roams on another network butits host MNO. The MVNO has to negotiate these roaming agreements itself, while the SP inherits the roamingagreements made by the host MNO.

7) This table is based on IMSI-routing, that is the routing used in the MVNO agreement between Tele2 and Sono-fon in Denmark. IMSI-routing means that the call is routed based on an analysis of the A-number. Thus the callis routed directly from the MNO’s network to the MVNO without analysis of the destination of the call (B-num-ber). It is possible to use ordinary routing, i.e. analysis of the B-number as in international roaming, howeverIMSI-routing has been preferred in MVNO concepts.

8) By “other" we mean another operator in the market, i.e. another mobile network operator, a fixed networkoperator or another MVNO.


MVNO to a customer subscribing to any othernetwork, the MVNO would pay p1 for roaming-out to the MNO, and m2 in transport and inter-connect charges to the terminating network. TheMVNO itself must negotiate its interconnectcharges with other networks. Finally, for incom-ing calls, the MVNO must pay p2 for the roam-ing-in service provided by the MNO. In additionit receives q1 and q2 in whatever interconnectcharges it manages to negotiate with the par-ties.9)

The price structure for the SP is much simpler. Itpays r1 to the MNO for calls terminating in theMNO’s network, that is calls to MNO customersand other SP customers. Furthermore, it pays r2to the MNO for calls terminating in other opera-tors’ networks. Thus, the SP ‘inherits’ all theinterconnect agreements (including roaming) ofthe host network operator, and does not receiveor make any payments to other network opera-tors than its host MNO. This simple structurereflects the fact that the service provider is re-selling the services provided by the MNO. Incontrast, the MVNO controls its own facilities,as illustrated in Figure 1. The latter, issuing theirown SIM cards, will have to negotiate their owninternational roaming agreements and intercon-nect agreements.11)

5 SIM Card Ownershipand Second-DegreePrice Discrimination

The difference in SIM card ownership betweenan MVNO and an SP gives rise to a key differ-ence in the MNO’s ability to price discriminate.The MVNO issues its own SIM cards, while theSP does not, although it may re-badge, and evenput in some additional features in the SIM cardsit rents/buys from the host MNO. Thus, the MNOknows the subscription type of the SP customerand can set the prices in r1 and r2 according tothe kind of subscription of the SP customer.

This means that the MNO to a large extent (butrestricted by regulatory practise) can influence

the price structure the SP chooses to offer itscustomers. In the Norwegian case, the MNOsoffer their SPs the prices r1

H and r1L. The SP

must, for each customer, choose whether heshould belong to r1

H or r1L. The r1

H price bundlecomprises a high fixed monthly subscription rateand a low airtime tariff, while the r1

L bundle hasa low monthly subscription rate and a high air-time tariff. Hence, the structures r1

H and r1L

resembles the MNO’s own end-user prices,designed to make the customers self-select to a‘high-end’ subscription or a ‘low-end’ subscrip-tion. The SPs, facing the same market as theMNO, choose to offer similar price structuresas the MNO.

Consequently, the MNO can prevent the serviceoperators from stealing the attractive high-endcustomers, that is, to capture these customerswithout having to pay an accordingly high total(subscription + airtime) amount to the MNO.The MNO does not know the subscription typeof the MVNO customer. This is due to the factthat the SIM card is foreign and accordingly,prices, p1, p2 and p3 must be set uniformly for

9) The Sonofon/Tele2 agreement exhibits a very simple structure which we think may be replicated in futureMVNO agreements: Firstly, the roaming-in price equals the roaming-out price, i.e. p1 = p2. This probablyreflects the fact that the MNO’s marginal costs of providing these services are equal, and that the MVNO ingeneral will have a balanced call pattern. For calls from one Tele2 customer to another Tele2 customer,Tele2 gets a 25 % discount on (p1 + p2). The termination charges, p3 and q1 are not specified in the agree-ment. We expect that q1 is at least as large as p1, otherwise the MVNO would incur a loss every time an MNOcustomer called an MVNO customer. Finally, the agreement does not specify a fixed monthly payment persubscriber from the MVNO to the MNO.

10) Figure 3 describes the streams of payment for a regular voice service between an MNO, an MVNO andanother telecom operator (“other"). The MNO is host of the MVNO. The MVNO’s subscribers are conse-quently part of the MNO’s network, indicated by red dots in Figure 3. The red arrows indicate payments madeby the MVNO, the green arrows payments made by the MNO and the blue arrows payments made by the otheroperator. Payment to the MNO are marked by p, while m marks payment to the other operator and q indicatespayment to the MVNO. The direction of the arrows shows how the call is routed.

11) This may be difficult and costly for a virtual operator ‘starting from scratch’ since it cannot offer reciprocalroaming possibilities.

Other

MVNO

p1 - roaming out

p2 - roaming in

p3 - interconnect to MNO

q1 and q2 - interconnect to MVNO

m1 - interconnect to Other

MNOp1

p2

q1

p3m1

q2

Figure 3 Payments betweendifferent telecom operators10)


all subscribers. Thus, the MNO can continue itssecond-degree price discrimination towards SPcustomers but not towards MVNO customers.12)

6 The MVNO PriceAgreements and Entry

As described above, the prices for the MVNOcustomers’ use of airtime must be set indepen-dently of subscription type. In general, this givesthe virtual operator more freedom to segment themarket according to its own will, compared tothe second-degree price discrimination casedescribed above. However, a network operatoranticipating this may set the roaming-in – roam-ing-out prices such that the business stealingeffect is minimized. For instance, a networkoperator that has most of its customers in thehigh-end segment could offer a price structure(set the price) with a low subscription fee andhigh airtime rate. Any virtual operator thataccepts this price structure is forced to operate inthe low-end segment only. (Conversely and the-oretically, a network operator with most of itscustomers in the low-end segment could offer aprice structure with a high subscription rate anda low airtime rate, forcing the virtual operator toserve the high-end segment only.) Thus, in prac-tise, the great price flexibility of an MVNOemphasized by several sources, see e.g. Ovum(2000), may be quite exaggerated.13)

A network operator with a balanced assortmentof high-end and low-end customers and givingaccess to an MVNO is forced to accept eitherbusiness stealing in the high-end segment orintense competition in the low-end segment. Forsuch a network operator, and for an entrant thatwants to serve both ends of the market, the ser-vice provider concept may be the preferablesolution.

Finally, players that control some infrastructurein other parts of the telecom business may bewell suited to become MVNOs. Well-knownexamples include fixed network operators whichmay add value by converged fixed-mobile ser-vices, and MNOs in other countries utilizingtheir existing international roaming agreements.

7 Concluding RemarksSumming up the insights from the previous sec-tions we may picture the archetypical MVNOconstellation: The host MNO is a small operatorwith a competitive disadvantage in the low-endmarket. The MVNO is a pan-European, mobileand fixed operator with a highly competitivemass-market organisation. Not coincidently,this case fits well with the observed constella-tions today.

BibliographyForos, Ø, Hansen, B. 2000. Connecting cus-tomers and disconnecting competitors – Thefacility-based firms’ strategy towards virtualoperators. Kjeller, Telenor R&D. (R&D N33/2000.)

Ovum. 2000. Executive report: Virtual mobileservices.

Varian, H. 1989. Price discrimination. In:Schmalensee & Willig (eds.). The handbook ofindustrial organization, vol. 1. Amsterdam,North-Holland.

Sonofon and Tele2. 2000. National roamingagreement for a mobile virtual network operator(“MVNO”). [online] – URL:http://www.tst.dk/samtrafik/samtrafikaftaler.htm

Telenor Mobil. 2001. Service provision agree-ment. [online] – URL: http://telenormobil.no

NetCom. 2001. Videresalgsavtale. [online] –URL: http://www.netcom.no

12) See Varian 1989 for a discussion of the various forms of price discrimination.13) In addition, to plan its own network activity it is obviously important for the MNO to get estimates on the

MVNO’s minutes of usage. In the Sonofon/Tele2 agreement, the MVNO is punished if the actual usage is 15 %above or below prognosticated usage. If the usage is above prognosticated level no tariff adjustments aremade, but adjustments in network quality may occur: “Traffic amounts exceeding the prognosticated levelwill be handled provided that the quality of other subscriber groups are not affected", see Sonofon and Tele22000, ANNEX IV, page 7. Thus, to a certain extent, the MNO can restrict the MVNO from running unan-nounced very aggressive price campaigns, by refusing to handle traffic above prognosticated levels (a similarpoint is made by Foros and Hansen 2000).

39

1 IntroductionDistance is dead!

In earlier ages the people in a person’s physicalproximity defined communities. We are now liv-ing in the Connection Age! People can travel toany place in the world in matter of hours. Peopleon opposite sides of the Earth can talk to eachother like they were sitting next to each other.Teenagers nowadays have close friends not justdown the street, but also in other cities and othercountries, obtained through chatting, e-mail, andmobile communication. In the digital age peopleare collaborating over huge distances, and theobstacle of physical separation has for all practi-cal purposes vanished. Our thesis is that commu-nication will totally transform society as weknow it today, and physical distance is dead inthe Connection Age!

The new era will give new meaning to the con-cept of community. We have already seen ex-amples of virtual communities on the Internet:social networks coming into existence despitethe physical separation of the members. Thefuture will explode into new types of communi-ties, new ways of collaborating and new ways ofliving and interacting with each other. This willfully transform our society rightfully into theConnection Age [10].

A possible way to implement market segmenta-tion is to address the communities themselves astarget markets, with the goal of trying to utilizesome of the inherent structure in the communi-ties to create a business opportunity. One possi-ble attribute to opt for is the network externalityeffect [2] – also called positive feedback [3].This can be extremely powerful, since the cus-tomer will experience lock-in, and the winnertakes it all. We will look at the possibility of

using this as a possible entrant strategy for anMVNO. The incumbent that owns the networkinfrastructure will generally dictate the terms inthe negotiations with a possible MVNO, and willbe able squeeze the latter on margins. Hence, theMVNO will need a strong business opportunityto leverage the negotiations with the networkoperator (NO) in order to create a win-win deal.

This paper is organized as follows. In the nextsection we will discuss communities and a math-ematical model of communities based on so-called ‘small-world’ networks. The linkagebetween the community concept and the mathe-matical model will aid us in the understandingof observed phenomenon and supply us with apossible description of observed communities.We also reflect upon adoption and adaptationrates in these communities. In Section 3 weaddress the emerging virtual telecommunicationoperators – Mobile Virtual Network Operator(MVNO). There is no rigid definition of whatan MVNO really is, and hence we need to makeprecise our particular meaning of the concept.Section 4 describes an entrant strategy based oncommunities and triggering network externali-ties. The approach is general of nature, and wedetail out the strategy in the case of a possibleMVNO operation. The paper is concluded withsome general remarks.

2 Communities and NetworksA community or personal network is any collec-tion of people sharing some interest, like ahobby, a social framework, etc. Hence, everyone of us is "members" of thousands of net-works, depending on time of the day, geographiclocation, surroundings, your own choice, andmany other factors. A person belongs to somenetworks out of free will. However, the vastmajority of networks we belong to are of such a

An Entrant Strategy for an MVNO Basedon CommunitiesK E N T H E N G Ø , E I R I K B E F R I N G A N D K E N N E T H T I L L E Y

The purpose of this paper is to explore a possibly alternative entrant strategy into the telecommuni-

cation market as a mobile virtual network operator (MVNO). We seek new trends and models that can

give complementary models for the development and deployment of telecommunication services. In

particular we explore the community concept as an alternative way of segmenting the market, and

explicitly using tight connections and the relations in such communities to ignite a network externality

effect in order to lock-in customers. A possible mathematical model of the sought networks is the so-

called small-world network. These networks can be characterized by the fact that the average number

of connections to get from one arbitrary node to another is very small compared to the total number of

nodes, and that there is high local clustering, that is; many local connections among the nodes. The

practical implication of this is that information can spread fast, despite high local connectivity. We outline

a possible entrant strategy that an MVNO operation can use based on the community approach, and we

discuss some example communities to illustrate the ideas.

Eirik Befring (41) is an advisor atTelenor R&D. He holds a Masterof Political Science from the Uni-versity of Oslo, Norway, and aMaster of Technology Manage-ment in telecom strategy fromNTNU, Trondheim, Norway. Heworks mainly with the earlystages of concept developmentrelated to services and contentin the telecom sector.

[email protected]

Kenth Engø (31) is a scientist atTelenor R&D. He holds a Masterof Science (1995) in industrialmathematics from NTNU, Trond-heim, Norway, a Dr.Scient.degree (2000) in informaticsfrom University of Bergen, Nor-way, and a Master of Technol-ogy Management (2001) intelecommunication strategyfrom NTNU, Trondheim, Norway.Since joining Telenor R&D in2000 his interests have mainlybeen in risk and real options,modelling and investigatingmathematical properties of so-called ad-hoc networks, andexamining the effects on thebehavior of the ECN markingscheme in a network with usershaving inelastic demands.

[email protected]

Telektronikk 4.2001

Telektronikk 4.2001

character that the membership is inflicted uponus by “external” reasons. Everybody living inNorway belong to the same geographic network“People in Norway”. The criteria defining thisnetwork are the geographic location of the per-son. Whereas, the network of teenagers isdefined by individuals aged 13 to 20.

Networks defined by external characteristics arenot very interesting, since these networks areintrinsically static. The defining character isbeyond the “control” and influence of the mem-bers, and this again lead to no or only marginalinteraction/communication between the mem-bers. The phenomenon of network externalitywill probably never occur in such static net-works. The reason for this is that the definitionand value of the network to each of the membersis not of the positive feedback type [2] orincreasing returns type [3].

We will exclusively focus on networks of thefree will type, so-called communities. In thesenetworks the individual itself has full controlover the memberships, and also exercise someinfluence on the dynamics in the network. Thechances of experiencing network externalitiesare greater in these types of networks, becausethere is interaction between the members, andthe whole community is possibly gaining valuefrom the addition of any new member. In busi-ness; communities of practice is seen as a possi-ble reinvention of the earlier team-approach fororganisations, giving groups of people infor-mally tighter bound by sharing their expertiseand passion for a joint enterprise [4].

Characterizing communities is not an easy task,and the result depends on your goal in mind.

We believe that communities can be an appro-priate setting for triggering network externali-ties. We do not only want to observe the phe-nomenon, but also use it in a constructive man-ner, in order to possibly launch new, ubiquitous,telecommunication services. One example com-munity is the tourist segment. Only 18 % of totalinternational travel is business related, and over60 % are leisure! Of the total international travelsome 10 % is pure backpackers; people con-stantly on the move, going from place to place,staying a week or two, and using more moneyper day than the average Japanese tourist[11,12].

Small-World Network ModelsEverybody on this planet is separated by only sixother people. Six degrees of separation. Betweenus and everybody else on this planet. The presi-dent of the United States. A gondolier in Venice... It’s not just the big names. It’s anyone. Anative in a rain forest. A Tierra del Fuegan. AnEskimo. I am bound to everyone on this planetby a trail of six people. It’s a profound thoughtÖ. How every person is a new door, opening upinto other worlds.” From John Guare’s play SixDegrees of Separation (1990).

One way to analyse communities mathemati-cally is graph theory. However, graph theoryhas, as of today, not matured enough to cope infull with the problems we want to analyse, suc-cessfully. Intense work is carried out on devising

Kenneth Tilley (42) is programdirector in Telenor. He holds aMaster in Economics from BI,Oslo, and a Master of Technol-ogy Management in TelecomStrategy from NTNU, Trond-heim. He works mainly withconcepts in the CRM andknowledge management area.

[email protected]

Social l

ayer

Applicatio

n laye

r

Host

laye

r

Physic

al laye

r

RouterRouter

RouterRouter

Host Host Host

Application

Application

Application

Figure 1 Different layers of networks

40


new models for networks incorporating the com-plex random structure that is observed in reallife [6]. The regular – lattice – graph is a graphmodel with well-defined and regular connectionpatterns. More promising models are foundamong refinements of random graphs, the so-called ‘small-world’ networks [1].

One can consider two extremes of networks. Thefirst are regular networks, where “nearby” nodeshave large numbers of interconnections, but“distant” nodes have few. The second are ran-dom networks, where the nodes are connectedat random. See Figures 3 and 4.

Regular networks are highly clustered, i.e. thereis a high density of connections between nearbynodes, but have long path lengths, i.e. to gofrom one distant node to another one must passthrough many intermediate nodes.

Random networks are highly un-clustered buthave short path lengths. This is because the ran-domness makes it less likely that nearby nodeswill have lots of connections, but introducesmore links that connect one part of the networkto another.

In [1], Watts and Strogatz studied what happensbetween these two extremes. They started withregular networks and “re-wired” the nodes. Thatis, they decided whether to leave each edge con-necting a pair of nodes in place, or to change itto connect the starting node to a different endingone, chosen at random. This decision was madeat random, with probability p for each edge.Thus, if p = 0, the original regular network isunchanged, but if p = 1, the resulting networkis completely random. Introducing a relativelysmall number of random connections dramati-cally changed the character of the graph. That is,for small values of p, the graphs retained theirproperties of being highly clustered, but the

average path lengths dropped dramatically. Forexample, for p = .001, (so that only 0.1 % ofthe edges in the graph have been randomlychanged), the “clustering coefficient” is over95 % of what it would be for a regular graph, butthe “characteristic path length” is less than 20 %of what it would be for a regular graph, see Fig-ure 5.

In a sense, locality is violated in ‘small-world’networks, since the ‘shortcuts’ in the networkrealize action at a distance.

Adoption and Adaptation RatesThe role of the telecommunication business inpersonal communication is to interconnect par-ties not in physical reach. This task can be ful-filled in at least two ways: by on-line telecom-munication services or delaying the service.

Regular network:connections to 4 nearest neighbours

Making connections

Small-word network:a few long-range connections

Random network:points connected haphazardly

Increasing randomness

Figure 2 Regular, ‘small-world’, and random networks

Figure 3 A regular grid withnearest-neighbour links

(196 nodes)


The last option means storing the communica-tion need as some sort of information.

The Tipping Point is the name given by epidemi-ologists for the dramatic moment in an epidemicwhen everything can change all at once.

In his work [14] Malcom Gladwell puts up anelegant framework of how social epidemicsemerges. The theory of tipping points explainswhy we have trouble estimating dramatic, expo-nential change.

Gladwell’s premise is that in addition to apply-ing to viruses, this type of pattern is observed inmany other situations. The book is filled withfar-reaching examples, from the resurgence of

Hush Puppies shoes to the popularity of SesameStreet to an epidemic of teen suicides inMicronesia.

Based up on the research of Stanley Milgram,the small-world problem, Gladwell identifiesthree principles of an epidemic, witch providesus the tools we need for how to go about reach-ing a tipping point.

The Law of the Fewis a law which states that only a few people mat-ter. There are people who are more socially con-nected than others, and thus have a much greatereffect on the spread of an epidemic. In the caseof a social phenomenon, they know the rightpeople to talk to, or they cross social boundaries,living comfortably in several worlds at once, orthey are just good salesmen. Their presencemakes an idea contagious.

The Stickiness Factoris the second law, meaning the length withwhich an idea or disease stays with you. Thelonger it does, the more time you will propagateit. People who hear about a new idea actuallyremember it, and in some way do somethingabout it. Gladwell suggests that little changes inthe presentation of ideas, designed to make theideas stickier, can have big effects. Gladwelltakes an interesting angle on this phenomenonby examining the success of Sesame Street.

The Power of ContextThe last law is the Power of Context, stating thatenvironmental influences play a far greater rolethan we give them credit for.

One section of the book that we found especiallyinteresting was a chapter in which the authordraws a parallel between the tipping point and thechasm described in Geoffrey Moore’s Crossingthe Chasm [15]. The premise of the chasm is thatduring the introduction of discontinuous techno-logical innovations, a chasm exists because themajority of users are unwilling to accept the rec-ommendation of very early users (innovators).

The need for technical tools or terminals to doeither the on-line communication or transferringthe stored information will give us differentadoption (buying the device) and adaptationrates (using the device). The simple voice termi-nals of today have high adaptation to adoptionrates, whereas the communication services on aPC have low rates. This means that the technicaltool/terminal in itself will have some level ofbarrier to utilise the service offered.

We believe that somehow igniting the networkexternality in free-will communities will possi-bly decrease terminal barriers. Along with

Figure 4 A random graph(200 nodes), with linkprobability p = 0.01

120

%

100

80

60

40

20

010-5 10-4 10-3 10-2 10-1

The Small-Wold Regime - 200 nodes!

ClusteringAvg. shortest path

Figure 5 Clustering andshortest path in a small-worldgraph (200 nodes)


proper support functions this can result in muchhigher adaptation rates and service efficiency forthe customers and communities. It is further pos-tulated that the communication needs of commu-nities, technical tool adoption and adaptationrates on a local scale is linked to global charac-teristics through chaos theory [5] and small-world network [1].

Targeting communities modelled as ‘small-world’ networks can be prolific in this setting.One of the characteristics of the ‘small-world’networks – the average distance between to arbi-trary nodes is small – can be interpreted, as thespread of information is more efficient in thesecommunities. Then the adaptation of new ser-vices is faster once the terminal has beenacquired.

3 The Relation to an MVNOWe believe the above ‘theoretical approach’ tocommunities enriches the view of communica-tion needs interesting for the telecommunicationbusiness. The focus on both communities andterminal adoption and adaptation rates gives us adirect link to the phenomenon network external-ity that is seen as one of the major characteristicsfor the telecommunication service deploymenttoday and in the future. Possible services thatenable the forming and growing of such net-works will be characterised by high adoption/adaptation rates and probably high mobility fac-tors – both personal, terminal and service mobil-ity. Technically, this means services independentof the existing core networks, service platforms,access networks and terminal operating system.Based on this an agent driven service platformmust be developed to link the terminal userinterface, service production and databasesneeded to run the service.

What is an MVNO?A Mobile Virtual Network Operator (MVNO)is a mobile operator that does not own its ownradio spectrum and usually does not have itsown network infrastructure. Instead, MVNOshave business arrangements with traditionalmobile operators to buy minutes of use (MOU)for sale to their own customers. Business agree-ments can be done with one or more networkowners operating in the same market.

The question we want to ask is: Why carry outvirtual community development through anMVNO?

According to Rheingold [13] “Virtual communi-ties are cultural aggregations that emerge whenenough people bump into each other oftenenough in cyberspace”. The communities incyberspace are more than self-sufficient and intheir virtual interactions are able to get similar,

if not truly the same, experiences as a socialinteraction.

• The MVNO adds value such as brand appeal,distribution channels, and other affinities tothe resale of mobile services. It may be signif-icant initial factors in community develop-ment.

• The MVNO usually offer no significant differ-ence in service or network performance, butoffer some special affinity to their customers.MVNOs can offer a product mix, or bundlingmix, that incumbent mobile operators cannotmatch. Both the product and the bundling mixmay differ from the incumbent, and favourcommunity building.

• The MVNOs are well-known companies, withmarketing and financial strength, and suffi-cient agreements with existing operators toprovide a good service coverage area.

• Some leading providers are actually deployingtheir own Mobile Switching Centres (MSC),Service Control Points (SCP) and mobile INinfrastructure in order to facilitate the meansto offer value-added services. In this way,MVNOs can treat incumbent infrastructuresuch as radio equipment as a commodity.

• Offering value-added services helps differenti-ate the MVNO versus the incumbent mobileoperator, allowing for customer acquisitionand preventing the MVNO from needing tocompete on the basis of price alone. AnMVNO may cultivate core competency andfocus areas like customer centricity, commu-nity service development, more than anincumbent.

Figure 6 Different adoptionrates vs. time


• MVNOs can have full control over the SIMcard, branding, marketing, billing, and cus-tomer care operations, but the entry cost islow. Flexibility makes the MVNO able totake the first mover advantage.

• The major benefit to traditional mobile opera-tors co-operating with MVNOs is to broadenthe customer base (sell additional MOUs) at azero cost of acquisition

Implementing a communication solution forcommunities like tourists and backpackers is ahuge challenge. A potential operator must even-tually obtain worldwide coverage, and the solu-tion should preferably be a mobile communica-tion system – that is if the traveller is to rely onhis/her own mobile terminal. An MVNO strat-egy now naturally enters. This is self-evidentbecause no single operator can build a world-wide mobile phone system with the costs thatsuch an operation will incur. This is just learningfrom several bankrupt satellite systems, and theenormous cost of acquiring UMTS licences incertain countries.

Lonely Planet has established collaboration withekit.com, a self-proclaimed MVNO that willexercise the above-announced strategy. Theircurrent solution is limited to commodity servicesfor people out travelling.

4 An Entrant Strategy into theTelecommunication MarketBased on Communities

There is a tight connection between the marketapproach and the task of entering a market.

The market entry strategy consists of:

1 market approach, marketing strategy, and

2 partnerships and who to join when approach-ing a market.

In the market strategy we utilize the networkexternality phenomenon. Network externality isthe effect usually described as the increase invalue for the people belonging to a community/network as a new member joins the population[2]. This can loosely be paraphrased as follows;the consumer is forced into using a specificproduct by addressing the consumer’s peers in away as to leave the individual without choice aswhat to choose.

Telecommunication network operators have notyet targeted this as an intentional market strat-egy, although the telecommunication businesshas for a number of years been the archetypicalexample of a business with an intrinsic networkexternality effect.

Partnership deals should be negotiated withcompanies that supplement and enhance yourown approach. This indicates that an entrantshould seek partnerships with market partici-pants that are reasonably modern in their cus-tomer handling and treatment.

Several ways of approaching the customer hascharacteristics that presumably will support thenetwork externality products in a better way thanordinary broadcast marketing.

The Tupperware model is a classical peer-to-peer type of spreading a product portfolio. Thebasis is a “franchise” model where ordinary peo-ple sell to friends and neighbours. This modelexploits normal everyday networks among peo-ple, and the there does probably not exist a per-son that has not been invited to a Tupperwareparty by a friend or colleague. Exploiting the“Tupperware” network effect in a mobile com-munication setting should be the goal of theentrant.

Hotmail has used a modern electronic alternativeto the Tupperware party. Hotmail specializes infree Internet-based email. On every email sentfrom a Hotmail subscriber there is a WWW linkthat takes the receiver directly to a signup ser-vice that lets you get your own account. It thusbecomes a short and personal link between theproducer of the service, the communicationchannel and the target(s). It is also a necessityfor the receiver to have an email service not toodissimilar from the sender’s service. In a lot ofinstances the receiver has to at least get a servicesimilar to the one already deployed.

A distinct possibility for any new entrant into atelecommunication market is to target the busi-ness or consumer market. In our concept this canbe viewed as a road map to an integrated market,not only one target in it self. The JapaneseI-mode concept is realized through the consumermarket with good fortune. It has been followedup by a B-mode targeted towards a more busi-ness type segment. The standard European wayis to establish the enterprise market as its first(and only (?)) target segment. This is a historicphenomenon that might not be the most validway of approaching markets in the future.

The MVNO CaseAn MVNO entering into the telecommunicationmarket with a new service must acquire a cus-tomer base – or exploit an existing customerbase. If the MVNO is in the unfortunate situa-tion of starting from scratch, this acquisitionshould be done fast and in a fashion exploitingthe natural network externality intrinsic totelecommunication services. Following the ideaintroduced in this paper about ‘small-world’ net-


works, peer-to-peer communities are then a natu-ral target community. The key issue is to addressthe way people communicate, mostly in adynamic group fashion, and use this as a basisfor technology, interface and marketing strate-gies. This is essential knowledge in starting upan enterprise like an MVNO. One of the mostimportant assets of an MVNO, being situatedclose to the customer, can be knowledge of thecustomer communication.

Hence, the communication solution shouldexpand the total market of telecommunicationservices, in order not to cannibalize the marketfor the network operators that own the infras-tructure. As an MVNO you will be dependent onthe incumbent operators, and you should avoid asituation where you compete on margins. Youhave to introduce something new when enteringthe market as an MVNO, and the overall goal isto create a win-win deal with the incumbentoperator. This should be a healthy symbioticrelationship, versus a deadly price war on theservices.

5 ConclusionsIn this paper we have discussed the above com-munity/network framework as a possible entrantstrategy for an MVNO into the telecommunica-tion market.

Communities that are suitable candidates for anMVNO initially entering into a market will typi-cally have high adoption and adaptation rates,giving the entering MVNO the possibility of cre-ating a customer base fairly fast for the preferredcustomer network. The above ideas are notsolely restricted to the case of an MVNO entry,but can also be applied in other situations whereone is faced with the discouraging task of enter-ing a new market segment with no previous orsimilar customer base.

Stochastic graph theory offers the strategicoption of analysing potential market segmentsin a different way than traditional demographicanalysis. We believe this to be important, be-cause it offers a more dynamic analysis of com-munities. This should however not be under-stood as demographic analysis not being impor-tant.

With the graph theory approach we have theoption to analyse aspects such as formation andevolution of personal networks and communi-ties, which are rather intractable issues in demo-graphic studies. Hence, using stochastic graphtheory we can try to better understand the innerworkings of the phenomenon of network exter-nality. This addresses the need for understandingthe phenomenon and how to utilize it, ratherthan just observing it. The literature is full of the

latter; examples of economic webs that displaypositive feedback, but random graphs and small-world networks are some of the first attempts intrying to understand the intricate and non-linearinner workings of network externalities.

6 References1 Watts, D J, Strogatz, S H. Collective dynam-

ics of “small-world” networks. Nature, 393,4 June 1998, 440–442.

2 Shapiro, C, Varian, H R. Information Rules– A strategic guide to the network economy.Harvard Business School Press, 1998.

3 Arthur, W B. Increasing Returns and theNew World of Business. Harvard BusinessReview, July-August 1996, 100–109.

4 Wenger, Snyder. Communities of practice:The Organizational Frontier. Harvard Busi-ness Review, January-February, 2000.

5 Gleick, J. Chaos – making a new science.London, Heinemann, 1988.

6 Watts, D J. Small Worlds – The Dynamics ofNetworks between Order and Randomness.New Jersey, Princeton University Press,1999.

7 Hayes, B. Graph Theory in Practice: Part I.American Scientist, 88 (1), 2000, 9–13.

8 Hayes, B. Graph Theory in Practice: Part II.American Scientist, 88 (2), 2000, 104–109.

9 Guare, J. Six Degrees of Separation: A Play.New York, Vintage Books, 1990.

10 Groove Networks. The Connection Age.January 18, 2002. [online] – URL:http://www.groove.net.

11 World Tourism Organization. Tourism High-lights 2000. August 2000.

12 Levy, M. No room at the inn. British Air-ways Business Life, December 2000/January2001, 63–66.

13 Rheingold, H. The Virtual Community.January 18, 2002. [online] – URL:http://www.rheingold.com.

14 Gladwell, M. The tipping point – How littlethings can make big differences. Little,Brown and Company, 2000.

15 Moore, G. Crossing the Chasm. 2000.

Telektronikk 4.2001

1 IntroductionIn this paper we describe a modelling approach,which provides decision support for evaluationof the strategies of an industrial agent that findsitself in complex relations of competition andcollaboration with other agents belonging to thesame industrial environment. This is exactly thesituation in which many telecom service pro-viders find themselves now, with deregulationprocess and convergence between telecommuni-cations, computer industry and content provisionbeing well under way. The objective is to pro-vide a set of quantitative decision support toolswhich would enhance the quality of strategic andtactical decisions, like decisions about productmix, pricing, and investment. Interplay betweenestablished network providers and virtual opera-tors is an important particular case of this situa-tion, when the necessity of quantitative decisionsupport is particularly evident.

Microeconomic theory [MW95] provides im-portant theoretical insights in these issues, espe-cially when the system under study finds itself inconditions of equilibrium. Unfortunately, equi-librium is a feature which is conspicuously lack-ing in the current telecommunication environ-ment which makes many established approachesinapplicable. Another weak point of the classicaltheory is often inadequate treatment of uncer-tainty, another central feature of today telecom-munications. This is the reason why we haveemployed techniques which were specially de-signed to incorporate uncertainty and dynamicsin decision models, and in particular stochasticprogramming [EW88, BL97]. On the theoreticallevel, this methodology was under developmentfor a few decades, but only relatively recentlythe state of software and hardware allowedlarge-scale applications. We supplement it byselected ideas of the game theory [Bin92]

because the part of uncertainty confronted by agiven decision maker results from the actions ofother decision makers.

Provision of services in telecom between two ormore parties requires that bits can be transportedand routed from one party to the other. In Inter-net, this is the purpose of IP (Internet Protocol).The service itself is offered transparently be-tween the parties in the packet structure offeredby the transport layer protocols TCP (Transmis-sion Control Protocol) and UDP (User DatagramProtocol). The service itself is not visible in thenetwork but hidden by the transport protocols.Offering services is thus a two-stage business:first, offering transport and routing of bits and,second, offering the application that existsbetween the communicating parties. The firstcapability is offered by the network operators.The second capability may be offered by anyother party, including the users themselves.

More precisely, e-mail is a service that existsabove TCP and is thus not directly visible to thenetwork. The network simply makes use of themail address to identify the areas where the com-municating users are located. The mail serviceis embedded in terminals and mail servers. Theservice may be offered via independent opera-tors increasing the value of the service beyondthat of routing and delivery. Such offers can besecurity in terms of access control and filtering,creation of SMS messages for alerting the useron a mobile phone that a mail has been received,and extending a LAN to incorporate home officesolutions and mobility.

The operator primarily offering the service mayalso want to market the network support. This isof course no problem if the service operator andthe network operator are the same. If they are

Modelling Market Uncertainty and Compe-tition in Telecommunication Environment:Network Providers and Virtual OperatorsJ A N A A U D E S T A D , A L E X E I A . G A I V O R O N S K I A N DA D R I A N W E R N E R

In this paper we describe a methodology for making optimal strategic decisions in the presence of

uncertainty directed specifically to telecommunication environment. We concentrate on the case when

a part of the uncertainty results from the actions of other actor(s) that pursue their own aims. This is

especially important issue in the telecommunications where uncertainty about market development is

coupled with emergence of new actors and new business roles. Our approach blends established

methodology for decision support under uncertainty known as stochastic programming with selected

ideas of game theory. The basic ideas are illustrated through the case study that involves relationship

between network providers and virtual operators.

We provide numerical examples and economic interpretations for static and dynamic cases under vari-

ous assumptions about such parameters as market sensitivity or quality of knowledge about the com-

petitor’s policies.

Alexei A. Gaivoronski obtainedhis PhD from Moscow Inst. forPhysics and Technology in app-lied mathematics specializing inoperations research and opti-mization. He has since worked inacademy (Ukraine, Austria, Italy,Norway) and telecommunica-tions industry (Italy). He is Pro-fessor of Industrial Economicsand Operations Research at theDept of Industrial Economicsand Technology Management ofNorwegian University of Scienceand Technology. His researchinterests are optimization anddecision support systems underuncertainty ranging from theoryto industrial applications to soft-ware and he has publishednumerous papers on these sub-jects. His applied interests arefocused on telecommunicationsand [email protected]

Jan A Audestad (60) is SeniorAdvisor for Telenor CorporateUniversity. He is also AdjunctProfessor of telematics at theNorwegian University of Scienceand Technology (NTNU). Hehas a Master degree in theoreti-cal physics from NTNU in 1965.He joined Telenor in 1971 after 4years in the electronics industry.1971 – 1995 he did researchprimarily in satellite systems,mobile systems, intelligent net-works and information security.Since 1995 he has worked in thearea of business strategy. Hehas chaired a number of inter-national working groups andresearch projects standardisingand developing maritime satel-lite systems, GSM and intelligentnetworks.

[email protected]

46

47

different, the service operator must lease net-work connectibility from operators owning thenetwork. There are many reasons why this canbe a consistent business logic. However, weshall not consider these aspects here.

We introduce this approach through the casestudy which describes relations between networkprovider, virtual operator and a population ofcustomers; see Figure 1. We develop a decisionmodel for this situation which is decomposed inthree sub-models: customer model, networkprovider model and virtual operator model. Forthe sake of transparency a number of simplifyingassumptions are made in these models, althoughthey are not essential and a number of more spe-cific details can be incorporated easily. In thesimplest case we consider two providers thatoffer the same type of service to a common mar-ket. This service is based on a telecommunica-tion network and the network capacity is utilisedfor its delivery.

Whereas one of the providers, which we callNetwork Operator (NO), owns the network,the other provider is Virtual Network Operator(VNO) not possessing her own network facili-ties. Therefore she must lease capacity from theNetwork Operator to be able to produce service.She sends a request for a needed amount ofcapacity to her rival and has to pay a price forthe received capacity units.

Customers in that market can subscribe to eachof the providers for delivery of service and paya price for it. For the sake of transparency thisprice has a simple and non-differentiated struc-ture, although arbitrarily complex price schemescan be incorporated in the similar manner con-sidering several customer types. We assume thatthe providers fix their prices at the beginning ofa time period for the whole period. The wholetime horizon under consideration consists ofseveral such periods.

The customer model assumes that the number ofcustomers wishing to purchase service from oneof the providers at a given time period is basedon the customer number in the previous period.It is corrected by the customer flow resultingfrom migration between the providers, unsub-scription or arrival of the new subscribers due tothe price differences between providers and/orthe price changes between periods made by thesame provider. This flow consists of a fixed anda variable part: some customers will (un-)sub-scribe to or from service or migrate betweenproviders regardless of price levels, whereas a(usually bigger) number decides on movementdependent on price changes. Feedback depen-dent on network effects can be also importantfor customer decisions: a greater amount of cus-tomers will provide more utility to these cus-tomers and thus attract even more subscribers.Other approaches are also conceivable so thatthe customer model itself could be seen as amodule providing detailed input for the completemodel.

Important part of the provider model for the Net-work Operator (NO) deals with the policy fordetermination of the service prices and the pricesfor capacity leasing within bounds determinedby the regulation authorities. Another importantpart of his policies deals with investment in theinfrastructure. The Virtual Network Operator’sdecisions cover service prices and the amount ofcapacity to lease. Both operators’ objectives areto maximise their profits, to increase their mar-ket share or to serve some other goals.

The usual economic view on a market is ”fromabove”, i.e. a general welfare function is formu-lated that will maximise the total welfare of thismarket [MW95]. We do not follow this approachhere because the objective is to provide decisionsupport tools for a given industry actor. There-fore we adopt the point of view of one of theoperators, and develop the decision model acc-ordingly. For the purpose of this case study wetake the view of the Network Operator, althoughthe point of view of the Virtual Network Opera-tor can be considered similarly. Naturally, hismain interest lies in maximising his own welfare(or market share), while the general welfare issecondary. To achieve this goal he formulatesthe model describing customer behaviour andhe has to predict his rival’s decisions as optimalresponses to his price policy. These modelsdepend inevitably on a number of parameterswith uncertain values. Therefore the adequatetreatment of uncertainty becomes of paramountimportance here.

In the subsequent sections we develop thisapproach in more detail. In the next section weconsider more formally the general structure of

Figure 1 Relations between service providersand customers

Adrian Werner (26) holds a Mas-ters degree in mathematics fromTU Bergakademie Freiberg,Germany. He is now studyingfor his PhD at the Department ofIndustrial Economics and Tech-nology Management at NTNUand is working on a thesis withthe working title “Modelling toolsfor evaluation of telecommunica-tion enterprise strategies”.

[email protected]

Virtual NetworkOperator (VNO)

Customers

Network Operator(NO)

subscription

price

price

pric

e

subscription

Telektronikk 4.2001


the model. After that the simplest possible caseis considered which deals with one time periodand a deterministic set-up. It is assumed for nowthat the providers follow a profit maximisingprice policy. Later this model is extended byallowing for uncertainties both about the net-work owner’s decisions and about the cus-tomers’ behaviour or “external” entities likeopportunity costs. A further extension is madeby introduction of a multiperiod model. Theoret-ical considerations are supplemented by resultsof numerical experiments.

Quantitative decision models for competitivetelecommunication environment recently be-came a subject of intensive research effort. Analternative and complementary approach is con-stituted by simulation models of systems ofinteracting agents known as agent nets [BEG98,Gai98, Gai99]. Different models which utilisethe game theoretical concepts were proposed in[LS92, QR01, SL88]. The distinctive feature ofthe approach presented here is the utilisation ofstochastic programming methodology for ade-quate treatment of uncertainty and the absenceof equilibrium coupled with selected notions ofthe game theory.

2 Top Level Description ofModelling Approach

Seen from point of view of the Network Opera-tor, the decision model consists of three blocks:enterprise model, customer model and competi-tion model shown on Figure 2.

Each period the decision process of the NetworkOperator consists of the following steps:

• Predict the customer response for a givendecision and a given competition responseusing the customer model;

• Predict the competition response for a givendecision using the competition model. In thisparticular case study the Virtual NetworkOperator constitutes the competition;

• Select the optimal policy from the enterprisemodel using as an input the predictions of cus-

tomer and competition response obtained onthe previous two steps.

The following notations are utilized here:

y – decisions of Network Operator (NO). Thesedecisions are the price y1 for service provisionto own customers and the price y2 for capacityleased by his rivals.

z – decisions of Virtual Network Operator(VNO). They comprise the price z1 for serviceprovision and the amount z2 of capacity leasedfrom the NO.

n = (n1,n2) – the total number of customers ofNO and VNO respectively. These numbersdepends on their respective decisions y and z.

F2(y,z,n) – the objective function of the VNO,it depends on respective decisions y and z ofproviders and on the number of customersn = n(y,z) obtained from the customer model.Examples of such a function are profit, marketshare, etc. Note that this function incorporatesthe knowledge of the network operator about theaims of his rival, namely the NO thinks that theVNO chooses his decisions from maximizationof this function. More formally, the NetworkOperator takes as predictions z(y) for decisionsof the Virtual Network Operator the solution ofthe following problem:

where Z is the set of admissible decisions of theVNO.

F1(y,z,n) – objective function of the NO, itdepends on respective decisions y and z of theproviders and on the number of customers nobtained from the customer model. Examplesof such a function are profit, market share, etc.Prediction z(y) of response of the VNO yieldedby the competition model and predictionn(y) = n(z(y),y) of the customer number providedby customer model are utilised for computationof its value for fixed decision y. Consequently,decision y is selected by maximisation of thisobjective function solving the problem

where Y is the set of admissible decisions of theNO.

In what follows we develop specific modelsusing this general structure paying special atten-tion to uncertainty and dynamics.

maxy∈Y

F1 y, z y( ),n y( )( )

maxz∈Z

F2 y, z,n y, z( )( )

Figure 2 General structureof decision model

maxyF1(y,z,n)n = n(y)z = z(y)

maxzF2(y,z,n)n = n(y,z)

n = n(y,z)

customer model

enterprise model

model of competition


3 Single Time PeriodWe start with the simplest possible case, whichinvolves the single decision period and assumesthat the Network Operator has full informationabout different parameters which define the cus-tomer model and the competition model. Thiscase will be used as a benchmark for more com-plex cases also because it allows an analyticsolution.

3.1 Deterministic CaseThe Network Operator’s profits consist of therevenue from the service provision and the rev-enue from the capacity leasing diminished by thecosts of service provision. These costs for ser-vice provision consist of a fixed and a variablecomponent. Similarly the profits of the VirtualNetwork Operator consist of revenue obtainedby service provision, diminished by costs of ser-vice provision and by costs of leasing capacity.

We assume now that the virtual operator has thefull knowledge about the price policy of the Net-work Operator. The amount of capacity availableto lease and the VNO’s service prices have to bechosen within some limits. Furthermore, it hasto be taken into account that this provider mustlease enough capacity to serve all demand of hercustomers (Quality of Service constraint). Thenthe Virtual Network Operator's response to theNetwork Operator’s prices and the assumed cus-tomer behaviour can be determined as the profitmaximising solution of her problem. This solu-tion can be stated explicitly.

The Network Operator knows that the VirtualNetwork Operator settles her policy as a profitmaximising response to his own prices. Takenthis into account he too will choose his decisionsfor a maximal profit and solves the same modelas the VNO. Thus he knows this provider’s deci-sions which he can substitute in his own model.Under certain assumptions to ensure concavityof the goal function we can derive an explicitsolution also for this model of the NetworkOperator.

Let us describe the model more formally, fol-lowing the structure defined in the previous sec-tion.

3.1.1 Customer ModelWe describe this model for the multiperiod case.For the sake of simplicity we assume that thecustomer behaviour is governed by price consid-erations only. Schematically the customer deci-sion process is described by Figure 3.

In order to describe the decision process mathe-matically, we need to introduce indexes for eachof the operators. With the Network Operator we

associate index i = 1, while with the Virtual Net-work Operator we associate index i = 2.

Here is the price charged by oper-

ator i at time t for its service and is

the number of customers by the end of period twho utilise a service provided by operator i. This

number depends on the prices charged by

operators and is structured as follows

(1)

where is amount of the first time customers

subscribed to operator i at time t, and is the

amount of customers that switch from operator j

to operator i at time t. Both and depend

on prices . In the general case this depen-

dence is derived from price/demand relationshipshown on Figure 4.

Here f(h) is the total demand for service in thecase if the unit price is h and d0 is the value ofdemand which corresponds to some referenceprice h0. We linearise this curve in the vicinityof the point h0, which gives the following app-roximate relation between price and demand

d = d0 – ch

where c is a coefficient to be estimated from themarket statistics. Using this approach we obtainthe approximate linearised relations between thenumber of customers and the price increments,

qit

mijtmi

t

mijt

mit

nit +1 = ni

t + mit + mij

t

qit

nit ,t = 0,1,...

qit ,t = 0,1,...

Figure 3 Customerdecision process

Figure 4 Price/demandrelationship

qi1

qi0,ni

0ni

1

t=0 t=1 t=2time

qi2 ni

2

price

f(h)

h0

d 0

h


which are valid in some vicinity of the referenceprice under conditions that the average serviceutilisation by a single customer remains con-stant. This last assumption can be easily relaxedby considering a “physical” customer consistingfrom several “logical” customers. This leads tothe following relations:

mti = k

ti + ci (x

t-1i – x

ti) (2)

mtij = k

tij + cij(x

tj – x

ti ) (3)

where the price structure is the following:

q ti = q + x t

i

Here q is a reference price, x ti is a price incre-

ment of operator i at time t, kti is the amount of

new customers which would subscribe to theservice of operator i at time t in the absence ofprice change, k t

ij is the number of customerswhich would migrate from operator j to operatori in the absence of price difference between ser-vices offered by these operators, and c t

i , ctij are

coefficients to be estimated from the marketdata. Through kt

i and kti it is possible to model

different aspects of the customer behaviourbesides the response to the price change.

In the case of a single period it is convenient todenote y1 = x1

1, z1 = x12 which together with the

relations (1)-(3) will yield the following expres-sion for the number of customers n1 and n2 ofoperators 1 and 2 (the NO and the VNO) respec-tively:

n1 = k1 – r1y1 + r12z1 (4)

n2 = k2 + r12y1 – r2z1 (5)

where parameters k1, k2, r1, r2, r12 can be ex-pressed through parameters found in (1)-(3).

3.1.2 Competition ModelThis is a profit model for the Virtual NetworkOperator as seen by the Network operator. Theprofit is defined as the difference between therevenue and costs which are the costs for leasingthe network capacity and the cost of service pro-vision. Denoting

y2 – price to charged by the NO for a unit ofleased capacity;

z2 – amount of capacity leased by the VNO;

Revenue and costs of the VNO can be expressedas follows:

Revenue. Given that the price charged by theVirtual Network Operator for its service is q + z1and the number of his customers is given by (5)his revenue is

(q + z1)(k2 + r12y1 – r2z1)

Costs. They are composed from two compo-nents:

• Cost for leasing of the network capacity y2z2;

• Cost of service provision g2 + e2(k2 + r12y1– r2z1), where g2 and e2 are respectively thefixed service provision costs and the variableservice provision cost per customer for theVNO.

Therefore the expression for profit of the VirtualNetwork Operator is

(k2 –(q – e2)r2 + r12y1)z1 – r2z12 – y2z2

There are two decisions of the VNO that affecthis profit in this model:

• The price difference z1 between the referenceservice price and the price charged by theVNO;

• The amount of the network capacity z2 tolease from the NO.

In order to predict these decisions the NetworkOperator assumes that the Virtual NetworkOperator maximises his profit under constraintson the quality of service. Therefore he obtainsthe prediction z(y) as the solution of the follow-ing optimisation problem:

Find z1 and z2 which maximise

(k2 –(q – e2)r2 + r12y1)z1 – r2z12 – y2z2

subject to constraints

z2 ≥ d(k2 + r12y1 – r2z1)k2 + r12y1 – r2z1 ≥ 0– ∆ ″ z1 ″ ∆

where d is the average amount of capacityrequired for the service provision for one userwith admissible service quality and ∆ is thebound for the price change. Note that the solu-tion of this problem depends on the price deci-sions y1 and y2 of the Network Operator. In thecase when the solution lies within these bounds,it can be expressed analytically as follows:


(6)

(7)

which means that the decisions of the VNOdepend linearly on the decisions of the NO.

3.1.3 Enterprise ModelThis model describes aims of the Network Oper-ator and how these aims depend on the decisionsof customers and competition. In order to bespecific we consider the profit model similar tothe competition model described in the previoussection, which requires a definition of revenueand costs of the Network Operator.

Revenue. It is composed from two components:

• Revenue from the service provision to the cus-tomers. Given that the price charged by theNetwork Operator for its service is q + y1 andthe number of his customers is given by (4)this part of revenue is

(q + y1)(k1 – r1y1 + r12z1)

• Revenue from the leasing of capacity to theVNO:

y2z2

Costs. They are the costs of the service provi-sion:

g1 + e1(k1 – r1y1 + r12z1)

where g1 is the fixed cost of service provisionand e1 is the variable cost of service provisionper customer for the network owner. There aretwo decisions of the NO which affect his profitin this model:

• The price difference y1 between the referenceservice price and the price charged by the NO;

• The price y2 charged to the VNO for a unit ofleased capacity.

Besides, his profit depends on the decisions z1and z2 of the VNO. Now the Network Operatorcan use the predictions of decisions taken by theVirtual Network Operator from (6)-(7). Substi-tuting these predictions into the expression forprofit and assuming that the Network Operatormakes decisions which maximise his profit weobtain the following optimisation problem:

z2 (y) =d

2k2 + r12 y1 + q − e2( )r2 − dr2 y2( )

z1(y) =k2 − q − e2( )r2 + dr2 y2 + r12 y1

2r2

Find y1 and y2 which maximise


k1 – r1y1 + r12z1 ≥ 0– ∆ ≤ y1 ≤ ∆0 ≤ y2 ≤ U1

where a1 and a2 are expressed through parame-ters introduced before, ∆ is the bound on the ser-vice price change and U1 is the upper bound forthe price charged for the leased capacity fixed bythe regulation authorities. It is assumed here thatthe Network Operator has unlimited amount ofnetwork capacity, later we will dispense withthis assumption.

This problem is a simple quadratic programmingproblem which can be easily solved analyticallyif it is concave. This, however, is not always thecase and concavity conditions should be derivedfrom the parameters which describe the cus-tomer and competition behaviour.

This simple model is used as a benchmark andas a starting point for development of moreadvanced and realistic models.

3.2 Allowing for UncertaintiesIt is very likely that the Virtual Network Opera-tor has no such a broad knowledge about the sit-uation as assumed in the previous section. Therecan be several sources of uncertainty like

• Pricing policy of the NO;

• Customer behaviour as described by the co-efficients for the customer flow;

• Customer behaviour with regard to demand.There may exist different customer types,each with a specific demand pattern, demandmay change over time, the demand behaviourmay be not clear to the operators, etc.;

• Costs for the provision of service.

In the next model the first two sources of uncer-tainty are taken into account. Both operators areuncertain about the customer behaviour (marketsensitivity, the amount of customers that are(un-)subscribing for service). Since at the deci-sion moment the operator has no exact knowl-edge about the customers’ behaviour, the neededamount of capacity cannot be predicted exactly.We still assume that the Network Operator hasunlimited amount of capacity available, but theVirtual Network Operator may face the problem

− r1 −r12

2

2r2

⎛

⎝⎜

⎞

⎠⎟ y1

2 + dr12 y1y2 −d2

2r2 y2

2 + a1y1 + a2 y2


that some demand remains unserved. Thereforeopportunity costs may occur and have to beincluded in her model. We assume that the VNOis also uncertain about the exact level of theseopportunity costs.

Furthermore, the Virtual Network Operator canonly estimate pricing decisions of the NetworkOperator to a certain extend. However, we stillassume that the Network Operator has the com-plete knowledge about the decision behaviourof the VNO. As described previously, the NOwants to find out decisions of his rival. Then hecan choose his policy such that he can maximisehis expected profits given the response of theVNO on his decisions that will maximise herprofits too. Observe that only average or ex-pected profits can be maximised due to thedescribed uncertainties. Therefore the NetworkOperator describes how the Virtual NetworkOperator perceives the NO’s decisions by helpof uncertain parameters and his actual price. Butwe have to bear in mind that revenue and costsare caused only by those customers that are actu-ally served. This number is limited by theamount of available capacity. Furthermore thereis a constraint ensuring nonnegative customernumbers. When he has solved this problem ofpredicting his rival's response the Network Oper-ator again can substitute this solution into hisown profit maximisation problem and solve it tofind his best decisions under the given response.

4 Two Stage ModelNow the described model will be extended bythe second time period. However, it representsan intermediate stage between the one periodmodel and a general two period model: at thebeginning of the first period all pricing and leas-ing decisions are made and remain unchangedalso throughout the second period. But, when thedemand becomes known after the first period,the Network Operator has the possibility to ex-tend capacity. This means that the number ofdecision variables of the Virtual Network Opera-tor remains the same, at the beginning of the firsttime period she decides about service prices andamount of leased capacity. However, the Net-work Operator decides at the beginning of thefirst time period about the service and capacityprices and amount of capacity he makes avail-able for leasing. The introduction of the latterdecision variable gives a new quality of the con-

tract between the two providers: so far, only theVNO has determined how much capacity shewants to lease; an upper bound was settled by an“outer force” like a regulation authority. Nowalso the Network Operator can intervene by set-ting an upper bound on capacity he has availablefor leasing. At the beginning of the second timeperiod he has to make a further decision:whether to extend the network and to whatextent.

For the Network Operator the available capacityin the second period depends also on a possiblenetwork extension. This has to be taken intoaccount when formulating the new objectivefunction. Then the decision process proceeds asdepicted before: the NO tries to predict how hisdecisions are perceived by his rival and whichstrategy the rival chooses as response to his pol-icy for formulating his own prices accordingly.

Since the decisions depending on the other part’spolicy remain constant throughout all time peri-ods, both prediction and the decision problemcan be solved as a combination of both period’sproblems by the use of a discount factor.

5 Mathematical Descriptionof Modelling Approach:General Case

In this section we describe formally our mod-elling approach for the general multiperiod casewith uncertainty. In order to simplify our exposi-tion only two periods are considered, but exten-sion for the case of more periods is straightfor-ward. For the case of two periods the decisionprocess is presented on Figure 5.

We introduced here the following notations:

ωt, t = 1,2 – uncertain parameters from the pointof view of the NO at time period t. These param-eters describe the quantities from customer andcompetition model about which the NetworkOperator has uncertain knowledge. The informa-tion about these parameters which are availablefor the Network Operator is described by proba-bility distributions. It is assumed that the valuesof these parameters become known at the end ofperiod t.

yt, t = 1,2 – decisions taken by the NO at thebeginning of period t before the values ωt ofuncertain parameters become known. Thesedecisions are taken with the aim to improve thevalues of some enterprise performance criterion,like profit or market share, averaged with respectto the values of uncertain parameters.

zt, t = 1,2 – reaction of competitors to decisionsof the NO. This reaction is forecast by the Net-work Operator using the model of competition.

Figure 5 Decision process fortwo periods in the presence ofuncertainty

Period 1 Period 2

Prediction

Decision

z1(y1)

y1 ω1

z2(y1, y2, z1, ω1)

y2(y1, z1, ω1) ω2


The Network Operator assumes that the com-petitors take decisions with the aim to improvethe values of some enterprise performance crite-rion, averaged with respect to the values ofuncertain parameters.

The decision that is going to be implementedimmediately by the Network Operator is y1. Inorder to take this decision it is necessary to fore-see its influence on the decisions taken in the sub-sequent periods by the Network Operator, com-petitors and customers. This is done through thedecision/prediction process depicted on Figure 5,which consists of the following steps. We presentit here in the most general form, while a morespecific model is considered in the next section.

1. Prediction of competitor’s reaction duringPeriod 2

For given y1, y2, z1, ω1 obtain prediction z2(y1,y2, z1, ω1) for decision of competitors duringPeriod 2 by solving the following problem.

Find z2 ∈ Z2 which minimizes


F2i (y1, y2, z1, z2, ω1) ≤ 0, i = 1 : m22

where F20(•) is the enterprise performance crite-rion of competition for Period 2 and the func-tions F2i(•) define different constraints on deci-sions of the competition like, for example, qual-ity of service constraints.

2. Finding optimal decision for Period 2For given y1, z1, ω1 and prediction z2(y1, y2, z1,ω1) for decision of the competitors during Period2 find the optimal decision y2 = y2(y1, z1, ω1) forPeriod 2 by solving the following problem.

Find y2 ∈ Y2 which minimizes


f2i(y1, y2, z1, z2 (y1, y2, z1, w1), ω1) ≤ 0, i = 1 : m21

where f20(•) is the enterprise performance crite-rion of the NO for Period 2 and the functionsf2i(•) define different constraints on decisions ofthe NO. Let us denote the optimal value of theperformance criterion by Q(y1, z1, ω1).

Eω2f 20 y1, y2 , z1, z2 y1, y2 , z1,ω1( ),ω1,ω2( )

Eω2F20 y1, y2 , z1, z2 ,ω1,ω2( )

3. Prediction of competitor’s reaction duringPeriod 1

For given y1 obtain the prediction z1(y1) for deci-sion of the competitors during Period 1 by solv-ing the following problem.

Find z1 ∈ Z1 which minimizes


F1i (y1, z1) ≤ 0, i = 1 : m12

where F10(•) is an enterprise performance crite-rion of competition for Period 1 and the func-tions F1i(•) define different constraints on thedecisions of competition.

4. Finding optimal decision for Period 1Having a prediction z1(y1) for decision of thecompetitors during Period 1, find the optimaldecision y1 of the NO for Period 1 by solvingthe following problem.

Find y1 ∈ Y1 which minimizes

(8)


f1i (y1, z1(y1)) ≤ 0, i = 1 : m11 (9)

where f10(•) is the enterprise performance crite-rion of the Network Operator for Period 1 andthe functions f2i(•) define different constraints ondecisions of the NO. Observe that the integratedcriterion which is optimised incorporates criteriafor both periods.

This decision process looks fairly involved.However, given the present state of the art in theoptimisation methods and related software it isfeasible to build a decision support system basedon this approach. It is in this connection that thenumerical approaches developed in the field ofstochastic programming become pivotal. Onepossible way to proceed consists of the follow-ing steps.

• Approximate probabilistic distributions of theuncertain parameters by a finite number ofscenarios that take the form

(10)

where it is assumed that ω1 takes the value

with probability p1i and ω2 takes theω1i

p1i ,ω1i( ), p2ij ,ω2

ij( ),i = 1:N , j = 1:Mi

Eω1f10 y1, z1 y1( ),ω1( ) + Q y1, z1 y1( ),ω1( )( )

Eω1F10 y1, z1,ω1( )


value with probability p2ij under the con-

dition that ω1 takes the value .

• Construct the so-called deterministic equiva-lent of the problem (8)–(9) [EW88,BL97]which makes the problem amenable to solu-tion. A specific form of the deterministicequivalent depends on the structure of theproblem and one example can be found in thenext section.

• Use commercial software as building blocksfor the solution of the deterministic equivalentand for the development of a decision supportsystem.

Alternative stochastic programming approachallows working directly with continuous distri-butions by application of sampling techniquesand stochastic gradient methods [Gai88].

6 Two Period Model withUncertainty and Investmentin Infrastructure

Let us develop now a specific decision model forthe Network Operator in the presence of VirtualNetwork Operator(s) using the approach fromthe previous section. This model includes themodels which were described informally in sec-tions 3.2 and 4 and constitute the further devel-opment of one period deterministic model fromsections 3.1.1 – 3.1.2. Notations remain essen-tially the same with some changes due to theconsideration of scenarios and two time periods.

We start by describing decision variables that inthe case of the network operator include also thepossibility to invest in the infrastructure.

Decisions of the Network Operator:y1 = (y11, y12, y13) – decisions during Period 1;

y11 – price to charge for its service to customersduring Period 1;

y12 – price to charge for capacity to the VNOduring Period 1;

y13 – maximal amount of capacity to lease to theVNO during Period 1;

y2 = (y21, y22, y23, y24, y25) – decisions duringPeriod 2;

y21 – price to charge for its service to customersduring Period 2;

y22 – price to charge for capacity to the VNOduring Period 2;

ω1i

ω2ij

y23 – maximal amount of capacity to lease to theVNO during Period 2, we assume that y23 ≥ y13;

y24 – amount of capacity to add at the beginningof Period 2, this additional capacity is availableduring Period 2;

y25 – binary variable which equals 1 if decisionto add capacity is taken and 0 otherwise.

Decisions of the Virtual Network Operator:zt (zt1, zt2) – decisions during Period t, t = 1,2;

zt1 – price to charge for its service to customers;

zt2 – amount of capacity to lease from the NO.

Next, let us define uncertain parameters ωt, t =1,2. These are the parameters that define the cus-tomer model and the competition model fromsections 3.1.1 and 3.1.2. More specifically,

where and describe the uncertainty

related to enterprise and competition model

respectively. Here are taken from

the relations

that describe the customer model in the case oftwo periods similar to the model (4)-(5) and

. Besides,

where are the variable cost of service pro-

vision per customer taken from the profit expres-sions for operators

which are similar to the ones presented in sec-tions 3.1.2 and 3.1.3 devoted to the enterprisemodel and the competition model respectively.Other parameters have the following meaning:

– the opportunity cost of not meeting one

unit of demand for NO during Period t;

r15t

g1t + e1

t k1t − r11

t yt1 + r12t zt1( )

g2t + e2

t k2t + r12

t yt1 − r22t zt1( )

e1t ,e2

t

r13t = k1

t − r11t q − e1

t( ),r14t = r12

t q − e1t( ),

r23t = k2

t − r22t q − e2

t( ),r25t = k2

t

r21t = r12

t

n1t = k1

t − r11t yt1 + r12

t zt1

n2t = k2

t + r12t yt1 − r22

t zt1

r11t ,r12

t ,r22t

r2tr1

t

ω ηtt t t t t t

t t t

r r r r r

r r r t

= ( ) = ( )= ( ) =

1 2 1 11 19

2 21 26 1 2

, , , ,..., ,

,..., , ,


– the variable cost of adding a unit of

capacity;

– the fixed cost of adding capacity, this

parameter together with is defined only for

t = 2;

where b is the current network capacity of the

NO and are the amounts of capacity

required to satisfy a demand from one customerfor the NO and the VNO respectively;

– the opportunity cost of not serving one

unit of demand for the VNO during Period t;

Variable ηt is used to describe uncertainty in theknowledge that the Virtual Network Operatorhas about the decisions of the Network Operatoras seen by NO. More precisely, if decision of theNetwork Operator is yt then he assumes that theVirtual Network Operator thinks that this deci-sion is

and uses this value in his competition model.

Generally, uncertain parameters ωt are describedby continuous probability distributions. Follow-ing the scenario approach outlined at the end ofthe last section, we approximate the possible val-ues of uncertain parameters by a finite numberof values with given probabilities as in (10):

In accordance with the framework described inthe previous section, decisions of both providersin period 2 depend on the values of random vari-ables ω1. In the case of a finite number of sce-narios i = 1:N they will depend on the index of ascenario. Therefore we denote

– decision of the NO in Period 2 under sce-

nario i;

– prediction of the response of the

VNO to decision of the NO in Period 2

under scenario i.

y2i

z2i = z2

i y2i( )

y2i

ω1i = r1

1i ,r21i ,η1i( ),ω2

ij = r12ij ,r2

2ij ,η2ij( ),i = 1:N , j = 1:Mi

yt = yt + η t

r24t

d1t ,d2

t

r18t = k1

t −b

d1t ,r19

t =1

d1t ,r26

t =1

d2t ,

r10t = q − e1

t ,r28t = q − e2

t

r16t

r17t

r16t

1. Decision Model for the NetworkOperator

Now we are ready to describe the deterministicequivalent of the decision process outlined in thelast section for the case of an environment con-sisting of the Network Operator and the VirtualNetwork Operator. We start by defining the inte-grated decision model of the NO that combinesdecision models for both periods. Its objective isto find the optimal decision y1 of the NetworkOperator for Period 1 and his optimal decisions

for Period 2 and scenarios i = 1:N having

predictions z1(y1) and for decisions of

the VNO by solving the following problem.

Find y1 and which maximizes

(11)


– ∆ ≤ y11 ≤ ∆, (12)

(13)

0 ≤ y12 ≤ U1, (14)

(15)

0 ≤ y13 ≤ b, (16)

(17)

(18)

(19)

(20)

Similarly to (8), the objective function in (11)consists of two terms. The term f1(•) describesthe profits of the Network Operator during

Period 1. Function describes the profits

of the NO during Period 2 under scenario i.Therefore the second term describes averageprofits of this operator during Period 2 dis-counted with the discount coefficient α ≤ 1.Function f1(•) can be expressed as follows.

f1 y1, z1 y1( )( ) = r131 − pi

i=1

N

∑ w1i⎛

⎝⎜

⎞

⎠⎟ y11 −

r111 y11

2 + r141 + r12

1 y11( )z11 y1( ) + y12z12 y1( )−

p1ii=1

N

∑ r151i + r10

1( )w1i

f 2i (•)

k12ij − r11

2ij y21 + r122ij z21 ≥ 0,

i = 1:N , j = 1:M,

k11i − r11

1i y11 + r121i z11 ≥ 0,i = 1:N ,

y24i ≤ Y24

i y25i ,i = 1:N ,

y13 ≤ y23i ≤ b + y24

i ,i = 1:N ,

0 ≤ y22i ≤U1,i = 1:N ,

−∆ ≤ y21i ≤ ∆,i = 1:N ,

f1 y1, z1 y1( )( ) +α p1ii=1

N

∑ f 2i y2

i , z2i y2

i( )( )

y2i ,i = 1:N

z2i y2

i( )y2

i


Observe that the costs for the infrastructureupgrade are taken into account in the profit cal-

culation for Period 2 through the variables

and .

Constraints (12)–-(20) consist of five groups.Constraints (12)-(13) define the bounds for theservice price change during Periods 1 and 2respectively, while (14)-(15) define the boundsfor the price charged for the leased capacity.Constraints (16)-(17) fix the bounds on theamount of the leased capacity which assures thatit does not exceed the total network capacity.Constraint (18) assures that the amount of newly

built capacity will be positive only when

the decision to expand capacity is taken, namely

= 1. Parameter is the maximal amount

of capacity increase under scenario i. Constraints(19)–(20) ensure a nonnegative customer num-ber.

Predictions z1(y1) and of decisions of

the Virtual Network Operator which enter thismodel are obtained through prediction modelsfor Periods 1 and 2.

2. Prediction model for Period 1For a given y1 obtain a prediction z1(y1) for deci-sion of the VNO during Period 1 by solving thefollowing problem.

Find z1 and v1 = (v11, ..., v1N) which maximizes

(21)


(22)

v1i ≥ 0, i = 1 : N (23)

– ∆ ≤ z11 ≤ ∆, (24)

0 ≤ z12 ≤ y13, (25)

(26)k21i + r21

1i y11 − r221i z11 ≥ 0,i = 1:N

v r z r z

r r y r i N

ii i

i i i i

1 221

11 261

12

251

211

11 211

11 1

+ + ≥

+ + =η , : ,

F10 y1, z1,v1( ) =

r231 + r21

1 y11 + p1ii=1

N

∑ r211i η1

1i − p1ii=1

N

∑ v1i⎛

⎝⎜

⎞

⎠⎟ z11 −

r221 z11

2 − y12 + p1ii=1

N

∑ η21i⎛

⎝⎜

⎞

⎠⎟ z12 −

p1ii=1

N

∑ r241i + r28

1( )v1i

z2i y2

i( )

y24iy25

i

y24i

y25i

y24i

where w1i are potential customers of the Net-work Operator which are lost due to the lack ofcapacity for the service provision during Period1 under scenario i:

and are expected values of uncertain parame-

ters :

The first part of the expression for f1(•) (every-thing except the last sum) is the profit defined asthe difference between the revenue from the ser-vice provision and the leasing of capacity minusservice provision costs exactly as it was definedin the enterprise model of Section 3.1.3. Sincethe revenue stems only from those customersthat are actually served, the number of servedcustomers is determined as the number of poten-tial customers diminished by the number of cus-tomers that cannot be served due to the lack of

capacity. Function for profit during

Period 2 is very similar to f1(•) and can beexpressed as follows:

where w2ij are the potential customers of theNetwork Operator which are lost due to the lackof capacity for service provision during Period 2under scenario j provided that the uncertainparameters during Period 1 followed scenario i:

and are expected values of uncertain param-

eters conditioned on scenario i of Period 1:

r12i = r11

2i ,...,r192i( ) = p2ij

j =1

Mi

∑ r12ij .

r12i

r12i

w2ij = max 0,r182ij − r11

2ij y21i +(

r122ij z21

i y2i( )− r19

2ij y24i + r19

2ij z22i y2

i( ))

f 2i y2

i , z2i y2

i( )( ) =

r132i − p2ij

j =1

Mi

∑ w2ij

⎛

⎝⎜

⎞

⎠⎟ y21

i − r112i y21

i( )2+

r142i + r12

2i y21i( )z21

i y2i( ) + y22

i z22i y2

i( )−−r16

2i y24i − r17

2i y25i −

p2ijj =1

Mi

∑ r152ij + r10

2( )w2ij

f 2i (•)

r11 = r1

1,...,r191( ) = p1i

i=1

N

∑ r11i .

r11

r11

w1i =

max 0,r181i − r11

1i y11 + r121i z11 y2( ) + r19

1i z12 y2( )( )


where v1i are potential customers of the VNOwhich are lost due to the lack of capacity for theservice provision during Period 1 under scenarioi. The structure of the profit function F10(y1, z1,v1) is very similar to the one period competitionmodel from Section 3.1.2. The new elements are

the sums which contain parameters and

which are used to model imprecise knowl-

edge of the Virtual Network Operator aboutdecisions of the Network Operator. Besides, thelast term is appeared which represents averageopportunity costs for not meeting demand.

Constraints (22)–(23) connect the amount ofunserved demand with total demand and avail-able capacity. Constraint (24) fixes bounds onadmissible price changes, while (25) assures thatthe leased capacity does not exceed capacity thatthe Network Operator is ready to lease. Con-straint (26) ensures a nonnegative customernumber.

3. Prediction model for Period 2For a given scenario i of the Period 1 and a

given decision of the Network Operator dur-

ing Period 2 obtain a prediction for

decision of the Virtual Network Operator duringPeriod 2 by solving the following problem.

Find z2 and which maxi-

mize

(27)


(28)

v2ij ≥ 0, j = 1 : Mi (29)

– ∆ ≤ z21 ≤ ∆, (30)

0 ≤ z22 ≤ y23, (31)

v r z r z

r r y r j M

ijij ij

ij ij ij iji

2 222

21 262

22

252

212

11 212

12 1

+ + ≥

+ + =η , : ,

F20 y2i , z2 ,v2i( ) =

r232i + r21

2i y21i + p2ij

j =1

Mi

∑ r212ijη1

2ij⎛

⎝⎜ −

p1ii=1

N

∑ v1i⎞

⎠⎟ z21 − r22

2i z212 −

y22i + p2ij

j =1

Mi

∑ η22ij

⎛

⎝⎜

⎞

⎠⎟ z22 −

p2ijj =1

M1

∑ r242ij + r28

2( )v2ij

v2i = v2i1,...,v2iMi( )

z2i y2

i( )y2

i

η21i

η11i

(32)

where v2ij are potential customers of the VNO,which are lost due to the lack of capacity for theservice provision during Period 2 under scenarioi of Period 1 and scenario j of Period 2. Thestructure of the profit function and constraints(28)-(32) is very similar to the prediction modelfor Period 1.

Observe that both prediction problems arequadratic programming problems which areeasily solvable with standard software.

7 Implementation andNumerical Experiments

The methodology described above underlies adecision support system for analysis of strategicand tactical decisions in a competitive telecom-munication environment, which is currentlyunder development. The top-level architectureof this system is depicted in Figure 6. It consistsof the following main components.

• Spreadsheet front. It is used for data entry andcommunication with user.

• MATLAB engine. It provides connectionbetween blocks, common implementationplatform, quick prototyping capability for newmodels, and customised top-level algorithmsfor model solution, scenario generation andpostprocessing.

• Model suite. Contains a set of models basedon the approach described in the previous sec-tions.

• Problem solvers. This block utilises commer-cial and custom developed solvers for modelsolution and analysis, in particular commerciallinear and non-linear programming solvers.

We present here some results of experimentswith several models from the Model suite of thissystem. All models are variants of the modelspresented in sections 3,4,6 and are implementedin Matlab. The quadratic programming solvernecessary for solution of prediction models ofvirtual operator (21)–(26) and (27)–(32) wastaken from Matlab optimisation toolbox.

We illustrate here several capabilities of the sys-tem and in particular

• Graphical presentation of enterprise perfor-mance and other important characteristics asthe function of decisions. This can help con-siderably in the decision process because itallows analysing the impact of decisions on

k22ij + r21

2ij y21 − r222ij z21 ≥ 0,

i = 1:N , j = 1:Mi


performance having much more informationthan just the optimal values of decisionparameters derived from the solution of model(11)–(20).

• Analysis of dependence of results on differentimportant characteristics of the decision prob-lem. Some examples of such characteristicsare market parameters, degrees of uncertainty,and parameters of enterprise and competitionmodels. This analysis is particular importantdue to the imprecise nature of the input data.

Two sets of numerical experiments with twomodels are presented here: the single periodstochastic model and the two period stochasticmodel with capacity expansion.

7.1 Single Period Stochastic ModelThis is the model which is obtained from themodel described in Section 6 by discarding allthe elements which refer to the second period.

In particular, in the objective function (11) thesecond term is eliminated together with the pre-diction model of the second period (27)–(32).Results are presented on Figures 7–11. Theyshow the effect of the Virtual Network Opera-tor’s uncertainty about the Network Operators’price decisions and of customers’ sensitivity toboth providers’ prices.

Price uncertainty denotes the quality of informa-tion flow in the market, in this case the knowl-edge of the virtual operator about the networkowner’s prices. It is expressed by the randomparameters that represent uncertain knowledgeof the VNO about the Network Operator’s ser-vice price and his price for leasing capacity.This uncertainty enters the prediction model

(21)–(26) through the parameters and .

Market sensitivity is expressed by customer flowfrom or to operators in dependence of the singleoperators’ prices. This customer flow is deter-mined by the help of parameters in customermodel denoting the amount of customers migrat-ing between the operators, the new subscribersor those which abandon the service. In terms ofthe model of section 6 these are the parameters

. A highly sensitive market is

characterised by the high values of these param-eters.

The figures show dependence of the profit of theNetwork Operator (in the upper left sub-graph),

k1t ,k2

t ,r11t ,r12

t ,r22t

η21iη1

1i

Figure 7 One stage stochasticmodel: the referenceparameter set

Figure 6 Top-levelarchitecture of decisionsupport system

MATLABengine

Modelsuite

Problemsolvers

Spread-sheet frontUser


Figure 8 One stage stochastic model:the case of high market sensitivity

profit of the Virtual Network Operator (in theupper right sub-graph), the VNO service price(in the lower left sub-graph) and the amount ofleased capacity (in the lower right sub-graph) asfunctions of the service price y1 of the NetworkOperator and of his leased capacity price y2. Wefixed the reference values of the model parame-ters and changed the level of the price uncer-tainty and market sensitivity with respect to thereference values.

Figure 7 shows the results for the referenceparameter set. Figures 8 and 9 show the resultswith the same parameter set with the exceptionthat the values of the market sensitivity parame-ters were increased and decreased respectively.The values of the market sensitivity parameterswere fixed again to the reference level in theexperiments reported on Figures 10 and 11, butthe level of price uncertainty was increased anddecreased respectively.

Figure 9 One stage stochastic model:the case of low market sensitivity


Figure 10 One stage stochastic model:the case of high price uncertainty

Figure 11 One stage stochastic model:the case of low price uncertainty

We will now have a closer look on the four sub-graphs in each of the figures. Our explanation isbased on Figure 7, but similar analysis can beeasily made also for all the other figures.

Let us consider the uppermost left sub-graph,which depicts the profit of NO as the function ofhis service price y1 and leased capacity price y2.The profit functions of NO consist of two main

components: service provision and capacity leas-ing. Whereas the profits caused by the capacityleasing part depend mainly on the NetworkOperator’s capacity prices, the profits of serviceprovision depend on the Network Operator’s ser-vice prices. This means we can give the follow-ing description of the main areas in the uppertwo sub-graphs:


• VNO leases capacity, NO has positive cus-tomer number (the front part in the graphwhich corresponds to low service prices y1and low capacity prices y2 of the NO). Thisis the mainstream situation.

• VNO leases capacity, NO serves no customers(left part of the graph, for high service pricesy1 and low capacity prices y2 of the NO). Thiscorresponds to the case when the service pricedifference between two operators is so highthat all customers go to VNO and NO obtainshis profit only by leasing capacity to VNO.

• VNO leases no capacity, NO has positive cus-tomer number (right part in graph display, forlow service prices y1 and high capacity pricesy2 of the NO). This corresponds to the situa-tion when VNO is driven out of the market byNO, who applies prohibitive leasing price.

• VNO leases no capacity, NO serves no cus-tomers (back part, for high service prices y1and high capacity prices y2 of the NO). Thiscorresponds to the situation when technologyis not capable yet to provide a service at rea-sonable price and the service does not takeoff.

These areas and the respective dependencies canalso be found in the graph of the Virtual Net-work Operator’s profit function (upper right sub-graph).

A look at the service price decision z1 of the Vir-tual Network Operator (the lower left sub-graph)shows that she nearly always offers her serviceat a lower price than the Network Operator; withincreasing prices y1 of the Network Operator thisprice difference even increases. The model ofcustomer behaviour includes among others cus-tomer migration due to the price differencesbetween the providers. A comparison of the cus-tomer numbers of each operator shows that theNetwork Operator has a positive customer num-ber and the Virtual Network Operator has nocustomers to serve when the Network Operatorcharges low service prices whereas these rela-tions are reversed when the Network Operatorcharges high service prices. However, the Vir-tual Network Operator maybe cannot serve alldemand of the customers because she has onlylimited amount of capacity. For this demand thatshe cannot meet, she faces opportunity costs thatdiminish her profits. This effect is especially vis-ible in the rear part of the graph depicting theVNO profit function (upper left sub-graph): TheNetwork Operator capacity prices are so highthat the VNO does not lease any capacity any-more and thus can not serve any customers. Theonly component in her profit/loss function isthen the opportunity costs.

The profit functions show now only the profitsof the Network Operator and the Virtual Net-work Operator that would be obtained for certainpricing decisions of the Network Operator. Withthe given set of parameters we observe that theVNO often has a negative profit. This means hecould stay on the market only for low capacityprices and high service prices. With another setof parameters (e.g. a higher general price level qor other opportunity costs) the area of nonnega-tive profits of the VNO could be extended, sothat she takes part in the market also when theNetwork Operator charges higher prices forcapacity.

The amount of capacity leased by the VNO fora certain pricing policy of the Network owner isinfluenced by a trade-off between costs for lostcustomers and the costs for capacity leasing. Forlow capacity prices the virtual operator wouldlease even more capacity but she is limited bythe decisions of the Network Operator.

This capacity limit also gives some effect on theVNO decision on her service prices; the edgevisible in the sub-graph depicting the serviceprice decision coincides with the limits on theamount of capacity available for leasing.

The irregularities that are visible in the graphsillustrating the Network Operator’s profits(upper left sub-graph) and the decision of VNOabout the amount of the leased capacity (lowerright sub-graph) are due to the fact that uncer-tainty about opportunity costs, customer move-ments, price uncertainty etc. was represented bya finite number of scenarios. Continuous distri-butions will give smoother graphs, but will cre-ate problems with numerical solution of relatedoptimisation problems.

Reasonable economic interpretations can befound also for finer details on the graphs. Let uslook again at the upper left sub-graph, whichdepicts the profit function of the Network Oper-ator. The crate visible approximately in the mid-dle of the graph depicts the highest possibleprofit that can be obtained by leasing capacity torivals. If the capacity price is chosen higher theVirtual Network Operator will decide to leaseless capacity. The saddle in the fore right area ofthe graph denotes the service price which willcause highest possible profits obtainable fromservice provision to own customers. When theNetwork Operator chooses a higher service pricehe will lose customers to his rival. Even here thedifferent parts of the graph will change withanother choice of the parameter set, so that forexample the maximum profit obtainable bycapacity leasing and the maximum profit obtain-able by the service provision show different rela-tive height.


7.2 Simplified Two Period StochasticModel

This model is informally described in Section 4.It is obtained from the model of Section 6 by fix-ing the second period price decisions of the Net-work Operator to his prices during the firstperiod. Figure 12 shows the results with the ref-erence parameter set where the network exten-sion was fixed to zero, while in the case shownon Figure 13 the network extension was fixed toa given positive number.

7.3 Analysis of ResultsThe profit function of the Network Operator asdepicted above is influenced by several compo-nents and has therefore a complicated shape thatmakes a mathematical treatment difficult. Theimportant effects that come into play are

• Number of customers purchasing service fromthe NO;

• Amount of capacity leased by the VNO.

Market sensitivity. The easier customers willmove between the single providers the morecapacity is reserved to meet the demand and thelower are the service prices. This strategy willensure to obtain as many customers as possible.The highest profits for the Virtual NetworkOperator are achieved at a low level of marketsensitivity with high service prices and lowcapacity prices. In the case of highly sensitivemarket the Network Operator obtains highestprofits by choosing high capacity and service

prices. In the case of low market sensitivity thepart of the profits due to the service provisionbecomes more important since the customers aremore loyal to him even in the presence of higherservice prices.

Price uncertainty. In the case of higher priceuncertainty, i.e. bad information flow in the mar-ket, the Virtual Network Operator will leasesomewhat more capacity than in the case oflower uncertainty and better information flow inthe market. However, her decision about the ser-vice price will be affected only marginally. Thelevel of service price will always be below theNetwork Operator’s prices, except for very lowvalues of the Network Operator’s service prices.Besides, the Network Operator can obtain higherprofit in the case if the VNO knows less abouthis policies. In the case of a longer time horizonwith two decision periods these observations arereinforced.

Some additional observations which can begiven as advice to the Network Operator forchoosing a profit maximising policy for thegiven set of model parameters:

• The case of low market sensitivity: choosecapacity price at approx. 75 % of the limitand a high service price;

• The case of good information flow, i.e. lowprice uncertainty of the VNO: choose highservice price and medium capacity price;

Figure 12 Two periods, referenceparameter set, no network extension


• The case of the very bad information flow /high uncertainty: choose capacity priceapproximately at the middle of the region,choose the service price either high or low. Inthe first case the profits are obtained by leas-ing capacity, in the latter case by service pro-vision to own customers.

These considerations are given here as an exam-ple of the kind of recommendations the decisionmaker can get by using modelling system. Fordifferent values of the model parameters theserecommendations may change.

8 Maximising Market ShareIt is conceivable that the service providers pur-sue some other aims than the maximum profit.One of such aims may be the largest possiblemarket share. Generally, it is defined as the salesof a company in a defined market compared withtotal sales in that market. Therefore, here themarket share is based on the amount of servicesold to the single providers’ customers, i.e. theserved demand. The models, which are similarto the models in the case of the profit maximisa-tion, can be formulated. For example the enter-prise performance measure of the NetworkOperator will have the following shape for thesingle period case:

F1 = Number of customers served by NetworkOperator / Total number of customersserved by both operators

The Virtual Network Operator’s market sharefunction is constructed similarly. Besides of thedifferent objective functions the models are setup on the same premises as for the models withprofit maximising operators. However, we haveto bear in mind that in the case of uncertainty theoperators have to estimate also the number ofcustomers of their rival in order to determinetheir own market share. Thus the Network Oper-ator’s model for the prediction of the VNO strat-egy includes a term for her estimated customernumber and a term for the NO customer numberperceived by the VNO. The total number ofserved customers may therefore be estimateddifferently by both operators depending on thequality of information flow in the market.

The models as formulated here are purely marketshare oriented, there is no dependency on thecapacity price anymore and all economic aspectswere put into background. To be more realistica combination of both presented ideas may beappropriate. Another important feature of themarket share approach is that the goal functionsare not quadratic anymore, and a solution tech-nique will therefore be more complicated.

9 Conclusions and FutureWork

A framework was developed for modelling ofcomplex competition relationships and evalua-tion of strategic decisions in the telecommunica-tion environment. Its methodological foundationdraws upon selected ideas of the game theoryand the stochastic programming.

Figure 13 Two periods, referenceparameter set, network extension by

5000 units


[Gai88] Gaivoronski, A A. 1988. Implementa-tion of Stochastic Quasigradient Methods. In:Numerical Techniques for Stochastic Optimiza-tion. Ermoliev, Y, Wets, R J-B (eds.). Springer.

[Gai98] Gaivoronski, A A. 1998. Agent Nets: amethodology for evaluation of enterprise strate-gies in information economy. Telektronikk, 94(3/4), 33–48.

[Gai99] Gaivoronski, A A. 1999. Modeling ofComplex Economic Systems with Agent Nets.In: Banzhaf, W et al (eds.). GECCO-99: Pro-ceedings of the Genetic and Evolutionary Com-putation Conference, July 13–17, 1999, Orlando,Florida, USA. San Francisco, CA, Morgan Kauf-mann.

[LS92] Leleno, J M, Sherali, H D. 1992. Aleader-follower model and analysis for a two-stage network of oligopolies. Annals of Opera-tions Research, 34, 37–72.

[MW95] Mas-Colell, A, Winston, M D. 1995.Microeconomic Theory. Oxford UniversityPress.

[QR01] Quan, V, Rogers, J S. 2001. AdvancedIntelligent Network Vendor-Telco-CustomerSupply Chain: Analyzing the Effects of NewVendor Entrants. In: Proceedings of the 9thInternational Conference on TelecommunicationSystems, Dallas, Texas, USA, March 2001.

[SL88] Sherali, H D, Leleno, J M. 1988. Amathematical programming approach to a Nash-Cournot equilibrium analysis for a two-stagenetwork of oligopolies. Operations Research, 36(5), 682–702.

Decision support system based on this frame-work is under development. In this paper we uti-lized it for analysis of interactions between net-work operator and virtual operators. Although atpresent the models have a quite simple structure,they allow obtaining fairly interesting and non-trivial insights into this situation.

The further work that we are planning is orga-nized along the following interrelated directions.

• Further development of the model suite. Itincludes development of more sophisticatedcustomer models with feedback and more dif-ferentiated pricing structures.

• Study of the mathematical properties of themodelling framework. This will provideinsights into the structure of the optimalstrategies and facilitate the development ofefficient solution techniques.

• Implementation of the architecture of the deci-sion support system. The objective is to createa user friendly and robust tool suite for evalu-ation of strategic decisions in a competitivetelecommunication environment.

Literature[Bin92] Binmore, K. 1992. Fun and Games. AText on Game Theory. D C Heath.

[BL97] Birge, J R, Louveaux, F. 1997. Introduc-tion to Stochastic Programming. New York,Springer.

[BEG98] Bonatti, M, Ermoliev, Y, Gaivoronski,A A. 1998. Modeling of multi-agent market sys-tems in the presence of uncertainty: The case ofinformation economy. Journal of Robotics andAutonomous Systems, 24, 93–113.

[EW88] Ermoliev, Y, Wets, R J-B (eds.). 1988.Numerical Techniques for Stochastic Optimiza-tion. Springer.

65

1 IntroductionThe purpose of this paper is to analyze howoperators and customers may benefit from open-ing the value-chain in telecom and establishingvirtual network operators (VNOs). Virtual net-work operators are facilities-less service pro-viders that own and control only a smaller coreof selected network devices and service applica-tions, while contracting out the remaining net-work and application services that are necessaryto provide the final service bundle to customersas promised. Although net social benefit may bepositive, incumbent will generally benefit lessand consequently be less receptive to virtual net-work operation than customers and new entrants.If all were likely to benefit, virtual network oper-ation should emerge naturally. Since some standto benefit more than others, virtual operationwill not appear as viable strategy until facilitat-ing regulation and appropriate operating condi-tions have been established. In this paper ourfocus will be more on operating conditions thanon facilitating regulation.

First of all, to profit from virtual business prac-tice, the price difference between final user priceand rental price must cover more than the VNO’sown service production cost. Generally, thismargin is meagre, and the chances for extraprofit smaller, when prices and services aretightly controlled by the network operator than

when they are less tightly controlled and corre-spondingly more influenced by the VNO itself(or by the regulator on behalf of the VNO).Under certain conditions concerning servicetransactions, production resources, organiza-tional capabilities, competition and governance(to be further laid out below), this separationmay contribute both to more efficient utilizationof existing network resources, and to more effi-cient development of new networks and ser-vices. Many still tend to believe, however, thatthe net benefit of virtual network operation ishighly uncertain and probably negative.1) Themain reason for such scepticism seems to be theexpected negative incentive effects that such aregulation may have on investment in future net-works such as in 3rd generation mobile network(UMTS network), combined with weaker infras-tructure competition. Although these negativeincentive and competition effects will be moder-ated by the positive effect that such regulationwill have on service innovation and competition,sceptics (most incumbents and some regulators)expect positive effects to be outweighed by neg-ative ones. However, the negative incentiveeffects on infrastructure investment will dependon both the financial arrangement chosen forbuilding and operating such networks and theindustrial dynamics released by opening thevalue-chain in general, and introducing virtualnetwork operators in particular. My intention

The Future of Virtual Network OperationLessons from the ICT-IndustriesS V E I N U L S E T

This article describes and analyzes critical conditions for achieving net benefit from opening the value-

chain in telecom by introducing virtual network operators (VNOs). These are facility-less operators that

outsource most of basic service production, while focusing on a smaller core business consisting of

intelligent network operation, innovative services and creative marketing. Whereas granting the VNO

significant influence both over service offerings and price margins is one essential condition for

profitable return, this is far from a sufficient one. Not only should technical interfaces between networks

and between higher and lower technology layers be standardized to facilitate trade in the use of such

resources, but contracts should also be designed that combine (i) incentives for future network invest-

ment and service provision among facility-based network operators with (ii) conditions that do not place

undue service and pricing restriction on facility-less operators when renting those networks. The major

argument for designing such contracts, however, is not primarily transaction cost minimization in the

static sense, although this is also an issue, but rather the subsequent learning and industrial dynamic

effects released by such contracts. By servicing a broader range of demanding external service provider

customers, facility-based operators will also be exposed to a broader range of service related problems,

stimulating the search for more innovative and customer friendly network solutions. Facility-less network

operators, on the other hand, may develop into market-led service providers that no longer can rely on

subsidies from their facility-based sister affiliates, but will be forced to develop profitable service and

marketing solutions for a wide range of demanding end customers.

Svein Ulset (49) holds a doctoraldegree in Business Administra-tion and is currently associateprofessor at The NorwegianSchool of Economics andBusiness Administration. Hisresearch interests are in organi-sational economics and corpo-rate strategy, with a specialfocus on the telecommunica-tions sector and multinationalcorporations. He has publisheda number of articles and reportsdealing with strategy and organ-isation of telecommunicationcompanies.

[email protected]

1) This was the conclusion reached by the British regulator Oftel and Norwegian Department of Communicationthat recently evaluated the concept of virtual operators in mobile communication.

Telektronikk 4.2001


with this paper is to start a more open and criti-cal assessment of this difficult, but very impor-tant question.2)

2 The Essence of VirtualBusiness Practice

2.1 The Core Issue of SeparationAs indicated above, the crucial question iswhether basic network services and enhanced(advanced) services are technologically separa-ble, and if separable, whether they still are toointerdependent to justify full corporate separa-tion. On the other hand, if basic and enhancedservices are technologically separable, but onlymoderately interdependent, what could then bethe most efficient mechanism for safeguardingthe underlying investment in physical infrastruc-ture, supplementary resources and service capa-bilities? Would simple contracting suffice, orwould more hierarchical structures be requiredsuch as strategic alliances or long-term exclusivecontracts? Transaction cost economics, the pre-mier theoretical approach in analyzing suchproblems, will serve as our theoretical guide,here applied in a somewhat more resource-cen-tric and dynamic fashion than usual (William-son, 1991).

As indicated in Figure 1, positive cost and/orincome performance is achieved by aligningtransacting actors who differ in their personaland transactional attributes with organizationalforms (market, hybrid, firm) which differ in theirgovernance mechanisms (incentives, administra-tive control and contract laws). Compared to thefirm (as one out of three alternative genericforms), the market will be characterized bystronger economic incentives, weaker adminis-trative controls and stronger reliance on the court

system as conflict resolution mechanism (Will-iamson, 1991). Hybrid forms such as joint ven-tures fall in between firms and markets.

In general, transaction hazards arise whenincomplete contracting is carried out betweenopportunistic, bounded rational and interdepen-dent actors. As interdependencies and the associ-ated transaction hazards decrease, integratedfirms (fully integrated facility-based operators)are gradually replaced by hybrid forms andarm’s length contracting. That is, facility-less“virtual” network operators that buy basic net-work services from facility-based “non-virtual”operators will normally do this under incompletecontracts that specify some, but not all the obli-gations under some, but not all future conditions.Instead of specifying every possible future deci-sion and condition, difficulties are dealt with,and conflicts resolved, as the future unfolds.This may work reasonably well under most ordi-nary supply contracts as long as problems aresimple and easy to solve and potential lossesfrom switching partner are small. In situationswhere problems are more difficult to solve, andswitching costs are large, simple contracts willno longer suffice. Should one of the parties, dueto possible failures or defects by the other partywish to exit from the relation, this may not onlylead to time-consuming and costly conflicts,arbitration and possible litigation, but also to theloss of all the non-redeployable assets. To avoidsuch transaction costs, simple contracts shouldbe replaced by more complex contracts withstronger safeguards, such as long-term contract,joint ventures or fully integrated corporations,dependent on the level of contractual difficultyand the size of potential losses from separation(the choice of integrated corporation beingreserved for the most difficult cases with highestloss potential).

2.2 Defining Virtual NetworkOperators

Virtual network operators (VNOs) are operatorsthat own and control a minimum of higher-layernetwork infrastructure while contracting out theremaining lower-layer infrastructure (cables,switches, etc.). Alternatively, a VNO may bedefined as an enhanced service provider thatowns and controls only those facilities that pro-duces the enhanced features that may differenti-ate its services from those of its competitors,whereas all the remaining basic infrastructure is

2) The more general question of outsourcing also apply to upstream activities such as network building and mobile tower operation as reported by TELE-COM A.M., a daily news bulletin from Warren Publishing Inc: “Nextel signed a master lease agreement November 5 with Crown Castle Internationalvalued at $250 million over 10 years for colocation on 1,264 wireless sites on Crown Castle's towers. Colocation will begin immediately and continueover three years, with Crown Castle expecting annualized leasing revenue of $23-$27 million under the 25-year contract, the companies said. As partof the contract, Crown Castle said it will provide installation services for all the sites. The agreement with Nextel comes after a series of contracts thatBell Atlantic, BellSouth, Powertel and the U.K.'s One-2-One have signed with Crown Castle to outsource their tower networks,”(TELECOM A.M., Vol.5, No. 215. November 8, 1999).

Figure 1 The core issue

Actors- bounded rational- opportunistic

Transactions- interdependent- incomplete

Organization- market: virtual- hybrid: semi-virtual- firm: non-virtual

Performance- cost- income

Transaction hazards


rented, or the associated network service bought,from facility-based operators. Since intelligentfacilities and service applications are the primarymeans by which VNOs make revenue and prof-its, these facilities and applications should alsobe owned and controlled by VNOs. Normally,expected profit margins will be lower on trans-port and access service that virtual operatorsonly rent or buy, than on intelligent nodes andapplication that the operators privately own andactively manage, unless prices are set below costby the regulator.

Physically, the VNO’s own network devicesand service applications can either be hosted bythe incumbent or by the VNO. In the latter case,VNO devices and applications are connected tothose of the incumbent across more or less stan-dard interfaces, making the former more or lessindependent of the latter. To create a virtual net-work operation, access not only to the incum-bent’s physical network but also to his Opera-tional Support Systems (OSS) is required. Theseare computer databases and systems that provideservices and network management, administra-tion, planning and repair functions, as well asfunctions related to customer operations, suchas customer care and billing. In other words,whereas network facilities and service applica-tions are the hardware and software needed toproduce and deliver telecom services to finalusers, interface standards are specifications thatmore or less seamlessly interconnect the hard-ware and software of the VNO with those of theincumbent or other complementary networkoperators, making the former interoperable withthe latter. In this respect, a full and unconditionalopening of the value-chain essentially meansturning previously closed and proprietary inter-faces into open and non-proprietary ones, whilesimultaneously offering network facilities andservice applications that newcomers can affordto rent or buy. Under exclusive contract, thevalue-chain is closed to all except the selectedpartner, whereas the underlying core technologywith associated interface may still be more orless proprietary. Except for the renting and pric-ing issue, opening the value-chain in telecoms isstrikingly similar to opening the value-chain inthe computer industry. In particular, as telecomsconverge with computing, lessons from the com-puter industry may increasingly become relevantfor the telecom industry.

Also facility-based operators are forced to buynetwork services from each other in situationswhere subscribers of one network call sub-scribers of other networks. By renting or leasingnetworks, costly duplication is avoided, existing

infrastructure more fully exploited, and servicesprovided at lowest possible production costs perunit (economies of scale). Only substantialincreases in transaction cost between competingnetwork operators should prevent such rentingand leasing contracts from replacing networkduplication. Thus, virtual network operators thatown only a minimum of infrastructure will onlysurvive to the degree such economies of scalefrom higher utilization of the larger network canbe attained just as well by selling to external asto internal service providers; that is, by separat-ing network operation from service provision.Since this cannot be excluded, some incumbentshave already started to explore the profitabilityof VNOs more systematically, especially in for-eign market where the roles are reversed, and theneed for an alternative VNO-strategy is morepressing.3)

2.3 The Separation IssueDividing integrated telecom enterprises into dis-tinctive businesses, separated by technical andcontractual interfaces, has for quite a while beenthe standard approach for introducing competi-tion in the telecom sector, strongly promotedby regulator and new entrants, and graduallyexcepted by incumbents. Considerable differ-ences still exist, however, concerning whichactivities to separate and to what extent. Whileregulators and new entrants seem to prefer thattechnologically separable activities should alsobe organized as independently owned busi-nesses, most incumbents are strongly againstthis. Their objection would generally be thatmost activities that seem to be technologicallyseparable, are considerably less so, and of thosethat are, many are still too interdependent to jus-tify full corporate separation. Forcing corporateseparation between activities sharing the use ofcommon non-redeployable and/or non-tradableassets would simply cause transaction costs toexceed the associated separation benefit (posi-tive competition effects). Incumbents have there-fore been considerably less reluctant in accept-ing the location of interconnection points outsidesystems of highly interdependent or non-tradableresources than inside these systems as illustratedby local loop unbundling and mobile networkroaming initiatives (Ulset, 1998a).

Recent interconnect arrangements, especially ofthe less intervening kind, have not only made iteasier for telecom brokers and resellers to buyand sell excess capacity, but also enabled newentrants to seamlessly interconnect their smallernetwork with the larger network of the incum-bent. Now, as the telecom technology has turnedmore complex, higher technical layers have been

3) This led Telenor and BT to develop a business plan for a software-based telco, named Facet, which illustrateperfectly the VNO concept (Halbo et. al., 1999a, 1999b).


added to the basic infrastructure, and more spe-cialized service and support functions havedeveloped, the question of separation has alsobeen extended to include these layers, servicesand functions. Dividing the infrastructure intoindependent layers would help to create technol-ogy-independent services and service-indepen-dent technology. It would also make it possiblefor independent operators to specialize on differ-ent layers and to combine their services, undersome kind of bundling contract, into more orless customized packages for sale to final users.

Whereas the regulator have regarded corporateseparation as a last resort measure for develop-ing competition, incumbents have claimed thatsuch separation will seriously weaken their tech-nical abilities and economic incentives todevelop new services. Their argument would bethat network facilities, although technically sep-arable, are still too interdependent, or too diffi-cult to rent out, to justify full corporate separa-tion. Albeit less openly debated, a more threat-ening possibility would be that such regulatoryenforced separation would enable virtual opera-tors to take full control, not only over value-added services, but also over the customer base,thereby reducing physical network operation to alow-margin commodity business (pure trans-porters of non-differentiable bit streams).

As technologies and markets have kept evolving,the issue of separation has also become morecomplex. This relates not only to organizationalseparation which may vary in degree fromaccounting to full corporate separation, but alsoto technological separation which may vary interms of effectively separable layers, from thelowest physical substrate layer to the highestfinal user applications layer.4) Due to the enor-mous complexity of modern telecom networkand services, along with the large share of non-distributable sunk cost, defining and pricingunbundled elements and services unambigu-ously, has been very difficult, almost impossible.If the infrastructure were forced open by regula-tors, incumbents would fear that the associatedprice for renting or leasing network elementswould no longer cover their cost, thus destroyingtheir incentives for investing in new infrastruc-ture and in the development of new services.No wonder that initiatives such as local loopunbundling, asymmetric roaming and virtualnetwork operation have been strongly rejectedby most incumbents.

2.4 The Competitiveness Issue

2.4.1 Competitive Advantage in generalTo answer the above separation question, wefirst need a clearer understanding of possiblesources of competitive advantage and how thesemay be affected by separation. First of all, toattain competitive advantage in general, compa-nies must be able to develop and deliver prod-ucts or services at lower price or higher qualitythan their competitors. This can be achieved bydeveloping superior functional competence (e.g;technology, management, marketing etc.), bybuilding stronger market positions (e.g. domi-nance, legal patents, exclusive licenses etc.), orby some combination of the two (i.e. lower costdue to earlier entry and deeper learning overlonger time). For example, the previous mono-poly operators could operate their nation-widenetworks at lower accounting cost per minutethan their competitors, not only due to over-depreciation during the previous monopolyperiod, but also due to superior competence inrunning those network accumulated over alonger period of time.

Then, superior competence can be used partly toreduce production costs, partly to develop morefunctional products at higher price. Whereas inthe first case, competitive advantage and extraprofit is the outcome of cost leadership, in thelatter case this is the result of exceptional perfor-mance in product differentiation and innovation.In both cases, sustainable surplus is created byassets that not only are superior compared tothose of the competitors, but also inimitable orotherwise protected from leaking out to competi-tors. In the cost leadership case, these resourcesare used to produce an increasing number ofidentical or closely related products at morerapidly declining unit costs than could beachieved by competing operations. In the inno-vation leadership case, common resources areused for developing faster or less costly pro-cesses than those of competitors. To the degreespecific resources that contribute to the develop-ment of cost leadership or innovation leadershipare difficult to trade in ordinary markets, com-petitive advantages will be more rapidly devel-oped and more efficiently exploited by units ofthe same enterprise than by comparable unitsorganized as independent suppliers in the market(or as part of another and differently diversifiedand integrated enterprise).5)

4) Neither should telecom services include only network services made separable through initiatives such as localloop unbundling, mobile network roaming and the widespread use of Internet technology. Relevant servicesshould also contain distribution of information, education, entertainment and electronic commerce over theInternet, as well as the full range of specialized downstream customer services and support such as wholesale,retailing, customer support, system integration, outsourcing and consulting, the latter being probably moreseparable from network infrastructure than content distribution.


In other words, the non-tradability of the ser-vices associated with the development and thesubsequent utilization of critical resources is themajor condition for corporate integration. Notethat “non-tradability” is here used compara-tively, meaning that development and/or utiliza-tion of these resources is considerably morecostly or time-consuming to carry out by inde-pendent firms in a market than by integratedfirms. This distinction corresponds of course tothe more general distinction in transaction costeconomics between autonomous and coordinatedadaptation (Williamson, 1991) and, in particular,to the distinction between autonomous and sys-temic innovation in the innovation literature(Teece, 1986; Chesbrough and Teece, 1996).In the systemic case, technical improvements orinnovations in single components would requiresimultaneous changes of the other componentsmaking up the larger system. Under autonomousinnovation, however, individual components canbe replaced by new ones without negative effectson the other components or on the system as awhole. In the autonomous case, efficient down-stream service provision will develop withoutthe downstream service providers having theslightest knowledge of upstream productiontechnology and capabilities. In the systemiccase, however, efficient downstream service pro-vision would require that downstream serviceproviders utilize the technology and capabilitiesof upstream network operators. To achieve suchknowledge sharing corporate integration may beneeded when critical knowledge is basicallytacit, sticky and dispersed over a larger numberof locations and team members. If less tacit,sticky and dispersed, sufficient sharing may beachieved by buying from the same supplier oradopting the same basic technology platform.

2.4.2 Competitive Advantage in VirtualNetwork Operations

In general, by focusing on a smaller core ofresources, virtual as well as non-virtual operatorsmay achieve extra profit to the degree theseresources are valuable, rare, inimitable and wellorganized (Barney, 1997, chap. 5). Resourcessuch as physical and human assets are valuableto the degree their effects on profit or valueadded are potentially large. They are rare in thesense of being scarce (few competing suppliers),different (rather unique compared to similarassets at the competitors’ disposal), non-substi-tutable (no alternative kind asset or technologyto perform the function) or specific (less produc-tive value for alternative users or uses). Beingrare in the absolute sense means that no useful

substitute exists (e.g. telephone lines before cel-lular and satellite). Resources such as technol-ogy, knowledge or competence may be difficultto duplicate or copy either for “technical” rea-sons: to the degree they are invisible, tacit,sticky or diffused; or for economic reason: to thedegree they constitute a natural monopoly orexhibit extreme degree of economies of scale orscope.

Given that the above resource conditions are ful-filled, appropriate organization and managementare still needed to transform potential sources ofcompetitive advantage into sustainable competi-tive advantage with above-normal return. A rent-ing/leasing contract may suffice, but only to thedegree competitively critical resources of thevirtual operator can easily be separated fromthose of the non-virtual operator, and otherwisemade tradable through some kind of leasing orrental contract. If not, corporate integrationwould replace contractual governance as themost efficient governance form. Consequently,integrated operators should also replace a systemof virtual operators buying network service fromfacility-based (non-virtual) operators.

In Figure 2 the arrows linking organization,essential assets and performance summarizethese relationships. Here the term “essentialassets” are used because of the close relationbetween these assets (resource-based competi-tive advantage) and the “essential facility” con-cept in the telecom regulation literature. Thelatter are facilities that (i) are controlled by amonopolist, (ii) are considered a necessary inputfor the provision of downstream services, and(iii) cannot be duplicated in any technical or eco-nomical feasible way. According to the received

5) Resources such as network and competence in running the network can be traded either as objects in which case the respective physical or humanassets are permanently sold, or as services in which case they are only rented out for a specified period of time. When competence is rented out, theobjective may sometimes be to transfer or duplicate competence to the benefit of the client, in which case the competence both is rented out and sold.

Figure 2 The competitive issue

Essential assets- valuable- rare/unique/non-sub.- non-tradable I & II- inimitable (otherwise)

Organization- market- hybrid- firm

Performance- cost- income

Transaction hazards

Regulation

ICT convergence

non-tradable I: specific/tacitnon-tradable II: leaky


doctrine, appropriate candidates for regulationare only those facilities that are “essential” in theabove sense because of the power abuse poten-tial (and associated super-normal profit poten-tial) such facilities represent.

The existence of independent VNOs, however,poses a rather crucial dilemma. By specializingin the more advanced higher-layer features, vir-tual operators may on the one side reduce lower-layer physical operation to a low-profit com-modity business without leaving sufficient sur-plus to finance next-generation infrastructureinvestment. On the other hand, by specializingon more advanced features, virtual operatorsmay also contribute to a higher innovation ratesin the deployment of higher-level applicationsand development of value-added services.

To reach expected growth in advanced servicesand content, large-scale investment in broadbandcapacity will be needed both in the transport andaccess network. Most observers agree, however,that the chances of achieving higher margin bymeans of product differentiation is far better onadvanced service and content than on transportservices and local access. Being deeply embed-ded in, and practically inseparable from, thelarger physical network, service developmenttools will for some time remain an integral partof the installed IT-systems. This, however, maygradually change as the older circuit-switchingnetwork is replaced by next generation packet-switching IP-network. The Internet technologymay then cause intelligence in service develop-ment to be transferred from the IT-systems inthe network to intelligent servers and computersbelonging to virtual operators and final users,thus making the underlying physical networkbasically dumb (and correspondingly unprofit-able).

Which company structure will be best positionedor endowed to achieve such differentiation ad-vantage on the next generation IP-platform –large integrated enterprises or networks ofsmaller specialists – is still unsettled, althoughmost operators and observers seem to believe inthe large-scale integrated structure as the mostefficient and profitable one. For example, localincumbents still prefer merging with other in-cumbents, so that larger synergies and cost re-duction (5 – 10 %) can be attained, instead ofcompeting on duplicated facilities and personnelin each other’s home markets. The synergy frommerging, however, can seldom be achieved with-out negative competition effects, including theattended risk of rising costs in the longer run.At least from a welfare economic point of view,

mergers should not be recommended unless thepositive synergy effects substantially exceednegative competition effects.

The most efficient alternative to compete on thebasis of duplicated resources may, however, notbe corporate integration, but rather duplication-less competition, based on leased lines, intercon-nection, unbundling, mobile network roaming orsimilar shared network usage. In particular, tothe degree virtual operators develop into a ratherefficient solution for new service creation andnetwork capacity utilization, the synergy effectsof mergers may easily turn negative. Virtualoperators may win extra profit if cheap transportand access services are combined with advancedfunctionality and value-added services, producedby the VNO’s own facilities and capabilities. Onthe other side, virtual operators will probably notbe able to capture extra profit to the degree thesupply of supplementary resources and servicesare monopolized, unless the price on those re-sources and services is regulated and offered atprices below historic full-cost (e.g., equal to longrun incremental cost). At the same time, VNOswho rent, rather than own basic infrastructure,are compelled to specialize on a narrower fieldof core assets to achieve competitive advantage,the effects of which could be that additional ben-efits from higher specialization exceed addi-tional transaction costs from more dispersedand specialized operations.

2.4.3 Expropriating Economies of Scaleand Scope (Core Assets)

The crucial question is, however, not the size ofeconomies of scale and scope, but whether phys-ical operators of larger networks can preventVNOs, as well as facility-based competitors,from “expropriating” a significant share of sucheconomies. Expropriation of surplus will be pre-vented to the degree the quality is lower and/orthe price higher on services that the incumbentdelivers to external customers than to internalones. For several reasons, this may often be thecase.6) Even though the regulator officially pre-scribes and expects that the same quality andprice should be offered external customers asinternal subsidiaries, this objective may never befully attained, due to technical difficulties andsever conflicts of interests. In turbulent timeswhen technology standards, regulation and busi-ness practice are still evolving, both technicalincompatibility and conflicts of interests willnormally work against external customers.

So long as the incumbent controls the onlyfixed-line access network available, competingphysical network operators can be prevented

6) This has been the experience in USA ever since the FCC tried to implement the unbundling regime as part ofthe Telecommunication Law of 1996.


from achieving economies similar to that of theincumbent, simply by providing access servicesof lower quality or higher price to competitorsthan to himself. The conditions and mechanismsmaking this possible are the same as for virtualoperators mentioned above. In the access mar-ket, however, alternative radio, cable and cellu-lar access may soon develop into a competitivealternative to fixed lines, and when this happens,the incumbent’s first mover advantage will soonbe gone. This will subsequently also help virtualoperators that will be in a stronger position tobuy competing high-quality access and supple-mentary support services at a lower price. Netbrokers and retailers will also buy and sell net-work capacity, but contrary to VNOs, these willonly compete with physical operators on serviceprovision, not on network operation. Conse-quently, brokers and retailers will represent amore promising complementary growth opportu-nity than VNOs that most often will be regardedas competitors (Foros and Hansen, 2000).

3 Virtual Operation in ICTTo get a better understanding of the economiceffects of separation and outsourcing, let us takea closer look at the development of the ICTindustry, especially the computer industry andthe Internet.

3.1 Lessons from the ComputerIndustry

3.1.1 Open Proprietary Standards,Innovation and Growth7)

Until the late 1970s nearly all computers werelarge machines used for mind-numbing calcula-tions and bookkeeping, mostly bought by largerorganizations that could afford their price andservice costs. Computers such as mainframesand minicomputers were the most complicatedmachines ever produced. They were sold in arelatively small number, and mostly produced bylarge companies that were vertically integratedfrom basic circuitry, computer platforms andoperating system software to application soft-ware and distribution. Newcomers had a hardtime breaking into the business for several rea-sons. First, few had the resources necessary toenter at all levels simultaneously. Second, forthose who entered at one or two levels, the smallnumber of independent supplier and distributorsmade it immensely costly to operate. Third, dueto the machines’ complexity and service needs,most customers were reluctant to buy from any-one but large, established suppliers.

So far vertical integration had served two pur-poses. First, by internalizing the developmentprocess, computer makers controlled technology

leakage so that proprietary and incompatiblesystems could be developed and sold to increas-ingly captive customers. Second, by internal-izing the process, the computer makers couldalso develop and deploy firm-specific assetsmore efficiently for the production of differenti-ated products. Since each chain of productionwas vertically chained together by closed inter-face technology, bilateral monopolies had arisenbetween component suppliers and assemblersthat were more efficiently managed by hierarchythan by market contracting. Competing produc-tion systems remained therefore vertically inte-grated until challenged by a significantly moreefficient technology, the personal computer (PC)technology. With this technology installed,distributed computing and networking soonbecame a more flexible and efficient alternativeto the mainframe system for ever-more complexand power-consuming tasks. Since the new com-puters could be assembled from hardware piecesand software components supplied by indepen-dent firms in the market, the new computerindustry was never vertically integrated as theold computer industry. Coordinated by openinterface standards, interchangeable componentswere outsourced to achieve economies of scalewithout the risk of excessive transaction costs.The need for vertical integration to control tech-nology leakage and to manage firm-specificassets along the value chain decreased.

From its inception, the personal-computer mar-ket assumed a different pattern from the estab-lished industry, mainly because of rapid diffu-sion of new technology. The chip manufacturerswere now able to cram a simple version of acomputer’s central processing unit, the circuitsthat did most of the actual computing, on to asingle chip, a so-called microprocessor. Aroundthis, a small cheap machine could be assembledfrom readily available parts used to supply theconsumer-electronics industry. The most suc-cessful of all personal computers, the IBM-PC,based on Intel’s microprocessor and Microsoft’soperating system, became the industry standardfor which a large number of application softwarefirms wrote their programs. Fortunately for bothfirms, full property rights to the basic technologywere retained in the initial contracts with IBM,and due to this, a large number of chips and soft-ware copies could be sold at a premium price toa booming computer industry. A parallel rapidgrowth in compatible application software,developed by innovative third parties, createdthe “network externalities” that substantiallyimproved the value of the underlying technolo-gies. Since Intel and Microsoft through incre-mental innovations succeeded in keeping theirtechnologies both proprietary and in strong

7) This subsection is previously published in Ulset (1994).


demand, huge monopoly profits ensued. AsIBM’s losses kept skyrocketing, it became grad-ually clear to everybody that Intel and Microsofthad profited substantially more from the successof the PC than any PC-maker, including thelargest of them all until then, IBM.

As additional software, equipment and networkproducts were developed, the value of possessingan IBM-compatible computer continued to grow,and so did the sales and use of such computers.Spawned by a constant stream of technical inno-vations and improvements, mainframes and mini-computers lost out to PC-network and stand-alone PC for steadily more complex tasks. Asdistributed computing and networking continuedto replace the old mainframe system, the demandfor open interface standards and intersystemcompatibility continued to rise, stimulated by amore open system strategy. This was moststrongly demonstrated by Sun Microsystems.Soon open product standards and open networksystems were demanded by most customers andsupported by most computer makers. When opentechnology standards were supplemented withconversion programs and internetworkingtechnology, the level of interoperability increasedeven more. Consequently, most personal-com-puter makers were never vertically integrated.Separate groups of firms supplied parts, fullyassembled machines (platforms), operating-sys-tem software and application software.

By attracting a larger number of potential inno-vators, open standards contributed to the growthof network externalities, and thereby to theprofitability of the firms that controlled thesestandards, such as Intel, Microsoft and Novel.Companies that did not control technology stan-dards, including most computer makers, bene-fited less. When open and dominant standardsare owned, patented or difficult to imitate, theowner of the standards will also tend to get rich.The assembler, however, will not tend to get richunless some additional proprietary technology orcompetence is added to the system. While earn-ing extra profit became constantly harder forcomputer makers, the owners of the original andincrementally improved standards got constantlyricher. The emergence of open, but still propri-etary standards, thus fundamentally redistributedcompetitive advantages and economic profits,from computer makers such as IBM (and NorskData) to chip producers such as Intel, and soft-ware firms such as Microsoft, Novel, Word Per-fect and Lotus. Also the system-users benefitedimmensely. As long as they stayed within thedominant IBM-standard, customers could freelychoose between a growing number of high-performing low-priced computer systems, andno longer needed to be exploited by their chosencomputer maker.

However, proprietary standards will still exist,and as more companies start outsourcing a largershare of their component production, leadingcomponent technology may even develop intoworld standards. To the degree these are ownedby one supplier, considerably higher profits canbe earned after the production systems havedisintegrated than before (which was and stillis the case for Intel in microprocessors andMicrosoft in operating-system software).Although positions as profitable as those of Inteland Microsoft are extremely rare, many firmsregularly develop proprietary technology withsignificant profit potential. Even if future tech-nology standards should be less proprietary thanbefore, technological innovations within thesestandards can still be kept proprietary.

3.1.2 Dominance and AntitrustSuccesses as impressive and positions as domi-nant as those of Intel and Microsoft can seldombe attained without the active use of some kindof monopoly power. Clever tricks and ploys thatmay pass when performed by non-dominantfirms, may not pass, however, when carried outby dominant businesses. Gradually, this was alsorealized by Intel who wisely moderated its prac-tice when the US Federal Trade Commissionrequested them to do so. Microsoft, however, didnot moderate its business practice when askedto, but rather continued to punish companies thatdeveloped competing products (e.g. Netscape,Sun Microsystems) or customers that sold com-peting products (e.g. IBM, Gateway). The find-ings of the subsequent antitrust case of U.S. v.Microsoft thus unambiguously showed thatMicrosoft routinely used its monopoly power tocrush competitors, even leading the judge to por-tray the company “as nothing less than a socialmenace” (Business Week, 1999, Nov. 22: 45).

After the court officially declared Microsoft amonopolist, regulators started to discuss reme-dies of which there are two major types: onebehavioral type and several structural ones. Thebehavior remedy will require close supervisionover issues such as pricing and contracts withother companies, eventually making the govern-ment the permanent overseer of Microsoft.Being very difficult to monitor, supervision ofsuch behavior may either become overly costlyand stifle innovation, or unreliable if not fullyimplemented. Structural remedies contain sev-eral dramatic measures such as breaking up thecompany into three vertically disintegrated com-panies (operating systems, application software,Microsoft Internet business) or three verticallyintegrated Mini Microsofts (“Baby Bills”), orforcing Microsoft to auction or license out pro-prietary technology to competing companies.The question is how to punish Microsoft andstop its abusive conduct while encouraging inno-


vation and protecting the consumer. Althoughthe different remedies may help to restrictmonopoly pricing and power abuse they are notwithout costs and limits. Disintegration will notcreate competition in the market for operatingsystems, partitioning Baby Bills and auctioningWindows may fracture the Windows standard,and open-source code licensing may facilitateillegal copying. In any case, remedies can onlybe recommended if significant net benefit can beexpected. Benefit can be gained from structuralremedies both in terms of making the core tech-nology more accessible and less costly fordownstream customers, and in terms of makinginnovations in complementary products easier todevelop and more profitable to commercializefor related businesses. Significant investmentdisincentives are, however, also involved sincestructural remedies will make investments inoperating systems less profitable for the mainfirm. A final decision is still pending and mayremain so for quite a while.

3.2 Lessons from the InternetUnder the virtual network operation model,modern technology enables coordination acrossfirm boundaries as if the respective individualfirms were parts of the same enterprise. TheInternet has not only enabled the development ofthe World Wide Web and electronic commerce,but increasingly also a more advanced divisionof labor and inter-firm specialization, also inter-nationally, without the usual increase in transac-tion costs. Through Web-based searching andordering (internet/extranet/intranet), a larger por-tion of peripheral and semi-peripheral activitiescan be outsourced to external specialists, leavinga smaller and more focused portion of coreactivities to internal specialists, the effects ofwhich would be quality improvement and pro-duction cost reduction, without the usual off-set-ting increases in transaction costs. A number ofindustries are now exploring these opportunities,but nowhere are these more challenging than inthe multi-layered infrastructure of telecom ser-vice provision of which the Internet Protocol andWWW themselves are parts.

3.2.1 Open Non-proprietary Standards,Innovation and Growth8)

Internet was born about 30 years ago out of aneffort to connect together a US Defense Depart-ment network called the ARPAnet and variousother radio and satellite networks. The objectivewas to build networks that could withstand par-tial outage (like bomb attacks) and still work. Inthe ARPAnet model, communications occur byhaving computers talk to each other and ensurethat the communication is accomplished. Thenetwork itself was assumed to be intrinsically

unreliable as any part of it could disappear atany moment. To send a message, the computersimply had to put its data in an envelope, calledan Internet Protocol (IP) packet, and address thepackets correctly. The demand for networkingthen spread quickly, and Internet developersfrom US, UK and Scandinavia, responding tomarket pressure, began to put IP software onevery conceivable type of computer. By then theInternational Standards Organization (ISO) hadalready spent years designing the ultimate stan-dard for computer networking without makingmuch headway. Users, however, adopted the IPinstead. So did companies that developed work-stations for local area computer networks(LANs), allowing all computers on such LANsto access ARPAnet facilities. One of thosenewer networks was NSFnet, commissioned bythe National Science Foundation, with the objec-tive of connecting computers of major universi-ties. Due to bureaucratic and staffing problems,NSF decided to build its own network based onthe ARPAnet’s IP technology.

Demand grew rapidly until the computers con-trolling the network and the telephone lines con-necting them were overloaded. The networkneeded upgrading and professional management,and the contract was awarded to Merit NetworkInc., which ran Michigan’s educational network,in partnership with IBM and MCI. The numberof connecting networks kept growing, graduallyalso including non-IP-based networks connectedby special gateways technologies. The partici-pating networks financed by governments or pri-vate users, connect without a charge, only byadopting the open IP technology or some non-IP-technology with gateway to the Internet.The modern web servers and web browsers havespawned further growth. Although the Internetwas for several years mostly used for informa-tion exchange and marketing, new communica-tion software was gradually introduced allowingthe Internet to also carry interactive voice traffic.Already by mid-96 the Internet reached nearly 5million host computers and approximately 20million users, and its growth is still accelerating.

The ultimate authority for where the Internet isgoing rests with the Internet Society, or ISOC, avoluntary membership organization whose pur-pose is to promote global information exchangethrough Internet technology. ISOC appoints acouncil of elders, called the Internet ArchitectureBoard, or the IAB, which meets regularly to con-firm standards and allocate resources, such asaddresses. It decides when a standard is neededand what it shall be. When a standard is re-quired, it considers the problem, adopts a stan-dard, and announces it via the network. The

8) This subsection is previously published in Spiller, Ulset and Zelner (1996).


Internet Engineering Task Force (IETF) isanother volunteer organization. It operatesthrough working groups that anyone can join.These groups makes different recommendationsthat either are made available to anyone or sentto IAB to be declared a standard. These stan-dards make computers from different vendorscommunicate, favoring no one in particular,whether IBM, Sun or Macintosh.

3.2.2 The Bearer Service Concept of theOpen Data Network

In addition to its primary information purpose,the widespread acceptance of the Internet Proto-col has organizational ramifications. By imple-menting the Internet Protocol, a spanning layeris being built that separates lower-layered basicinfrastructure (network services) from higher-layered value-added applications (customer ser-vices) thus creating the basic condition for thedevelopment of separate markets. The InternetProtocol enables applications to request networkservices independent of underlying physical net-work technologies. Moreover, new underlyingnetwork technologies may either substitute for,or co-exist with, existing network technologieswithout significantly affecting the broader sys-tem, enabling so-called autonomous innova-tions.9)

In other words, the “bearer service” functionsas a technology-independent network layer thatresides above the technology substratum andenables interoperation between diverse, high-level applications and various underlying net-work infrastructure. Separation of applicationsand services from the physical network will also

imply that the creation and supply of new ser-vices can be separated, physically as well asorganizationally, from network operation.10)

As noted by Gong and Srinagesh (1996, 1997),this bearer service market may not prove sus-tainable unless competing services are differenti-ated. If not sufficiently differentiated, Bertrandcompetition will lead to destructive pricing fornetwork services with close-to-zero marginalcost. One way to avoid Bertrand competition isthrough bundling of bottom-layer transport withhigher services, closer to final customers (verti-cal integration). As facility-based companiesintegrate with others at higher layers, variablecosts rise significantly along with minimum effi-cient size. Policies that promote competition inthe provision of unbundled bearer service willtherefore eventually fall. However, since bearerservices that includes Quality of Service guar-anties for bandwidth, delays and losses, is not acommodity as pointed out by Kavassalis et. al.(1998), it can be differentiated by (i) choosingdifferent substrate technologies, (ii) designing adifferent network typology than their competi-tors, and (iii) designing a different pricing policyfor their service.11) The crucial point, however,is not whether the service can be differentiated,but whether such differentiation can be copiedor imitated by competitors (Barney, 1997).Probably it can, but more easily by facility-lessproviders (virtual network operators) that areless restricted in choosing between competingsubstrata and networks, than by facility-basedoperators that are primarily restricted to choos-ing among their own substrata and networks.12)

9) This kind of architecture separating service offerings from infrastructure facilities, is by the NationalResearch Council more generally described as Open Data Network with four levels: ... “i) at the lowest levelis an abstract bit-level service, the bearer service, which is realized out of the lines, switches, and other ele-ments of networking technology; ii) above this level is the transport level, with functionality that transformthe basic bearer service into the proper infrastructure for higher-level applications (as is done in today’s Inter-net by the TCP protocol) and with coding format to support various kind of traffic (e.g. voice, video, fax); iii)above the transport level is the middleware, with commonly used functions (e.g. file system support, privacyassurance, billing and collection, and network directory services); and iv) at the upper level are the applica-tions with which users interact directly. This layered approach with well-defined boundaries permits fair andopen competition among providers of all sorts at each of the layers”.

10) For example, the equipment supplier Ericsson recently launches such a service creation and supply solutionfor mass-market Mobile Internet, called Service Network solution (Press-release, Wed, 21 Feb 2001).

11) Such differentiation can also be achieved by using the Service Network offered by Ericsson: “The ServiceNetwork is designed to help operators and service providers bring services to market in the most efficient waypossible. For example, a third-party application developer working closely with a mobile operator will be ableto enter the operator's Developer's Zone solution, hosted within the Service Network, and log on to their ownpage. From there they can check existing services and create new applications, downloading them to the oper-ator's application server to be approved. When ready, the operator can make the application available publiclyand send targeted alerts to customers whose profiles match that of the application categories.”

12) In the same way as travel agencies owned by one of several competing airlines would be more restricted andless competitive than independent agencies

13) Since this hypothesis is more thoroughly discussed elsewhere (Ulset, 1986), only a few introductory remarksand tentative conclusions are included here. ICT means Information and Communication Technology


3.2.3 Internet and ConvergenceAccording to many observers, the extraordinaryrapid diffusion of Internet will soon also speedup the long expected convergence of the ICTindustries.13) That is, as the technologies forproducing voice, data and video services con-verge, so might also the respective companiesproducing the associated devices and services.If so, we should already now have witnessed alarge number of telephone companies expandingthrough mergers, acquisitions and genericgrowth into closely related fields. These wouldinclude not only a larger number of nearly iden-tical operations and closely related wireless andcable networks, but also upstream equipmentand content production as well as downstreamdata processing services and content distribution.Similarly vertical and horizontal expansionscould have been expected by other ICT compa-nies, such as equipment suppliers expanding intotelevision distribution and telephone networkoperation, and cable-TV companies into networkoperation, telephone services and data processing.

Over the last decade, there have been manyattempts at integrating these and similar activi-ties, but until recently with only moderate suc-cess. As suggested above, interface standardiza-tion and tradability of complementary assets isone plausible explanation for the lack of com-mercial success. That is, by diversifying intorelated activities that share common resources,significant costs can be saved, superior assetsdeveloped, and extra revenue created, but onlyto the degree essential conditions are fulfilledconcerning the attribute of underlying resources,and the corporate organization employed forusing those resources. First of all, superior assetsshould be safeguarded by contractual, legal, andstrategic means, that either (i) prevent suchassets from leaking out to competitors or (ii)protect the corresponding profits from beingcaptured by more strongly positioned suppliersor customers in the subsequent commercializa-tion phase. Second, activities should be orga-nized into business units and divisions, andinternal governance systems (financial versusbehavior control) designed and chosen for thoseunits and divisions according to tradability ofcommon resources so that financial control isselected over behavior control for non-related,but not for highly related activities where theopposite relation holds. As attributes of criticalresources or services change, so should also theirdivisional structure and governance systems.I expect efficient telecom operators to be thosethat adjust their boundaries and restructure theirdivisions accordingly, and inefficient operatorsto be those that do not.

In a previous report (Ulset, 1998) I questionedthe value of integrating primary telecom servicesnot only with upstream equipment production,but also with upstream content production, asthere are little common resources to be sharedwith these. As far as I could observe then, alsomidstream distribution of media products waslittle more than a pure conduit, where telecomoperators basically provided a transport networkof sufficient capacity for the respective mediaproducts. Downstream IT-services are probablymore closely related to electronic equipmentproduction, than to network operation, andshould therefore normally be more efficientlycarried out by computer companies or special-ized distribution and service companies, thanby network operators.

Since considerable physical and human assetsmay be shared, the economies from integratingvoice services with data communication andvideo distribution should normally be signifi-cant. However, as also these assets are becomingincreasingly tradable, the cost savings from inte-grating may gradually decline. Quite similarbenefits could be obtained from expanding thenumber of local networks and from integratingtransport networks with local access networks.However, also here tradability of critical re-sources is increasing, and the value of integra-tion thus declining due to technology standard-ization and pro-competitive interconnection pol-icy. If this continues, the range of activities fromwhich most telecom operators may gain a com-petitive advantage may gradually shrink towardsthat of a pure conduit. However, until more sys-tematic and recent data is available, the abovestatements should be regarded as largely un-tested hypotheses.

3.3 Preliminary Development inTelecom

3.3.1 Norwegian and British InitiativesSense Communications14) is a Norwegian-basedfirm with the objective of providing new andinnovative fixed and mobile services to theEuropean market based on contractual agree-ments with existing fixed and mobile networkoperators. The company seems to have made agood lesson from its forerunner, here namedOld Sense, which went bankrupt in March 1999,even before becoming operative. The businessidea of Old Sense was to offer small andmedium sized enterprises (SME) tailor-madecommunications services on an all-inclusivebasis. Old Sense classified itself as a virtualnetwork operator (VNO) with the intention ofoffering services using other network operators’

14) Information sources for this company profile is Halbo et. al. (1999a, 1999b) and the company’s home page(http//www.sense.com).


access and switching networks, only equippedwith a minimum of network elements such as (i)Gateway Mobile Switching Centre (GMSC), (ii)Home Location Register (HLR) with Authenti-cation Centre (AUC), and (iii) Subscriber Iden-tity Modules (SIMs).

Old Sense wanted to use Telenor Mobil’s net-work to provide its own customers with directnetwork access and additionally offer one-stop-shopping of telecom services including; VirtualPrivate Networks (VPN), GSM, Centrex, FMC-services and voice-mail. High demand servicescould be offered from their IN-platform, othersvia contractors. Thus, the normal modus for OldSense’ customers in Norway would be roamingin Telenor Mobil’s network, while connection inSweden would be provided through TelenorMobil’s international roaming agreements.Old Sense planned to invest NOK 1.3 billion(£100m) by year 2000, totally dependent, how-ever, on access and interconnection with the net-works of existing operators.

After a series of difficult negotiations withTelenor for interconnection, and with ownersand creditors for additional money and equip-ment, Old Sense went bankrupt in March 1999.Soon after the company was re-established bynew owners as Sense Communications. Theprevious NetCom assistant marketing director,Nadir Nalbant, was elected CEO.

On August 27, 1999, after Elkjøp, a large electri-cal equipment chain (now Dixons) had becomeowner and distributor of services, Sense pio-neered the launching of the first free Internetaccess (no subscription fee) in Norway, namedSense.Wave. The competitors Telia, Telenor andTele1 were completely taken by surprise. Thedemand was enormous, and by November, Sensehad delivered 50,000 Free Internet packages. Inrelative terms, this was equal to the success ofthe British company Freeserve one year ago.Soon after free Internet access were also offeredby Telenor, Tele 2 Norge, Telia Norge, Dag-bladet/Powertech, X-stream/Domino, TV2, Net-Com GSM and CyberCity, and many others.

Later in October Lucent Technologies and SenseCommunications became partners and Lucentagreed to deliver equipment and provide finan-cial backing worth 100 million NOK.

Less than a month later, Sense, Telia Mobile inSweden and Telenor Mobil in Norway agreedon conditions for mobile operations by Sense.Sense could now access these networks and ser-vices necessary to become a serious challengerand competitor in Norway and Sweden. InDecember Sense launched sense.home andsense.homeFree, the former product is a product

that helps Internet users without any priorknowledge to set up their personal homepageon the web. These particular products representa smooth, fast and simple solution for the user.The latter product, Sense.homeFree, is a productfor the advanced user and is free of charge.

At about the same time, Sense concluded a pri-vate placement of 200 million NOK throughPareto Fonds. The interest for Sense is over-whelming and the company value is estimatedat 600 million NOK.

Then on January 17 the following year, SenseCommunications enters the Norwegian mobilemarket as the third but first independent com-petitor. Sense had prior to that signed a contractfor buying traffic capacity and SIM-cards fromTelenor that allowed them to sell mobile ser-vices in their own name. This time, both partieswere seemingly well satisfied with the outcome.Stig M. Herbern, the CEO of Telenor/TeliaMobile for the Nordic countries, said that thisdeal would stimulate competition and contributeto the development of mobile communication.Like old Sense, the new firm requires access toTelenor’s mobile network on their own terms.They want to manage, not only resell, mobiletraffic on Telenor’s mobile network, but this wasturned down by the Department of Communica-tion, and subsequently by Stortinget. In Decem-ber Nadir Nalbant of Sense was elected “IT-manager of the year” by Telecom Revy for hiscontribution to bringing Free Internet to the peo-ple, and for stimulating innovation and competi-tion in the Norwegian market.

Sense managed to establish itself as the firstindependent mobile provider in the Norwegianmarket, with the objective of becoming a pan-European provider of integrated telecommuni-cations services. The Sense business idea wasbased on creating a network of partners, suchas Elkjøp, MCI WorldCom / UUNet, TelenorMobil, Telia Mobile, Siebel Systems and LucentTechnologies, that continuously are being devel-oped and evaluated. With backgrounds fromScandinavia telecom and Internet business, theSense staffs are well known for their enthusiasmand entrepreneurial approach. By choosing the“best from the best” the company aimed at be-coming the leading independent Communica-tions Service Provider in Scandinavia.

Stimulated by the Sense success, a growingnumber of virtual operators were subsequentlyestablished (counting 17 facility-less operators atits peak). Most of these are by mid 2001 stillloosing money. Consolidation through mergersand acquisitions into a smaller number of opera-tors is the likely outcome, if not stopped bybankruptcy and liquidation. Feeling the need


for larger traffic volume and sales revenue, alsoSense decided to grow bigger and thereforebought Site, another newcomer. At about thesame time, the company’s entrepreneurial CEO,Nadir Nalbant, resigned, but only to start assem-bling a new consortium with the intention ofbuying the bankrupt licence holder Broadbandand building the third UMTS network in Nor-way. Sense with its 200.000 subscribers wouldof course be one likely member of such a con-sortium.

NextCall15) is a British virtual operator, estab-lished two-three years ago by Andrew Harring-ton, a previous investment banker. Since 1997BT had then been under pressure from Oftel, theUK telecoms regulator, to offer a new wholesaleservice to rivals called Calls and Access. Har-rington’s idea was to make use of Calls andAccess (he describes it as “phase one local loopunbundling”) to take over from BT not just spe-cific calls such as international ones but a cus-tomer’s entire traffic, and their rental charges.Harrington clearly saw this as an opportunity todevelop a strong, simple marketing propositionto customers. The network owner, British Tele-com, resisted the idea. They felt they had moreto loose than to win by granting facility-lessoperators access to their network.

After a long and hard fight with BT NextCallcould finally start launching its services. Infras-tructure was rented from BT or Energis, andmaintenance provided by BT. The Calls andAccess product gave NextCall the opportunity touse BT’s world-renowned technical expertise todeliver a range of products and services to resi-dential and small business customers. The onlyinfrastructure owned and used by the companywas a billing system and a call center, leavingNexcall free to concentrate on marketing. WhileNextCall enjoyed the technical support of BT,its overheads were smaller, part of the savingsbeing passed on to customers. Although this mayturn out to be a cheaper way to offer service, vir-tual network operators such as NextCall cannotdevise its own services in the way a carrier thatowns infrastructure can.

The company almost immediately ran into prob-lems with BT. NextCall and another newcomercalled Localtel became involved in a high-pro-file tussle over BT’s slowness in letting cus-tomers switch to them. NextCall started com-plaining to Oftel in February 2000 and by the

summer was threatening to seek a judicialreview. Oftel criticized BT’s stance and eventu-ally NextCall’s persistence led to changes inBT's practices and enabled the virtual operatorto run its business more smoothly. By September2000 NextCall had 50,000 residential and small-business customers (increasing at a rate of over8k per month) to which it offers a range of ser-vices including local, long-distance and inter-national telephone services as well as Internetaccess services, all on the same bill.

Its approach was to bundle these services andoffer them at a price that undercuts what a cus-tomer would pay for each service individually.Initially its customers were in clusters in thesoutheast and the south-west/south Wales, areasit targeted using a highly localized marketingcampaign. It used the model established by cableTV operators of door-to-door selling supportedby local newspaper and radio advertising andsponsorship of local institutions (the WelshCricket Association and local boys’ clubs arebeneficiaries). It subsequently launched in theMidlands and Scotland, and moved later on toLondon. At the start of 2001 the Companyemployed over 400 people in a wide variety ofroles and were rapidly becoming a major forcein the UK communications industry.

Calls and Access was not as attractive a solutionfor service providers as unbundling. But a moveinto unbundling could pull NextCall away frombeing a pure VNO and towards a more facility-based operator by installing software in BT’slocal switches to create its own services. Al-though the company was not on principle opp-osed to this, its CEO (Mr. Harrington) thoughtit unlikely. “An infrastructure decision is purelyfinancial. If we can make money from installinginfrastructure I'll do it. But why should I whenwholesale rate are so low?”

Despite all the hard work and all the innovativeefforts, NextCall did not make it. On February13 the UK Activity Report announced thatNextCall has gone into receivership with the lossof up to 300 jobs at its headquarters in Boreham-wood.

Virgin Mobile16), was established as a 50:50joint-venture company between Virgin Groupand Deutsche Telekom’s One 2 One. Launchedin the UK on November 11, 1999, it was also theUK’s first mobile virtual network operator. In its

15) This company profile of NextCall draws heavily on Financial Times Survey of June 21, 2000, written byRichard Hanford (http://www.ft.com/ftsurveys) and the company’s home page (http://www.nextcall.co.uk/).

16) This company profile of Virgin Mobile draws heavily on Financial Times Survey of June 21, 2000, writtenby Niel McCartney (http://www.ft.com/ftsurveys) and the company’s home page(http://www.virginmobile.com/mobile/).


first seven months it signed up nearly 300,000cellular telephone users, making it the fastest-ever, and probably also the earliest-ever prof-itable start-ups, mainly due to the fact it does notbuild its own network. Instead it buys bulk air-time from One-2-One. Under this arrangement,it is able to issue its own subscriber identifica-tion module (SIM) cards, the smart cards inhandsets, which identify the customer and con-trol their relationship with the network. It is thiselement, which differentiates the company fromtraditional cellular service providers. The com-pany is thus responsible for its own tariffs, bill-ing and customer service infrastructure, and forthe full cost of acquiring customers. It carriesmore commercial risk than a service provider,but it also gains more benefit if it is successful.

Virgin Mobile sells its services through Virgin’sMegastores and Our Price shops, but it also usesdirect marketing and the Internet. A Sim-basedweb browser allows customers to access websites, which use the wireless application protocol(Wap) technology, such as BBC Online andMegastore Direct, Virgin’s online music retailer.This means that its customers will not need tobuy special Wap handsets to access these ser-vices. But it will sell Wap telephones to thosewho want them. For example, VirginXtras isVirgin Mobile’s unique range of mobile leisure,information and entertainment services. WithVirginXtras customers can save in the range of10 % off selected Virgin Holidays; 15 % offmusic, videos, DVDs and computer games; andup to 50 % off electrical goods such as TVs,stereos, fridges and microwaves compared totypical high street prices.

The company started with the considerable mar-keting advantage of a very strong image, cour-tesy of the Virgin Group. This claims to have thebest-known domestic brand in the UK, with ahigh reputation for value and customer service.So, unlike its competitors, Virgin Mobile didnot have to spend much on brand building. Thisallowed it to set a relatively modest advertisingbudget of £5m to cover its launch and the firsttwo months of operation – far less than most ofits rivals were spending at the same time. It ex-pected to spend a total of £20m on advertising inits first year of operation.

Unlike the UK’s four normal network operators,Virgin Mobile did not subsidize the cost ofhandsets. It sold them at cost price (plus VAT),charging £60 to £350. But it offered a cheap andsimple tariff package with no subscription fees

and low call charges.17) The company said itsrates were on average at least one-third lowerthan anyone else’s. The same prices wereoffered to all customers, regardless of how theypay and whether they paid in advance or werebilled at the end of the month.

As well as signing new cellular users, the com-pany is now putting more emphasis on poachingcustomers from its rivals. If these users have acompatible telephone, and are out of contract,they can switch to Virgin Mobile simply by buy-ing one of the company’s Sim cards for £12.50,a price that includes calls worth £10. The com-pany estimates that about 10m UK users couldswitch to its service without having to buy a newphone. Virgin Mobile has also sought to differ-entiate itself by offering advanced services suchas the facility to surf music and to send andreceive three-page e-mails. It is able to deliverthese services to standard handsets by exploitingthe high capacity of its SIM cards. These have32 kB (kilobytes) of memory, at least twice asmuch as those currently used by other UK opera-tors. The company admits its approach generatesless money than its competitors do. It says that itis getting close to £20 a month per user, signifi-cantly less than the average for the UK industry.But it argues that this is more than offset bylower costs. It also claims that the expense itincurs in acquiring each customer is 50 per centlower than that of its nearest rival.

This combination of simple but cheap tariffs andadvanced services plays a central role in VirginMobile's marketing message, which uses alladvertising media. Unlike most of its rivals, itdid not initially spell out its prices in its adver-tisements. Instead, potential customers wereencouraged to contact the company to find outmore. But the pricing message is likely to getsharper now that the service has establisheditself.

The Virgin Group, which holds the overseasrights to the Virgin Mobile brand name, is nowseeking to replicate this success in other coun-tries. It has set up a joint venture with Australianoperator Cable & Wireless Optus. It has alsosigned an agreement with Singapore Telecom-munications to set up a venture to offer cellularservices throughout Asia, starting in 2001. Moredeals are in the pipeline.

Already in November 2000, as the company cel-ebrated its 500,000th customer and first birth-day, Virgin Mobile was voted “Best Network

17) This package involves charges of 15p a minute for the first five minutes each day and 5p a minute after that(the initial rate was 15p a minute for the first 10 minutes, 10p a minute for the next 10 minutes, and then 5pa minute).


2000” by consumer magazine Mobile Choicefor offering customers the best deal and serviceavailable in 2000.18) Early the following yearthe company announced it had exceeded its salestargets for the fourth quarter running, by attract-ing over a quarter of a million customers in thelast three months of 2000, increasing its cus-tomer base by a market-beating 63 %. It claimedto be on course to meet its target of reaching 1million customers by the end of March. Analystssaid it may turn profitable already in 2001 andmight even be floated at a value well above£1bn.

As the Virgin case suggests, also virtual opera-tors may turn profitable if favorable conditionsprevail. These may at least include cheap tariffs,advanced services and popular content, sup-ported by a strong brand and a productive con-tract with at least one large facility-based opera-tor (One-2-One). As emphasized above, crucialconditions for creating such a productive out-sourcing contract include standard interfacesbetween technologically separable network facil-ities and service capabilities that connect mid-stream network operator with downstream ser-vices provider. After becoming sufficiently stan-dardized, these interface facilities and capabili-ties are no longer tacit and sticky and thereforeno longer candidates for full corporate integra-tion. Since the basic interface codes (sometimeseven the source code) are available to all, marketcontracting will suffice as governance form forsubsequent development, transfer and utilizationof the technology and the products sharing thiscode. No additional organizational capabilitywould be needed. This statement applies espe-cially well to network operators. By keepingtheir network open and accessible (though openinterfaces) rather than closed and inaccessible toexternal suppliers and customer, network opera-tors may maximize capacity utilization and mini-mize unit costs.

3.3.2 Responses by Regulators andIncumbents

As indicated above, the initial responses byincumbents and their owners to the variousVNO-initiatives have been strongly negative.As long as VNOs are only expected to captureexistent traffic, not to create extra revenue,

incumbents’ reactions can only be negative.Regulators, however, who would like to seemore infrastructure competition, tend to be morepositive. Gradually, also the negative responseshave weakened, especially in Telenor who havecome to realize that open access in terms ofunbundling and roaming would not only berequired by the regulator as a general conditionfor allowing Telenor to dominate the market, butincreasingly also by Telenor itself as part of itsinternational strategy. To become accepted asa VNO abroad, Telenor must also be willingto accept foreign as well as domestic VNOs athome. Besides, virtual operators may also playa more productive role. Under the appropriateconditions, virtual operators may develop intoa more efficient and innovative marketing forcethan corresponding units of facility-based opera-tors, generating extra profit not only for them-selves, but also for their facility-based partners.

4 Summary and TentativeConclusion

In this article I have analyzed how incumbentand new entrants may benefit from opening thevalue-chain; that is, by creating separate facility-less network operators that profit from serviceinnovation while outsourcing basic networkoperation to facility-based operators. Being arelatively recent phenomenon in telecom, theinstances of virtual business practice describedand analyzed above include not only examplesfrom telecoms, but also from the computerindustry and the Internet.

Quite obviously, to profit from outsourcing, spe-cializing and downscaling in general, and fromvirtual network operation in particular, the pricemargin of the resultant “virtual” operation mustcover its costs. Since this revenue margin islikely to expand with the VNOs own freedom inservice innovation, packaging and pricing, grant-ing such freedom can be regarded as one neces-sary, but still not sufficient condition for profit-ing from this type of business practice. In addi-tion, important conditions related to servicetransactions, to production resources, and tocompetition and governance should also be ful-filled. In particular, and according to transactioncosts economics, corporate integration would bethe organizational solution of choice for techno-

18) Orange falls to third place behind Vodafone, with One 2 One coming fourth, and BT Cellnet trumped in lastplace. This motivated Mobile Choice's Steve Gold to the following comment on the award: "It may be a sur-prise to some and a downright shock to others, but squeezing in at number one isÖVirgin Mobile. It's beenan interesting year for mobiles and undoubtedly a cracking one for consumers, thanks to the arrival of theground breaking new 15 to 10 to 5 pence tariff from Virgin - which soon slashed to a 15 to 5 pence deal thissummer. Virgin deserves its number one slot after an impressive first year." - Richard Branson, chairman ofVirgin Mobile responded: “We're thrilled to have received this award. It recognizes the enormous contribu-tion all the staff at Virgin Mobile have made in making Britain's newest mobile phone company the best. It's agreat compliment and a tremendous accolade to receive on our first birthday," (reported by Niel McCartneyin http://www.ft.com/ftsurveys).


logically separate operations (activities) only ifthe intermediate products or services producedare non-tradable or if the respective underlyingproduction resources are non-redeployable.

As indicated above, the resultant products andservices are non-tradable to the degree theirquality is difficult to measure and evaluate, orto the degree basic production resources usedto produce and deliver intermediate products orservices, are specific, tacit or diffused and there-fore also difficult to transfer or access, unlesssupported by additional organizational transferor access capabilities in terms of shared values,norms, codes and routines. On the other hand,production resources that are technically separa-ble may still be non-redeployable to the degreethey are partnership-specific (tailor-made) andthereby useless for other relations or applica-tions. Should similar partners appear, however,specialized resources would no longer be part-nership-specific, although they still may be non-tradable due to being tacit or diffused.

To the degree at least some of the respective pro-duction resources and support capabilities arealso valuable and unique, these may under theprotection of corporate integration develop intomore sustainable sources of competitive advan-tage. However, increasing outsourcing in gen-eral, and virtual network operation in particular,indicate the opposite trend: services are madeincreasingly separable due to the use of increas-ingly redeployable and/or tradable productionresources. Whereas some of the separated activi-ties and operating units are well positioned toprotect their already unique assets or to developnew ones (e.g. differentiable content or value-added services, developed and produced byfacility-less operators), others are less well posi-tioned (e.g. non-differentiable transport services,produced by facility-based network operators).Due to significant increases in expected trafficvolume generated by a growing volume of inno-vative services, supplied by virtual operators,also physical network operators may earn signif-icant rents by differentiating their network oper-ation and service provision. Increasing competi-tion, facilitated by the recent pro-competitiveregulatory reforms, is perhaps the strongest forcebehind this embryonic restructuring process,besides the concomitant advances in digitizedtechnology.

Obviously, both the computer industry and theInternet world illustrate that whereas innovationand growth are enhanced by open standards,

huge profits will only result from large-scalesales of products or services on the basis ofsome kind of upstream or downstream monopolyposition. Lacking such monopoly positions,most computer makers (assembler) and InternetService Providers (ISP) are typically character-ized by small profit margins. Even among themost successful and previously high-pricedonline and e-commerce companies (e.g. Ameri-can Online and Amazon.com) operating profitshave always been tiny, even negative, due tolow-priced introductory offer and large-scalemarketing investment, combined with unre-stricted competition in operating mostly non-proprietary production facilities.19) If protectedby branding and copyrights, digitized contentmay develop into a competitive advantage. This,however, will mostly benefit content producers,not network operators unless distributed underexclusive contracts or within fully integratedcompanies. Nor can advances in network tech-nology, operating support systems and serviceapplications constitute any sustainable competi-tive advantage so long as these are built on openand non-proprietary technology platforms.Under such regimes, the speed by which thesedevices, systems and applications can be turnedinto a seamless system that works may be theonly remaining source of competitive advantage.However, also here network operators are butone of several competing system managementoperators, the others being software makers andservice providers.

Whether integration will have the intendedeffects or not, will most likely depend both onthe innovation incentive of the network operatorand certain key attributes of the resources usedto produce the outsourced products and services.Increasing use of non-proprietary and open inter-faces between network devices, operation sup-port systems and service applications across dif-ferent technology layers (facilitated by theTCP/IP protocol) have not only contributed tomaking all of these more separable, but alsomore tradable. Consequently, this has alsoincreased the value of contracting (renting outshared devices, systems and applications or sell-ing only the associated services) compared tofull corporate integration. Furthermore, underoligopolistic network competition, service inno-vation will probably be more strongly promotedby facility-less operators, as above exemplified,than by facility-based operators. Then, the prof-itability of both virtual and non-virtual operatorswill depend on attributes of their core assets,especially the degree to which these assets are

19) On the other hand, the stock price of those companies may occasionally rise steeply and even fall steeper as itrecently did due to grossly inflated demand expectation combined with enormous scale economies (high fixed,but zero marginal costs) associated with providing digitized content over future broadband networks to mil-lions of consumers.


uniquely valuable, inimitable or otherwise pro-tected by patents or private contracts (co-owner-ship, quarantines, protective contract clauses).If inimitable, essential resources may still bemade tradable under protective renting or leasingagreement, or by spinning essential resourcesoff into a stand-alone unit capable of selling theassociated services to a larger number of com-peting customers without too much leakage ofprivate technology and knowledge.

Consequently, the conclusion to be drawn fromthe above examples and discussion should be asfollows: In the telecom services industry com-petitively critical assets are becoming increas-ingly separable, tradable and less monopolized,and thereby ready for being outsourced andmore widely dispersed among a larger numberof specialized players, along with private knowl-edge needed to further develop these assets. Sub-sequently, this may cause virtual (facility-less)operators to develop stronger innovation incen-tives and competencies in service provision, andnon-virtual (facility-based) operators to developstronger capabilities in digitized network opera-tions on the basis of increasingly more com-moditized and less monopolized productionassets.

BibliographyBarney, J B. 1997. Gaining and Sustaining Com-petitive Advantage. Reading, MA, Addison-Wesley.

Business Week. Special Report, Nov. 22, 1999.

Chesbrough, H W, Teece, D J. 1996. When isVirtual Virtuous. Harvard Business Review,January-February, 65–73

Foros, Ø, Hansen, B. 2000. Connecting cus-tomers and disconnecting competitors. The facil-ity-based firms’ strategy towards virtual opera-tors. Kjeller, Telenor R&D. (R&D N 33/2000.)

Gong, J, Srinagesh, P. 1996. Network competi-tion and industry structure. Industrial and Cor-porate Change, 5, 1231–1241

Gong, J, Srinagesh, P. 1997. The Economics ofLayered Networks. In: McKnight, L, Bailey, J(eds.). Internet Economics. Cambridge, MA,MIT Press, 63–77.

Halbo, G et al. 1999a. Exploiting the edge inintelligence to build a SoftTelco business.Kjeller, Telenor R&D. (R&D R 108/99.)

Halbo, G et al. 1999b. Business plan for ‘Fazet’,a software based telco. Kjeller, Telenor R&D.(R&D R 109/99.)

Kavassalis, P, Lee, T Y, Bailey, J P. 1998. Sus-taining a Vertically Disintegrated NetworkThrough a Bearer Service Market. In: Bohlin, E,Levin, S L. Telecommunications Transforma-tion: Technology, Strategy and Policy. IOSPress.

Teece, D J. 1986. Profiting from TechnologicalInnovation: Implications for Integration, Collab-oration, Licensing and Public Policy. ResearchPolicy, 15, December, 285–305.

Ulset, S. 1994. Commercializing Technology inNetwork Industries. (SNF Working Paper,104/94.)

Ulset, S. 1998a. Hvordan organisering av infra-struktur kan bidra til utvikling av bedre og bil-ligere teletjenester. (SNF-rapport 25/98.)

Ulset, S. 1998b. Optimal Organization of theConverging Information and CommunicationTechnology (ICT) Industries. Theoretic analysisand some illustrative data from the Norwegianmarket. (SNF-report 27/98.)

Spiller, P T, Ulset, S, Zelner, B A. 1996. Strat-egy, Structure and Global Networks. Studies ofthe telecommunications sector. (SNF-report27/96.)

Williamson, O E. 1991. Comparative EconomicOrganization: The Analysis of Discrete Struc-tural Alternatives. Administrative Science Quar-terly, 36, June.

Telektronikk 4.2001

PrefaceKenth Engø and Steinar Låg wrote this paper asparticipants in Telenor Corporate University’sMaster of Telecommunication Strategy Program2000–2001.

The Master of Telecommunication Strategy Pro-gram is a cross-disciplinary education programinitiated by Telenor and NTNU. The program isfocused on a project case where the students bycreating a “virtual company” shall write a busi-ness plan. In the 2000-2001 program, a businessplan for an application provider and system inte-grator called Move-On was designed. The workpresented in this paper was carried out as part ofthe analysis work for Move-On’s business plan.

1 IntroductionThe modern telecommunication company isevolving into a system which largest challengeis coping with the ever-increasing complexity ofthe telecommunication business. The reason forthe increase in complexity is due to several fac-tors; more advanced and complicated technol-ogy, more complex organisations, the multi-faceted game of co-operation and competitionamong actors in the market, intricate productsand difficult market situations, and the compli-cated task of value creation [Audestad, 1998].

Hand in hand with complexity walks uncer-tainty. The complex system of the telecommuni-cation business turns upside down on decision-making, because the methods used for forecast-ing and making projections into the future aremade for a linear system. In a linear situation theresponse is proportional to the adjustment, andin this way no surprises will occur. The new sit-uation that the telecommunication business isfacing breeds the need for decision makingunder truly random (stochastic) market condi-tions.

The most important thing to realise in a complexworld with thriving uncertainty is that random-ness and stochasticity imply possibilities. Hence,uncertainty enfolds economic value – a point

which it is important for managers to realize.Like financial options are used in the financialmarkets to tame fluctuating asset prices, realoptions can be used to value strategic optionsin a market operation. Real options is not thatmuch about numbers as it is about strategicthinking, managing flexibility, and decisionmaking [Amram & Kulatilaka, 1999].

Virtually any business project consists of a col-lection of real options. The problem is to iden-tify the options, value them, and manage themcorrectly. For example, the following are realoptions that can be of some economic value[Trigeorgis, 1996]:

• The option to defer the start of a project;

• The option to abandon a project;

• The staged investment option;

• The option to scale up/down a project;

• The option to switch between technologiesdepending on prevailing market conditions;

• Growth options;

• And multiple interacting real options.

In this paper we will discuss and explore someviable real options identified for the Move-Onbusiness case. For the reader with no priorknowledge of options we include in Section 2 anintroduction to financial and real options. Thiswill by no means be an exhaustive introductionwhere only the key points of options valuationare introduced. Section 3 will go through the realoptions identified in the Move-On business case,and a valuation will be attempted. We concludethe paper in Section 4 with strategic recommen-dations.

2 Options in GeneralIn this section we will introduce option valuationbased on the Black-Scholes theory. For a com-

MORO – Move-On Real OptionsK E N T H E N G Ø A N D S T E I N A R L Å G

As part of the Move-On business case we identify and investigate several options that are implicitly

contained in the business case. These options are so-called real options, and similar to financial

options, they have their own intrinsic value. We will in this paper give a brief introduction to option theory

and then use the theory to value the real options identified in the Move-On business case. The option

theory is introduced through the financial option theory, whereas for the real options we give specific

examples that relate to our special case.

Steinar Låg (33) was, at the timeof writing this article, student inNTNU and Telenor CorporateUniversity’s Master of TelecomStrategy Program 2000–2001.Since receiving his Master ofScience in Electronic Engineer-ing from NTNU in 1993, he hasbeen with Nera, Telenor andInmarsat (London), working withsystem and network aspects ofmobile satellite communicationssystems, with focus on systemdesign and development ofInmarsat services. Låg is cur-rently employed by TelenorSatellite Services, working onsystem and product develop-ment as well as future strate-gies.

[email protected]

Kenth Engø (31) is a scientist atTelenor R&D. He holds a Masterof Science (1995) in industrialmathematics from NTNU, Trond-heim, Norway, a Dr.Scient.degree (2000) in informaticsfrom University of Bergen, Nor-way, and a Master of Technol-ogy Management (2001) intelecommunication strategyfrom NTNU, Trondheim, Norway.Since joining Telenor R&D in2000 his interests have mainlybeen in risk and real options,modelling and investigatingmathematical properties of so-called ad-hoc networks, andexamining the effects on thebehavior of the ECN markingscheme in a network with usershaving inelastic demands.

[email protected]

82


prehensive and detailed introduction, see e.g.[Wilmott et.al, 1995; Hull, 2000; Dixit &Pindyck, 1994].

2.1 Financial OptionsThe owner of a financial option has the right –but not the obligation – to buy or sell a financialasset at a fixed price – the strike price – on orbefore a predetermined date. The option is calledEuropean if the holder can only exercise theoption on the date set forth in the option con-tract, and likewise, the option is called Americanif the holder can exercise the option at any timeup till, and including the last running day of theoption, agreed upon at the time of issuing theoption contract.

The value of an option can be described by theBlack-Scholes (BS) option pricing model. In thefollowing we will restrict our attention to thismodel only.

In the BS model, the value of an option, denotedby V(S,t), is a function of the current value of theunderlying asset S, and time t. The value of theoption is governed by the partial differentialequation (PDE)

This is the so-called Black-Scholes equation. Inthe above equation σ denotes the volatility of theunderlying asset and r is the risk-free interestrate (typically, it is the bank interest rate). Thetype of option will decide the boundary and finalconditions for the PDE, and the strike price –denoted by X – enters the equation through theseconditions.

A European call option is the right to buy theunderlying asset S on some future date T for theamount X. The final condition is that the value ofthe option at the exercise time t = T equals thepayoff

V(S,T) = max(S – X,0).

This says that if the underlying asset price is lessthan the strike price at t = T the payoff is zero,otherwise the payoff is the difference betweenthe asset price and the strike price. The boundaryconditions are applied for S = 0 and S → ∞. Inthe case S = 0 the call option is worthless, evenif there is a long time to expiry. Hence we havethat

V(0,t) = 0.

As the asset price increases without bound itbecomes ever more likely that the option will be

exercised and the magnitude of the exerciseprice becomes less and less important. Thus asS → the value of the option becomes that ofthe asset and we get

Having specified the boundary and final con-ditions it is known that the BS equation has aunique solution. Hence, this can be fed into acomputer and a solution can be obtained.

In the case of European options the solution isknown as a closed-form formulae both for thecall and put options. The value of the Europeancall option is

C(S,t) = SN(d1) + Xe–r(T-t)N(d2),

where N is the cumulative normal distributionfunction

and d1 and d2 are given as the parameters

The European put option the value is

P(S,t) = Xe-r (T-t)N(–d2) – SN(–d1).

In Figures 1 and 2 we have plotted the payoffcurve and the value of a call and a put option forthe Telenor stock (TEL) valued August 24,2001, with exercise date January 17, 2002,where the underlying stock value is NOK 36.50,strike price equal to NOK 35, estimated volatil-ity of 35 %, and risk-free interest rate of 7.3 %.

Checking the above results up against the marketon the Oslo Stock Exchange, we find that theabove call option – TEL2A35 – theoreticallyvalued to NOK 3.57 is traded at NOK 4 onAugust 24, 2001, and the put option – TEL2M35– theoretically valued at NOK 2.56 were offeredat NOK 2.85 by the seller and bought for NOK 2by the buyer on the same date. The differencesobserved between theory and observed trade ismainly due to the fact that it is American optionsthat are traded on the Oslo Stock Exchange, andthe market’s different judgements of the currentvolatility of TEL.

d2 =log S / X( ) + r − σ 2 / 2( ) T − t( )

σ T − t.

d1

log S / X( ) + r + σ 2 / 2( ) T − t( )

σ T − t

N(d ) =1

2πe−

12 x2

−∞

d∫ dx,

limS→∞

V S,t( ) = S.

∂V

∂t+

1

2σ

2S

2∂2V

∂S2+ rS

∂V

∂S− rV = 0.

∞


26 28 30 32 34 36 38 40 42 44 46Value of underlying asset - NOK

European put option value for TEL12

10

8

6

4

2

0

Val

ue o

f op

tion

- N

OK Option value

Payoff

26 28 30 32 34 36 38 40 42 44 46Value of underlying asset - NOK

European put option value for TEL8

6

4

2

0

Val

ue o

f op

tion

- N

OK Option value

Payoff

There are quite a few assumptions underlyingthe Black-Scholes option model. The modelassumptions are probably the most importantthing to know for users of option pricing models.The number crunching is easily fed into a com-puter program – and the user must be able toanalyse the output in order to validate theobtained answers.

The following is the list of assumptions made inthe BS option-pricing model:

• The asset price follows a lognormal randomwalk.

• The risk-free interest rate r and the assetvolatility σ are known functions of time overthe life of the option.

• There are no transaction costs associated withhedging a portfolio.

• The underlying asset pays no dividends duringthe life of the option.

• There are no arbitrage possibilities.

• Trading of the underlying asset can take placecontinuously.

• Short selling is permitted and the assets aredivisible.

How well are these assumptions fulfilled by themarket? This is very important to analyse, be-cause if the model assumptions are not observedin the market the BS model is not the correctmodel for the market!

The two first assumptions cause the most prob-lem for the BS option-pricing model to correctlycapture the behaviour of true financial markets.First, the assumption that the underlying assetfollows a lognormal random walk is quite oftena good approximation of the market performanceof an asset, but it is easy to find assets that donot conform to this assumption. Second, thereis no way of knowing the volatility and risk-freeinterest rate over the life of the option. The risk-free interest rate is not fluctuating much, and theimpact of incremental changes in the interest ratedoes not affect the option value to a large extent.However, the volatility has a large impact on thevaluation of options. Volatility also depends onthe data samples. To illustrate how much volatil-ity can change over time, and how much it candepend on the data, we have plotted the volatil-ity of TEL calculated (data including August 24,2001) with a 50-day data window, see Figure 3.

For the sake of comparison, we also plot corre-sponding graph for the OPC stock – a highlyvolatile stock on the Oslo Stock Exchange (Fig-ure 4). Note that for TEL the volatility range is

Figure 1

Figure 2


(20 %, 45 %) annual volatility, whereas for OPCit is (40 %, 200 %).

Another point to notice about volatility is theimplied volatility calculated from observingoption prices in the market. If the BS model is

correct and explains option prices observed inthe market the implied volatility of the sameunderlying asset should be the same for differentstrike prices at the same maturity date. Oneshould observe a horizontal line when volatilityis plotted versus strike price. However, in the

Figure 3

Figure 4


market one can observe a curve – the volatilitysmile – with increasing volatility for the largerand smaller strike prices.

Extensions of the BS model including stochasticinterest rates and stochastic volatilities havebeen proposed, but we will not follow this anyfurther here. We refer the interested reader tothe referenced literature.

To summarize; the five central variables in theBS option pricing model are:

1 S – current value of the underlying asset;

2 X – strike or exercise price;

3 T – exercise date of the option, also calledmaturity;

4 r – risk-free interest rate;

5 σ – volatility of the underlying asset.

2.2 Real OptionsA real option is an analogue of a financialoption, but the most important difference is thatthe “underlying” is not a financial asset but forinstance an investment project. The importantexercise in the real option case is to “map” theproject and its related variables onto a genericoption. In this sense a real option is somethingmore general than a financial option, but sourcesof uncertainty are also a lot more difficult toidentify. In terms of financial valuation realoptions can be viewed as an supplement to thetraditional discounted cash flows, where the realoptions approach also take advantage of the flex-ibility inherent in the project when valuing it.

In the following we will map a project invest-ment onto a European call option, and use theBlack-Scholes option-pricing model to value thisoption. One important question that the managershould ask himself is if the assumptions of the

BS option pricing model are fulfilled? Does theunderlying cash flow from the investment pro-ject follow a lognormal distribution? What canyou say about the volatility of the investmentproject? The results from the calculations shouldundergo heavy scrutiny – with reference to theassumptions of the option-pricing model used(Figure 5).

In fact, the mapping of an investment projectonto a call option is direct, once the DCF analy-sis is available. As seen in the above figure, thepresent value of a project’s free cash flow corre-sponds to the current price of the underlyingasset. Hence, referring to the assumptions of theBS valuation model, the variation of the presentvalue of the project should follow a lognormalrandom walk. This assumption can hardly everbe tested. Next, an option’s exercise price willcorrespond to the expenditures required toacquire the proposed project assets. The time ofexpiration for the option is the amount of timethe decision to take up the project or not can bedeferred. Time value of money corresponds tothe risk-free interest rate. The hardest thing toestimate for a real option is the risk of the pro-ject assets, which will correspond to the under-lying assets volatility. In the stock market his-torical data can indicate a reasonable choice ofvolatility, but for an investment project you willin general not have historical information avail-able.

Strategies for choosing the project’s risk can beto try out a range of possible values, or check thefinancial markets. One can estimate the histori-cal volatility of companies that are traded in thefinancial market and have similar characteristicsas to the proposed investment project. This willgive an indication of how the financial marketperceive and value the risk of companies operat-ing in the same sector as the proposed project.

Figure 6 shows some examples of common realoptions.

Present value of project´sFree Cash Flow

Expenditure required toacqurie project accets

Length of time the decisionmay be deferred

Time value of money

Riskiness of project assets

Stock price

Exercise price

Time to expiration

Risk-free rate

Variance of returns

S

X

t

Rf

σ2

Investment Opportunity Call OptionVariable

Figure 5 Real Options: Thelink between investments andBlack-Scholes inputs


Some examples of the above real options can bethe following. In the invest/grow category of realoptions, typical scale up options can be invest-ments in high technology, and research anddevelopment intensive projects. The scope upoption is relevant for the telecommunicationbusiness, since this option is relevant for busi-nesses experiencing lock-in, and for de factostandard bearers. The defer option is relevant inthe Move-On project, since it is desirable to waitand see how well the market accepts our prod-ucts before further investment. The scale downoption is also relevant here. If the product launchis not successful Move-On will want to scaledown the investments or maybe even terminatethe project. The last two real options illustratethat it can be difficult to identify the type ofoption that you are really interested in.

We will now go through an example of a realoptions calculation. Consider the cash flows inTables 1–4.

Carefully looking at the present values one cansee that something is happening in year 3. Alarge investment is planned for in year 3, andjudging from the net present value (NPV), thisis not that lucrative. Splitting up the above cashflow into cash flows relevant to the present oper-ation and to the investment project in year 3only, the situation is the following. The cashflows from the present operation are as shown inTable 2.

The cash flows relevant to the investment in thethird year are as shown in Table 3.

To get the correct valuation, we discount theinvestment in year 3 with a risk-free interest rateof 5.5 % instead of the discount factor of 12 %.The reason for doing this is that there is nouncertainty about the investment (strike price);hence no risk is involved in this investment.This results in the updated net present value forthe year 3 investments as seen in Table 4.

In order to value the option whether to invest ornot, we map the above investment onto a calloption. The choice is to buy the underlying pro-ject, and we can defer the decision about this tillyear 3. Returning to the five central variables inthe BS option-pricing model, we have collectedthe following information:

1 Current value of the underlying asset.S = 255.7

2 Strike or exercise price. X = 382

3 Exercise date of the option, also called matu-rity. T = 3

4 Risk-free interest rate. r = 5.5 %

5 Volatility of the underlying asset. s = 40 %

We find the four first variables in the DCFspreadsheets, and as an estimate of volatility wechoose 40 % per annum. This estimate can bebased on historic data from similar traded assets,implied volatilities on options, or just simply arange of values.

Well positioned businesses can scale up laterthrough cost-effective sequential investmentsas market grows.

Scaleup

RealOption

Category

RealOption Type

Switchup

Scopeup

Study/start

Scaledown

Switchdown

Scopedown

Invest/grow

Defer/learn

Disinvest/shrink

Description

A flexibility option to switch products, processon plants given a shift in underlaying price ordemand of inputs or outputs.

Investments in proprictary assers in one industryenables company to enter another industry cost-effectively. Link and leverage.

Delay investment until more information or skillis acquired.

Shrink or shut down a project part way throughif new information changes the expected payoffs.

Switch to more cost-effective and flexible assetsas new information is obtained.

Limit the scope of (or abadon) operations in arelated industry when there is no further potentialin a business opportunity.

Figure 6 Common Real Options(Source: “How much is Flexibility

Worth?” Thomas E Copeland and PhilipT Keenan, “The McKinsey Quarterly”,

1998, No 2; “Real Options”, NalinKulatilaka, Harvard Business School

Press, 1999

Year 0 1 2 3 4 5 6

Cash flows 0 9 10 11 34.7 37.5 40.7

+ terminal value 0 0 0 0 0 0 610.3

– investment 125 0 0 382 0 0 0

x discount factor 12 % 1.00 0.89 0.80 0.71 0.64 0.57 0.51

= PV (by year) -125.0 8.0 8.0 -264.1 22.1 21.3 329.8

NPV (sum of years) 0.1

Year 0 1 2 3 4 5 6

Cash flows 0 9 10 11 11.6 12.1 12.7

+ terminal value 191

– investment 125


= PV (by year) -125.0 8.0 8.0 7.8 7.4 6.9 103.2

NPV (sum of years) 16.3

Table 1

Table 2


Based on the above values an option calculatorwill value the call option to be worth 48.7, orabout 19 % of the underlying asset value. Hence,the total value of the investment project is16.3 + 48.7 = 65, which is a long way from theinitial NPV of 0.1. The project is in fact a veryattractive one.

In the next section we turn to the real options inthe Move-On business case.

3 Applications to Move-On

3.1 IntroductionThe Move-On project is based on the businessidea to develop and deliver community function-ality to mobile terminals. See [Move-On Busi-ness Plan, 2001] for further details. The NPVanalysis in the business plan indicates a positivebusiness opportunity.

As we see in Table 5, there are negative cashflows for the first four years, before experienc-ing positive cash flows from year 5 and on-wards. An important question is whether theMove-On Project contains any real options,which would enhance the business case, or atleast reveal possible opportunities that couldarise from Move-On’s initial business. The pur-pose of this section is to identify and explorethree such intrinsic real options, and considertheir value and possible implications to theMove-On Project.

In the following we have chosen to explore thesubsequent three options; the building IP zonesoption, the integrating IP zones option and theshutdown option. In Section 3.3 we consider theoption to build a network of IP-zones (usingWireless LAN equipment). For Move-On, thiswould mean expanding the product offering ina market where the initial business has alreadybeen established. Conventional NPV analysisdoes help in evaluating the attractiveness ofmaking such an investment, but offers no helpwhen it comes to considering whether there isgreater value in postponing this investment deci-sion. In Section 3.4 we consider another optionrelated to deployment of IP-zones, i.e. the optionto integrate IP-zones, which have already beenbuilt by others. This is a “lighter” version, in thatcapital investments and network rollout timesare considerably shorter. Again, the questionsare how much does it cost to buy and preparefor this option, how much uncertainty will beresolved while waiting and what is the value of

Table 5 Free Cash flows forthe Move-On Project(in kNOK)

Year 0 1 2 3 4 5 6

Cash flows 0 23.1 25.4 28

+ terminal value 419.3

– investment 382


= PV (by year) 0.0 0.0 0.0 -271.9 14.7 14.4 226.6

NPV (sum of years) -16.2 255.7

Year 0 1 2 3 4 5 6

Cash flows 0 23.1 25.4 28

+ terminal value 419.3

– investment 382


Risk-free rate 5.5 % 1.00 0.95 0.90 0.85 0.81 0.77 0.73

= PV (by year) 0.0 0.0 0.0 -325.3 14.7 14.4 226.6

NPV (sum of years) -69.6 255.7

Table 3

Table 4

Year 0 1 2 3 4 5

CFO -5909 -16937 -16198 -4642 31226 97947

Terminal value 216000

Invest 7066 1590 9826 21190 37337 45595

x discount (rf) 1 0.97 0.94 0.92 0.89 0.86

x discount (rd) 1 0.87 0.76 0.66 0.57 0.50

DFCF -12975 -16110 -19678 -16985 -3494 128994

ADFCF -12975 -29085 -48762 -65747 -69241 59753

NPV 59753


postponing the decision. Whereas the two firstoptions are expansion (or growth) options, Sec-tion 3.5 introduces the shutdown option, whichis almost the opposite. The shutdown option rep-resents the opportunity to stop the project, iftechnology, market developments or any internalor external condition turns out to significantlyreduce the value of the project. Unlike the ex-pansion options, the value of the shutdownoptions lies in reducing downside. However, theshutdown option does also require some prepara-tory actions and costs, which is further discussedin Section 3.5 below. Finally, in Section 3.6 wemention other real options relevant to the Move-On project.

Before exploring the identified real options, andattempting to evaluate them, we first need toderive and estimate various parameters, whichis done in the next section.

3.2 Parameter Estimation

3.2.1 The Risk Free Interest Rate (rF)From [Bøhren & Michaelsen 2001] we see thatbased on the Norwegian government bonds inthe period 1992–2000, the average value isapproximately 2 % after deflating with CPIgrowth (i.e. real rF). The real rF for the last twoyears (1998–2000) has been approximately 3 %.Judging the macro-economic conditions in Nor-way as of August 2001, there are no signs of anynear-term immediate reductions from this level,so in the following we will use rF = 3 %.

3.2.2 Discount Rate (r)A well-established method to evaluate projectsor companies is to find the net present value(NPV).

It is based on summing up discounted versionsof estimates for future cash flows. NPV is themost common tool to guide investment deci-sions, and the decision logic is very simple: “Ifthe NPV is greater than zero; invest, otherwisedo not”.

The discount rate (r) should be set according tothe owners’ growth requirements, which in turndepends on the risk the company or project isexposed to. There are various ways to deducediscount rates from financial data, see for exam-ple [Bøhren & Michaelsen, 2001]. However forthe purposes of this topic, we shall just assumer = 3 % for the Move-On project itself, as well

as the associated options. This number, which isnot an unusual one for relatively risky Telecom/IT projects, will be used in the remainder of thispaper.

3.2.3 Volatility (σ)Volatility is perhaps the most important parame-ter to consider when dealing with options. How-ever, it is not straightforward to estimate, and weneed to estimate equity volatilities of all threephases (σO1, σO2 and σO3). As shown in Section2.1, we can estimate the volatility by using avail-able historic trading data. We make the assump-tion that that volatility of the Move-On projectand the three identified real options will be simi-lar to the volatilities of some key Norwegian IT/telecom stocks. Calculating by applying a 10(trading) days sliding window over 1 year oftrading data until May 2001 we obtain:

σ NER = 64 %, σ TEL = 25 %, σ ENI = 86 %,and σ SOI = 47 %.

One may assume that a start-up company likeMove-On will face more uncertainties and beexposed to more risk than the average telecom/IT stock. We also have to take into account thatthe Enitel stock has been subject to extreme vari-ations over the last year. Based on this we endup estimating an equity volatility of 70 % for theMove-On project, which is slightly higher thanNera. Hence, as a rough estimate we use thisvalue for all the three identified options:σO1 = σO2 = σO3 = 70 %.

3.3 Building IP Zones

3.3.1 DescriptionIn Move-On’s final business plan, IP zones(Wireless LAN deployment) is considered fromthe perspectives of a possible additional commu-nication network available to Move-On’s termi-nals and applications, recognising the fact thatseveral telecom players have started various IPzones network initiatives. However, early on inthe project, the possibility that Move-On itselfcould build and operate a network consisting ofinterconnected IP zones was considered. Thisidea was rejected due to anticipated large invest-ments and time delays associated with networkrollout, and the need to focus on the businessidea of community applications. In this section,we consider these two versions of deploying IPzones as possible options, i.e. opportunitiesMove-On may or may not take up some timeafter launch of its initial business (communityservices). In the remainder of this section wewill call this option Building IP-zones option(or simply BI), as opposed to the Integrating IP-zones option (or II), which is dealt with in thenext section.

NPV = E(X0) +E(X1)

(1 + r)+

E(X2)

(1 + r)2+ . . .+

+E(XT )

(1 + r)T=

T∑

t=0

E(Xt)

(1 + r)t


UncertaintyThe uncertainty related to the potential successof deploying IP zones have several sources:

• Market risks (or systematic risks)- Will macro-economic market conditions

develop favourably?

- Will the demand for IT/Telecom continueto grow at the current pace?

- What will be the outcome of other IP zonesinitiatives; e.g. Telenor Mobile Communi-cations, Netcom and WAN? Will they to-gether grow a new market? Will one actorend up taking the whole market?

- What will become the positioning of IPzones business compared to GSM/GPRSand UMTS?

• Private risks (or non-systematic risks)- Will Move-On succeed with the initial

product portfolio, and thereby being capableof (and willing to) expand with new prod-ucts?

- Will Move-On be capable of retaining theneeded in-house competence and necessaryrelations to partners?

- Will Move-On’s owners appreciate andsupport the progress of the initial businessand Move-On’s suggested ideas for expan-sion?

Although not being an exhaustive list, the above-mentioned risk factors indicate that there is asignificant level of uncertainty related to the IPzones option for Move-On. It is also likely thatsome of this uncertainty will be resolved withtime. Hence, it does make sense, at least intu-itively, to postpone the decision about thedeployment of IP zones until some of the abovequestions can be answered. In the followinganalysis we want to find out whether it pays offto delay this decision.

Associated activities and costsIn [Move-On Business Plan, 2001] a server sys-tem called MPCS (Multi Purpose CommunityServer) is proposed as the required networksolution to support the initial community ser-vices. The expected costs associated with theMPCS are 4 MNOK, split equally between hard-ware and software.

The functionality required to support the integra-tion of IP-zones is envisaged to be related toaccess control, authentication, and roaming.These are all functions that could be imple-mented centrally, and we assume that the re-

quired functionality could be implementedentirely by software on the MPCS platform, atan additional cost of 500 kNOK to cover theextra software development. We assume thatthere is no additional development time, becausedevelopment time of integrated and complexHW/SW systems is typically dominated by fixedelements such as testing, configuration andinstallation. For the same reason if the IP-zonesoption is not identified at the initial investmentdecision, a new SW development contract mustbe placed with the manufacturer, resulting insignificant cost and development impacts later.

The above applies to both IP zones options. Forthe BI-option, additional aspects include the costand time involved in rolling out an IP zone infra-structure. Assuming a network of 100 IP-zoneswith an average establishment cost of about200 kNOK, we end up with an additional net-work investment of 20 MNOK. Network rollouttime is in this case considered to be approxi-mately 12 months.

The costs involved with core network infrastruc-ture required to interconnect the IP zones areignored.

T0 TLC T1 TL-BI

Time

Timeline:In the following we shall assume the followingsequence of events:

T0: time for initial investment decision (Move-On project incl. options)

TLC: time for launch of community services(i.e. initial service offering);

T1: time for decision whether to launch IPzones (integrating or building option);

TL-BI: time for launch of the building IP-zonesoption (only relevant if exercised).

Then we will make the following assumptionsabout the spacing between the events:

TLC = T0 + 6 months: typical development timefor a complex HW/SW system

T1 = TLC + 6 months = T0 + 12 months: assum-ing 6 months of operations

The earliest possible TL-BI would be 12 monthsafter T1 to cater for IP zones network rollouttime.


3.3.2 CalculationsWe assume that if the option is exercised at matu-rity the total investment (of 20.5 MNOK) willhave to be committed, in addition to annual costsof 5 MNOK to cover marketing, sales, R&D, andother operational activities, which are specific tothe IP zones option. If the option is not exercised,these costs will not be committed. We assume amodest customer take-up, increasing with 5000customers each year after launch. Further, weassume a terminal value of zero for the IP zonesoption for simplification purposes. Finally, weassume revenues corresponding to a fixed sub-scription fee of 80 NOK per month per user.

The parameters relevant to this option are:

• S = the present value of the cash-flows gener-ated by the BI-investment;

• X = 20.5 MNOK. The required investment forthe BI-option;

• T = postponement time. We consider T = 0and T = 1;

• rF = 3 %;

• r = 15 %;

• σ = 70%.

First, we consider the NPV of building IP zonesnow, i.e. corresponding to a straight forwardNPV analysis ignoring the option to postpone.We get NPV = 1550 kNOK, which according tothe conventional NPV logic, would result in thedecision to invest in building IP zones now.

Second, we develop the cash-flow analysis fordelaying the building of IP zones with 1 year.Note that the investment is discounted with therisk free rate. We also note that NPV is -729kNOK, indicating a negative business case.

Then we go on to find the value of the option towait 1 year. When we apply an option calculatorto T = 1 year, S = 19174 kNOK (sum of all BI-related cash flows), together with the otherparameters stated above, we obtain the optionvalue of 4993 kNOK. This means that there isgreater value in waiting one year, than buildingnow.

Recalculating for T = 2, we get an NPV of -2650kNOK, and the option value is this time 5577

Year 0 1 2 3 4 5

Inv 20500 0 0 0 0 0

OPEX 5000 5000 5000 5000 5000 5000

Revenues 0 4800 9600 14400 19200 24000

FCF -25500 -200 4600 9400 14200 19000

x discount (rf) 1 0.97 0.94 0.92 0.89 0.86

x discount (rd) 1 0.87 0.76 0.66 0.57 0.50

DFCF -25500 -174 3478 6181 8119 9446

ADFCF -25500 -25674 -22196 -16015 -7896 1550

SUM of NPVs 1550

Year 0 1 2 3 4 5 6

Inv 0 20500 0 0 0 0 0

OPEX 0 5000 5000 5000 5000 5000 5000

Revenues 0 0 4800 9600 14400 19200 24000

FCF 0 -25500 -200 4600 9400 14200 19000

x discount (rf) 1 0.97 0.94 0.92 0.89 0.86 0.84

x discount (rd) 1 0.87 0.76 0.66 0.57 0.50 0.43

DFCF 0 -24251 -151 3025 5374 7060 8214

ADFCF 0 -24251 -24402 -21377 -16003 -8943 -729

SUM of NPVs -729

Table 6

Table 7


kNOK. In the graph in Figure 7 we plottedoption value versus postponement time.

The reason for this curve shape is related to theinput parameters to the Black-Scholes equation.Partly there is a benefit from delaying heavyinvestments, but the relatively high volatilityalso plays a beneficial role and is an importantdriver for the value of postponing.

The option value (for 1 year) of 4993 kNOKalso indicates how much we should be preparedto commit in preparatory costs prior to possiblyexercising the option. In this particular case, weassume no preparatory costs, i.e. the optioncomes for free. This is because we regard thenecessary investments (i.e. MPCS software andIP zones network infrastructure) to form theexercise price, to be paid at the time of execu-tion, rather than being preparatory costs to bepaid in advance. However, in a realistic setting,there would potentially be time or cost advan-tages associated with carrying out some of theseactivities prior to the decision. For example, car-rying the 500 kNOK MPCS SW upgrade for-ward to the Year 0 would be one such possibil-ity. This could be justified easily, because thecost of making that preparation (500 kNOK) iswell below the option value of waiting one year(4993 kNOK).

ConclusionWhereas the conventional NPV approach givesthe somewhat faulty advice to invest now, theadvice from the real option approach is veryclear: Do not invest now, because there isgreater value in waiting. Reconsider the situa-tion later when some of the uncertainty has beenresolved.

3.4 Integrating IP Zones

3.4.1 DescriptionThe previous section discusses the option tobuild and operate an IP zones infrastructure. A“lighter” IP zones business alternative is to inte-

6000

5000

4000

3000

2000

1000

00 1 2 3 4 5 6 7 8 9 10

Delay [in years]

Building IP zones - option value

Figure 7

TimelineIn the following we shall assume the followingsequence of events:

T0: time for initial investment decision(Move-On project incl. options);

TLC: time for launch of community services(i.e. initial service offering);

T1: time for decision to launch IP zones ornot (integrating or building option);

TL-II: time for launch of the integrating IP-zones option (relevant only if exercised).

The following assumptions have been madeabout the timeline events:

TLC = T0 + 6 months: typical development timefor a complex hard-/software system

T1 = TLC + 6 months = T0 + 12 months: assum-ing 6 months of operations

T0 TLC T1 TL-II

Time

grate IP-zones which already has been rolledout, for example by providing functions such asaccess control, roaming, etc. on a central server.Other IP zone networks could then “hook up” toMove-On’s server, and thereby participate in anaggregated network of interconnected IP zones.This would require lower investments, but thisbusiness case has big challenges associated withit. For example, a consequence of not owningthe complete network infrastructure is thatMove-On will need to consider revenue sharingarrangements with the IP zone owners. In thefollowing analysis we explore if it pays off todelay this decision.

UncertaintyThe uncertainty related to the business opportu-nity to integrate IP zones is similar to buildingIP zones, see Section 3.3.1.

Associated Activities and CostsThe additional software change to cater forauthentication, authorisation, accounting androaming functionality in the MPCS, with an esti-mated cost of 500 kNOK, is also required forthis option. However, the heavy capital invest-ment for network infrastructure required in theBI-option, is not applicable here. As for the BIoption, costs associated with interconnecting theIP zones are ignored.


TL-II will happen no earlier than T1 (or shortlyafter), assuming the required functionality isimplemented and can be switched on at thematurity of the option.

3.4.2 CalculationWe assume that at maturity the total investment(of 500 kNOK) will have to be committed, inaddition to an annual cost of 5 MNOK to covermarketing, sales, Research&Development, andother operational activities, which are specificto the IP zones option. If the option is not exer-cised, these costs will not be committed. As forthe BI option, we assume a modest customeracquisition, increasing with 5000 customers eachyear after launch. Finally, we assume zero termi-nal value and revenues corresponding to a fixedsubscription fee of 40 NOK per month per user.The latter is 50 % of the fee used in the BIoption, based on the rationale that Move-On willhave to enter a revenue sharing scheme with theIP zone owners.

The parameters relevant to this option are:

• S = the present value of the cash-flows gener-ated by the II-investment;

• X = 500 kNOK. The required investment forthe II-option;

• T = postponement time. We consider T = 0and T = 1;

• rF = 3 %;

• r = 15 %;

• σ = 70%.

First, we consider the NPV of integrating IPzones now, corresponding to a straight forwardNPV analysis, ignoring the option to postpone.We get NPV = -355 kNOK, which according tothe conventional NPV logic, would result in thedecision not to invest in integrating IP zonesnow.

Second, we develop the cash-flows analysis ofdelaying the integration of IP zones with 1 year.In this case the NPV is -360 kNOK, still indicat-ing a negative business case opportunity. Nextwe find the value of the option to wait 1 year.Crunching the numbers: T = 1 year, S = 126kNOK (sum of all II-related cash flows), to-gether with the other parameters stated above,

Year 0 1 2 3 4 5

Inv 500 0 0 0 0 0

OPEX 5000 5000 5000 5000 5000 5000

Revenues 0 2400 4800 7200 9600 12000

FCF -5500 -2600 -200 2200 4600 7000

x discount (rf) 1 0.97 0.94 0.92 0.89 0.86

x discount (rd) 1 0.87 0.76 0.66 0.57 0.50

DFCF -5500 -2261 -151 1447 2630 3480

ADFCF -5500 -7761 -7912 -6466 -3835 -355

NPV -355

Table 8

Table 9Year 0 1 2 3 4 5 6

Inv 0 500 0 0 0 0 0

OPEX 0 5000 5000 5000 5000 5000 5000

Revenues 0 0 2400 4800 7200 9600 12000

FCF 0 -5500 -2600 -200 2200 4600 7000

x discount (rf) 1 0.97 0.94 0.92 0.89 0.86 0.84

x discount (rd) 1 0.87 0.76 0.66 0.57 0.50 0.43

DFCF 0 -4833 -1966 -132 1258 2287 3026

ADFCF 0 -4833 -6799 -6931 -5673 -3386 -360

NPV -360


the option value turns out as 1.7 kNOK. Thismeans that there is a certain value, although verysmall, in waiting one year, compared to startintegrating now. Repeating the exercise forT = 2, we get an NPV of -362 kNOK and anoption value of 6.4 kNOK.

We may wonder why the values of BI-optionsand II-options are so different? The only differ-ence is in S, the value of the investment projecttoday, and X, the exercise price of acquiring theinvestment project. The BI option has a muchhigher value of X (20.5 vs. 0.5 MNOK) so thethere is a more positive effect for BI by postpon-ing this investment. The BI option also has amuch larger S-value (19174 vs. 126 kNOK), sothe underlying investment project also has amuch higher value.

However, it is important to note that if the II-option had some other characteristics that weresignificantly different, then the option valueswould be affected accordingly. For instance,an increase in volatility or expected revenueswould give a higher option value.

For illustration, the graph in Figure 8 shows howthe II option value (here for T = 2) depends onvolatility.

ConclusionThe conventional NPV approach gives the cor-rect advice in this case, which is not to invest,and the real options approach strengthens thisconclusion by indicating that there is not muchvalue in postponing the decision.

3.5 Shutdown OptionIs there value in having the possibility to shutdown an operation? Absolutely! The decision toshut down is in many circumstances a very dras-tic decision, and one should also consider thepossibility of just scaling down or contractingthe operation. The option to shut down or con-tract is valued as a European put option on therelevant part of the project, with exercise priceequal to the potential cost savings. Hence, the

real value of shutting down a running project liesin the money that potentially is saved on futureinvestments. However, one also has to take intoaccount the lost revenues and different deals thathas to be negotiated with different parties likeemployees and customers. In fact this last pointis very important, and very hard to value. Howdo you compensate customers that no longer getthe telecommunication service that the companyhas promised them, and how to compensate theemployees?

The value of the shutdown option does not put atotal value on the project as a whole, but merelythe option value reflect how much investmentsthe management can undertake on securing theoption to abandon the whole project. Hence, ifthe real cost of obtaining (buying) the option isexceeding the put option value, it is not econom-ically viable to buy this option. Hence, it is pos-sibly too expensive to shut down the project, andyou are better of continuing the operation.

3.5.1 Description of Move-On Shut DownOption

We will in the following investigate the optionof terminating the project in two years time. Thecourse of the project is that there are heavy ini-tial investments followed by two years of justminor investments. From the third year substan-tial investments will be performed each year,hence it is of interest for the Move-On project toestimate the value of acquiring the option to ter-minate the project after two years. In the two-year period much of the uncertainty will dis-close, and future market conditions for our prod-ucts should be judged as either good or bad. Inthe case of good market conditions Move-Onshould continue the project with its investmentsand update the cash flow prognoses. In the othercase, one should be able to shut down the Move-On operation limiting the losses. An attempt tovalue the shut down option will indicate howmuch we are willing to spend on securing thedeals with employees and customers to be ableto stop the business in two years time. In thenext section we calculate the value of this putoption.

3.5.2 CalculationWe have to separate between the following:First, the option in itself has value, and we haveto be prepared to pay for negotiating the dealswith the employees and customers in order forthem to accept that the service is possibly shutdown in two years. Second, at maturity we willhave to pay the exercise price, which is theagreed compensation to involved parties. Inother words we should be willing to pay forbeing able to possibly pay compensation! Thismight seem awkward, but limiting the downsidein a risky investment does have its price. This

2

120

100

80

60

40

20

0

Volatility [%]

Opt

ion

valu

e

50 100 150 200 250 300 350 400

Figure 8


way we are only “buying” the upside potential inthe investment project.

Estimating the amount of compensation – thestrike price – is practically impossible. Here wehave to make an educated guess based on thenumber of employees and the number of cus-tomers. In the following (Table 10) we choose10 MNOK as the options strike price.

To find the value of the option today, recall theproposed free cash flows for the Move-On pro-ject (see the above table). The accumulated costsavings relating to the shut down will be the dif-ference in the accumulated discounted cashflows between year 5 and year 3. Subtractingthis we end up with the current value of our pro-ject assets as 1124 kNOK.

The Black-Scholes parameters with relevant val-ues to the present option are:

• S = 1124 kNOK• X = 10000 kNOK• T = 2• rF = 3 % (see 3.2)• r = 15 % (see 3.2)• σ = 70 % (see 3.2)

Feeding the numbers into an option calculator weend up with a value of the option of 8310 kNOK.Hence, we are in principle willing to buy theflexibility of possibly shutting down the Move-On operation in 2 years for 8.3 million NOK.

In the above calculation we have not includedan estimate of the terminal value. This will ofcourse have an impact on the valuation of theshutdown option. The larger the terminal value,the less the option value, this is because themore value that is coming in the future the lessinterested we are in shutting down the business.

ConclusionWe see that the conventional NPV approachdoes not really know how to value the option ofshutting down an operation. There is very muchvalue hidden in this real option, and hence our

strategic recommendation is to acquire thisoption. This is because it will add substantialvalue to the Move-On operation by limiting thedownside risk of the investments.

3.6 Other Move-On Real OptionsIn the Sections 3.3 and 3.4 above, we have con-sidered two versions of IP zones deployment,which both are typical growth or expansionoptions. Unlike the IP zones options, which con-sider expanding with new products, one couldalso consider the opportunity to expand the busi-ness into new geographical markets. In the caseof the Move-On project, the other Scandinaviancountries and Germany have been considered aspossible future target markets, where Move-Oncould reuse and benefit from the products andexperience developed in Norway. Instead ofmaking the decision to enter the German or theSwedish market initially, which would increasethe overall risk associated with the Move-OnProject, such an option would allow Move-Onto maintain the possibility to enter one of thesemarkets at a later stage, at the cost of carryingout the necessary preparations.

One could possibly treat Move-On as a stagedinvestment project within the same geographicalmarket. An initial launch of the IP zones will berestricted to e.g. a few cities in Norway. Furtherexpansion in Norway, to other cities and towns,can then be staged by selecting for instance 5new areas to be populated with IP zones in thenext stage. Depending on the popularity of theprevious investments, demographic profiles, etc.,further investments can be halted in the case ofunfavourable market conditions.

The size of the community server system is cru-cial in the IP zone business case. Investing in alarger server than actually needed is an ineffi-cient investment. However, the opposite case ofnot having a large enough server to supply for anunexpected and surprisingly huge customer baseis probably worse, because expanding an exist-ing system typically is more expensive thanallocating for the same capacity initially. Thus,incorporating a scalable server solution from

Year 0 1 2 3 4 5

CFO -5909 -16937 -16198 -4642 31226 97947

Invest 7066 1590 9826 21190 37337 54495

x discount (rf) 1 0.97 0.94 0.92 0.89 0.86

x discount (rd) 1 0.87 0.76 0.66 0.57 0.50

DFCF -12975 -16110 -19678 -16985 -3494 21603

ADFCF -12975 -29085 -48762 -65747 -69241 -47638

NPV -47638

Table 10


the start can turn out to be a very valuable realoption.

Having identified several different real options,a few words have to be said about the collectivevalue of all the real options acting together inthe same business case. Multiple interacting realoptions follow a non-additivity law. This meansthat the value of the individual real options cannot just be added together to get the collectivevalue of all the options that are active at thesame time. The options interact in a non-linearway.

4 ConclusionsIn the Move-On business case we have identi-fied three real options; two growth options andthe option to terminate the whole project. Thegrowth options are related to potentially expand-ing the product portfolio with IP-zones, and theoption to abandon is the strategic choice of shut-ting down after a period of operations.

The first two growth options are the IP-zone“wild cards”. IP-zones is emerging as a possiblecompetitor to the UMTS networks, and theimportance of recognising this as an importantstrategic option is attempted in valuing theoption to expand the product portfolio with IP-zones after some time of initial operation in therelevant market. In Move-On’s case this is Nor-way initially, but the arguments also apply toany other market. Two variants of this option areconsidered, the first being the option for Move-On to build and operate an IP zones networkitself, the second is to act as an integrator forother players building IP zones. To prepare forthese options, it may be necessary to commitsome additional investment costs to the multi-purpose community server platform (MPCS) ini-tially. Once the uncertainty about Move-On’sinitial products and ongoing IP zones initiatives(carried out by other players) have been revealed– which could be evaluated some time after theinitial service launch – the decision whether todeploy IP-zones can be made. The value of theseoptions then reflects how much Move-On shouldbe willing to spend on preparing for IP zonesdeployment.

Analysis of the building IP-zones option, re-vealed a greater value in postponing the decisionwith one year (option value 4993 kNOK) thanbuilding IP-zones immediately (NPV: 1550kNOK). Whereas the conventional net presentvalue (NPV) approach gave the somewhatflawed advice to invest straight away, the adviceresulting from using the real option approachwas very clear: “Do not invest now, becausethere is greater value in waiting. Reconsiderthe situation in one year”.

For the integrating IP zones option we foundthat the value of postponing the decision wasmarginal. The conventional net present valuefor integrating IP zones was found to be -355kNOK, and the option value associated withdelaying the investment decision was negligible;1.7 kNOK for one year and 6.4 kNOK for twoyears. So, in this case we conclude that the con-ventional net present value approach gives a sen-sible advice; “Do not invest”. The real optionsapproach actually strengthens this argument byindicating that there is not much additional valuein postponing the decision.

We have also investigated the shutdown option,i.e. the option to terminate the whole projectafter two years. During that period of timeMove-On will have launched its first line ofproducts and we will have a much better under-standing of how the market judges the productideas. Depending on how successful the productlaunch is, the project can be terminated at theend of the two-year period. The value of thisoption reflects the amount that Move-On shouldbe willing to invest in obtaining the possibilityof terminating business in two years time.

To conclude, the strategic recommendationsfrom the real options analysis are:

• Even though building IP-zones appears as anattractive investment opportunity, do not makethe investment decision now. There is greatervalue in waiting for some time, for exampleone year, and then reconsider the situation.

• With the assumptions made, integrating IP-zones does not appear like an attractive busi-ness opportunity. Neither is there any signifi-cant option value by postponing the decision,so there needs to be fundamental changes tothe assumptions made (e.g. improved cashflow estimates) before it is worth reconsider-ing this option.

• There is very much value hidden in the optionof shutting down the operation after two years,and hence our strategic recommendation is toacquire the option. This is because it will addsubstantial value to the Move-On operation bylimiting the downside risk of the investments.

References[Alleman & Noam 1999] Alleman, J, Noam, E.1999. The new investment theory of real optionsand its implication for telecommunication eco-nomics. Kluwer Academic Publishers.

[Amram & Kulatilaka 1999] Amram, M, Kulati-laka, N. 1999. Real Options – Managing Strate-gic Investment in an Uncertain World. HarvardBusiness School Press.


[Audestad 1998] Audestad, J A. 1998. Telecom-munications and Complexity. Telektronikk, 94(3/4), 2–20.

[Bøhren & Michaelsen 2001] Bøhren, Ø,Micaelsen, T. 2001. Finansiell økonomi – teoriog praksis. Skarvet Forlag.

[Dixit & Pindyck 1994] Dixit, A K, Pindyck, RS. 1994. Investment under Uncertainty. Prince-ton University Press.

[Fleten et al. 1998] Fleten, S-E, Jørgensen, T,Wallace, S W. 1998. Real Options and Manage-rial Flexibility. Telektronikk, 94 (3/4), 62–66.

[Hull 2000] Hull, J C. 2000. Options, Futures, &Other Derivatives. Fourth Edition. Prentice Hall.

[Luehrman 1998] Luehrman, T A. 1998. Invest-ment Opportunities as Real Options: GettingStarted on the Numbers. Harvard BusinessReview, July-August, 51–67.

[Move-On Business Plan 2001] 2001. Move-OnBusiness Plan, …with the Move-On communityacross networks and terminals. June 17.

[Trigeorgis 1996] Trigeorgis, L. 1996. RealOptions – Managerial Flexibility and Strategyin Resource Allocation. The MIT Press.

[Wilmott et al. 1995] Wilmott, P, Howison, S,Dewynne, J. 1995. The Mathematics of Finan-cial Derivatives – A Student Introduction. Cam-bridge University Press.

Telektronikk 4.2001

1 IntroductionNowadays, telecom services run in the telecomoperator’s domains. The networks belong to theoperators and are totally closed for other parties.The service providers wishing to offer service tocustomers through the networks must have theirservices running in the network operator’sdomain. In that case, the service providers havelittle or no control of the applications and it isdifficult to integrate the service with their data-bases and services. The development of servicesintended for theses networks is quite complexsince it must be carried out in close collaborationwith the operator and their suppliers.

With the advent of the Internet, which offers alarge and diverse range of applications, thedemand of more services in telecommunicationis increasing. Further, the introduction of 3rdgeneration mobile systems like e.g. UMTS cre-ates needs for new services. To be able to differ-entiate with competitors, a telecom operatorneeds to be able to offer more value-added ser-vices. The development of such services costsquite a lot in terms of resources but also in time.In order to cope with the situation the operatorsare forced to open their networks and let 3rd par-ties participate in the development, deploymentand offering of services (see Figure 1-1). Theservice providers benefit from the opening oftelecom networks in that they can develop theirown applications running at their own applica-tion servers. They can run the same applicationsthat earlier had to run in the telecom operator’sdomain, but now it is much easier for the serviceproviders to integrate the services with eachother and to use their own databases and infras-tructure. Such an opening brings also opportu-nity to Mobile Virtual Network Operators(MVNOs) with expertises in specific sector toenter into the mobile business.

This article starts with a brief explanation of thedevelopment and deployment of value-addedservices. A summary of the 3GPP architecture isgiven next. The article will then clarify how theOpen Service Access can be beneficial forMVNOs.

2 Development andDeployment of Value-added Services

In current mobile systems such as GSM there area number of network capabilities that are usefulfor value-added services/applications, such as:

• Setting up a connection between to devices;

• Monitoring the status of a terminal (switchedoff, switched on, busy etc.);

• Locating a user’s terminal;

• Sending messages to a user’s terminal.

An example of a value-added service that usessome of the capabilities mentioned above is aTaxi dispatch. A customer calls, via his mobilephone, a phone number associated with the Taxidispatch to order a taxi. The Taxi Dispatchapplication queries the location of the customerfrom the mobile system and then also requests

The Open Service Access andOpportunities for MVNOsD O V A N T H A N H A N D G U N V A L D M A R T I N G R Ø D E M

Gunvald M Grødem (25) obtain-ed his Engineering degree inelectronics and telecommuni-cation from Oslo College ofEngineering (1999) and his MScdegree in Communication Sys-tems from Univ. of Oslo, Dept ofInformatics (2001). During hiswork on the MSc degree heworked at Telenor R&D, focusingon Corba, Parlay/OSA, JAIN, 3rdparty service provisioning, net-work independent services andthe Session Initiation Protocol(SIP). He is now working in theconsultancy company eScienzaAS, which is specialised in prof-itable implementations of tech-nology based business models,focused on the convergencearea electronic channels, multi-media and mobility.

[email protected]


[email protected]

Aiming at opening up the value chain and allowing third parties development and deployment of value-

added services, the 3rd Generation Partnership Project (3GPP) promotes the Open Service Access

(OSA). This article will give a comprehensive introduction to OSA and discuss how OSA can benefit

third parties, especially Mobile Virtual Network Operators (MVNOs). The focus is especially on the

flexibility and the ability to adapt to customers needs like e.g. personification of services. Another

important issue is the integration of services and customer databases.

Figure 1-1 Opening of telco networks

The IT WorldRich in applications

and developers

The Telco NetworksRich in capabilities

API

98


the location of all taxis in order to choose theclosest one. The locations of all taxies inside acertain boundary could e.g. be shown on a mapof the area where the customer is located. Next,the Taxi dispatch calls the taxi in question andinforms the taxi driver about the order. Alterna-tively, the Taxi dispatch could also send a mes-sage informing the taxi about the order.

Another example of value-added service is apersonal forwarding service. A user can definewhich devices the calls should be forwarded tofor example in case of no answer from themobile phone. For example, the user could spec-ify that calls should be forwarded to the fixedphone at work between 0900 and 1700 and tothe fixed phone at home otherwise. If there isno answer at any phone, the call should be for-warded to the voice-mail. When someone callsto the user’s mobile phone, the service checksfirst with the mobile system for the status of themobile terminal. If the mobile phone is switchedoff the call is routed according to the rules de-scribed above. If the mobile phone is switchedon, the call is first routed to the mobile phone.If there is no answer then the call is routedaccording to the rules described above.

As shown in the examples, the network capabili-ties like call control, terminal status and userlocation are very valuable for value-added ser-vice but are unfortunately not available outsidethe GSM network domain (see Figure 2-1).

To build an application, which makes use of thementioned capabilities is quite difficult. The dif-ferent network elements that provides differentcapabilities, e.g. the IN system providing capa-bilities for connecting devices, the HLR provid-ing location information and terminal status, thelocation server providing location information,the SMSc providing capabilities to send mes-sages to users, etc. have different interfaces andprotocols. The developers of the value-addedservices need to have considerable knowledgeabout different network elements, interfaces andprotocols. Hence, many developers may beneeded since each developer may be expert ononly one single network element, protocol orinterface. An application developed for one ven-dor’s equipment is not necessarily portable toanother vendor’s equipment.

If a taxi company wants to provide a taxi dis-patch application to the customers, it needs tocollaborate with the network operator in orderto develop the service. The taxi company needsto specify the requirements of the application tothe network operator, who then has to examinewhether it is possible to develop. Only after thatthe network operator and the service providerhave agreed, the expert developers working at

the network operator can develop, deploy andrun the service. As we can see the developingprocess is complex and time and resource con-suming.

As stated earlier, the telecom operators need todevelop and deploy many new value-added ser-vices in order to cope future needs and the Inter-net’s ability to supply services. In order to

Figure 2-1 Applicationrunning inside a GSM

network operator’s domain

Figure 2-2 Application usingstandardised interfaces insidea network operator’s domain

Figure 2-3 3rd partyapplication using network

capabilities throughstandardised interfaces

IN/SCP

HLR

SMSc

GSMnetworkdomain

Taxidispatch

Proprietary interfaces

IN/SCP

HLR

SMSc

GSMnetworkdomain

Taxidispatch

Standardised interfaces

IN/SCP

HLR

SMScGSMnetworkdomain

3rd partyserviceproviderdomain Taxi

dispatch

CustomerDatabases

EnterpriseServices

Open Standardised interfaces


accomplish this there is a need to improve thedevelopment process for value-added services.One approach is to standardise the interfaces todifferent network capabilities. By standardisingthe interfaces the developers do not need to haveknowledge of the proprietary interfaces specificfrom each equipment vendors. The developmentand deployment of new services are going to bemuch easier, and consume less time andresources.

But standardising the interfaces to network capa-bilities does not only benefit the developersworking for a network operator. It will be easierto let 3rd parties get access to network capabili-ties through open standardised interfaces. Byusing open standardised interfaces the applica-tions may be developed, deployed and run by 3rdparty service providers. The network operatorshould supply mechanisms that ensure that onlyauthorised 3rd parties get access to these inter-faces. This solution makes it much easier for the3rd party service providers to integrate the value-added service with other enterprise solutions andcustomer databases, since the application is nowrunning in the value-added service provider’s(VASP’s) domain (see Figure 2-3).

3 3GPP’s Open ServiceArchitecture

The 3GPP (Third Generation Partnership Pro-ject), which purpose is to prepare, approve andmaintain globally applicable technical specifica-tion and technical report for a 3rd generationMobile System (UMTS), based on an evolutionof GSM/GPRS, has specified the Open ServiceArchitecture (OSA) [1], which enables networkoperators and 3rd party service providers accessto network capabilities such as Call Control,Messaging, Location Information, etc. in a sim-ple and secure way. OSA is 3GPP’s suggestionto solution to the problems in the developmentand deployment of value-added services men-tioned earlier. The 3GPP has also proposed OSAas one approach to allow 3rd party service provi-sion for UMTS.

ArchitectureIn OSA, 3GPP has standardised a set of openinterfaces to the network capabilities in the wire-less (GSM/UMTS) network [2]. The interfacesare specified in OMG IDL [3], which is a de-facto standard interface definition language. Theinterfaces are grouped into Service CapabilityFeatures (SCF). The SCFs are abstractions ofthe underlying network functionality. For exam-ple, the Call Control SCF includes all the inter-faces that deal with controlling calls i.e. call for-warding and call events. The SCFs run on a Ser-vice Capability Server(s) (SCS(s)), which serveas a gateway between the OSA interfaces andthe proprietary interfaces specific to each net-work capability, e.g. MAP, INAP, ISUP, CAP,etc. Hence, the standardised interfaces aremapped into the specific proprietary interfaces tothe corresponding network capabilities (see Fig-ure 3-2). For example, the interfaces in the CallControl SCF have to be mapped into the inter-faces to the CAMEL (Customised Applicationfor mobile network enhanced logic) system usedin the current GSM/UMTS network. The proto-cols and interfaces may vary from differentimplementation i.e. Ericsson and Alcatel.

Figure 3-1 Overview of OpenService Architecture [1]

Figure 3-2 SCS as gatewaybetween OSA interfaces andproprietary interfaces

OpenServiceArchi-tecture

Servers

framework UserLocation

CallControl Service capability server(s)

discovery

Applicationserver

Application

OSA API

Interfaceclass

OSA internal API WGWWPP

CSEHLR

E.g. Location serverMExE serverSAT server

HLR IN/SCP MSC/VLR

Applications

GSMSystem

SCS

OSA Interfaces

MAPInterface

INAP/CAPInterface

MAPInterface

ProprietaryInterfaces


ApplicationsTypically applications that can use OSA arepersonal forwarding services, personal commu-nication services, Taxi dispatch, pizza ordering,location based applications (find the nearest cin-ema, gas station, hotel, map navigation, etc.),Bank service, etc. The Service providers are freeto use the SCF in the way that they want as longas they follow the agreement with the OSAprovider. The Telco world, when opening theirnetworks with OSA, will benefit with many newand fancy “killer applications”.

Applications that wish to use SCFs through OSAcan either run on application servers locatedinside the network operators domain or on appli-cation servers running in a 3rd VASP’s Domain(see Figure 3-3). Hence, the OSA may benefitthe network operator as well as the 3rd partyVASP in the way that the developers only needto know the OSA interfaces, no matter whetherthey develop applications for the network opera-tor (OSA provider) or for the 3rd party VASP.This will hopefully lead to faster and simplerdevelopment both for network operators (OSAproviders) and 3rd party VASPs.

Service Capability Features (SCFs)The network service capabilities are available toapplications through the SCFs. Currently, thefollowing SCFs are specified in OSA Release1999 [1]:

• Call Control: Allows applications to controlnetwork-initiated calls e.g. call barring, callforwarding, pre-paid calls, etc.;

• User Location: Enables applications to gethold of users’ current geographical position,delivered by the network;

• User Status: Provides applications with mech-anisms to obtain the current status of a user’sterminal;

• User Interaction: Allows applications to inter-act with end users by sending messages/announcements to users and to collect infor-mation from them. The applications may alsocontrol a message sent by a user;

• Data Session: Enables applications to controldata sessions initiated by a terminal, e.g. pre-paid data sessions;

• Terminal Capabilities: Provides the applica-tions with mechanisms to obtain the currentcapabilities of a user’s terminal.

3GPP are currently working on SCFs for eCom-merce, Instant Messaging, Multimedia controletc., which will be included in future releases ofthe OSA specification when they are ready.

FrameworkIn order to ensure secure and simple access tothe OSA SCFs, 3GPP has specified an OSAFramework. The Framework is responsible forprotection against misuse of the network re-sources in the telco networks and to ensureintegrity of the system.

Before a third party can use any mechanismsprovided by the OSA framework, the OSA

Figure 3-3 OSA applicationsresiding inside or outside the

telco domain

IN/SCP

HLR

SMSc

GSMnetworkdomain

3rd partyserviceproviderdomain Taxi

dispatch

CustomerDatabases

EnterpriseServices

OSAGateway

OSA Standardised interfaces

Taxidispatch Customer

Databases

IN/SCP

HLR

SMSc

OSAGatewayGSM

networkdomain



provider and the third party needs to establish aService Level Agreement (SLA) between them.The SLA is an off-line business-level transac-tion, were the two parities agrees on terms ofuse of OSA Services.

Applications that wish to use network capabili-ties through OSA need to get authenticated withthe framework and visa versa. The OSA frame-work is able to differentiate between applica-tions, e.g. the applications running inside thenetwork operator’s domain could be marked“trusted”, which do not need to authenticateitself and applications running outside the net-work operator’s domain as “distrusted”, whichalways have to authenticate itself. The securitypolicy and mechanisms of the OSA frameworkare quite flexible and it is up to the OSA pro-vider to decide which security mechanisms toprovide (e.g. CORBA Security, SSL etc.) andwhich policies are going to be used.

The OSA framework has much in common withthe ODP Trading Functions described in [4].Hence, the OSA framework works like a“Trader” for the telco network domain. It pro-vides applications with mechanisms to list anddiscover which services the OSA frameworkcurrently supports. Hence, the different SCFtypes supported in the network domain, registerthemselves with the OSA framework using aninternal interface to the OSA framework. Theapplications also have the possibility to checkthe access right to each service at run- time(access to services are agreed in the SLA).Through the Framework’s interface an applica-tion can select the service that it wish to use, e.g.Call Control or User status, the framework thenuses the Service Factory to make an instance ofthe selected service and returns a reference to theservice to the application (see Figure 3-4). Theapplication can then use the reference to accessthe SCF.

Communication betweenApplications and OSAIn order to be able to use the framework and net-work SCF offered by OSA, the applications needto be able to communicate with OSA. This couldbe a difficult task due to different computers,operating systems, implementation language,both at the Application server and the OSAservers. The applications have to set up a con-nection to OSA framework and to send mes-sages to OSA, in a format that is understandablefor OSA. This could be a complex task and willmake it very difficult for applications to useOSA.

CORBA [3] is an object-oriented technique forclient-server communications. It enables a clientobject to invoke methods on Remote object(Server object)’s interface, specified in OMGIDL. Hence, CORBA is a Remote Method Invo-cation (RMI) middleware. A “remote object”means that the object is running in a differentprocess than the client object, either on the samecomputer or in any computer connected in acomputer network. CORBA hides the complex-ity of the communication between the client andthe server objects, hence the client does not needto worry about how the communication to theserver takes place e.g. it does not need to set upconnections and to send messages to the server.In addition both the server and the client objectmay be implemented in any object-oriented pro-gramming language like Java, C++, SmallTalketc. and they can run at any Operating System.CORBA hides all the differences between theclient and the server. By using a CORBA IDLcompiler on the interfaces specified in OMGIDL, is it possible to generate all the CORBAspecific code and the corresponding APIs, whichenables the communication between the clientand the server. There is a specific compiler foreach programming language, e.g. if the client/server are going to be implemented in Java, anIDL-to-Java compiler needs to be used. In orderto communicate with the server, the client onlyneeds to include the generated CORBA code(client stubs) in the client code and use anObject Request Broker (ORB). The developerscan then program (call methods) against theserver’s API, generated by the IDL compiler, asif the server object is a local object (running inthe same process as the client). The server objectalso needs to use an ORB and some generatedCORBA code (server skeletons) in order to beable to deliver the messages from the client tothe right server object.

By using CORBA as described above, the appli-cations can communicate and access the OSAinterfaces in a simple way. The OSA interfacesspecified in IDL can easy be compiled to OSAAPIs, in the programming language that the

Figure 3-4 The OSAFramework

CallControl

SCF

UserStatusSCF

UserInteraction

SCF

Applications

OSAFramework

(Trader)

Internal OSAinterfaces

service factory(get service manager)

registerservice

OSA Frameworkinterface

list services,discover services,

check access rights,select service

• • •

3rdparty

domain

ORB

OSA interfaces

Stubs

API

Application

GSMnetworkdomain

ORB

Skeletons

API

OSAserver(s)

CORBA

RMI


developers want to use in the OSA application(see Figure 3-5).

OSA Application ScenarioIn this section we will describe a scenario wherean application uses OSA Call Control (CC) SCF tocontrol the processing of a call. Note, that the sce-nario will be quite similar when other SCFs thanCC are used by the application. The purpose ofthis section is to give an example of how an appli-cation proceeds, in order to use an OSA SCF.

At first, a Service level Agreement has to beestablished between the VASP and the OSAProvider. In the agreement process the two par-ties agree on term of use of the CC SCF. Theapplication then needs to get a reference to the“initial point of access” to OSA, using a NamingService, Trading Service, etc. The application isnow able to start interacting with the OSAframework (see Figure 3-6).

At first the application calls ‘initialeAuthentica-tion’ (1), in order to start the authentication pro-cess. The application then calls ‘selectAuth-Method’ (2) specifying which authenticationmethod that should be used in the authenticationprocess. The framework always needs to authen-ticate itself to the application. In order to authen-ticate the framework, the application calls‘authenticate’ (3) on the framework’s ‘Authenti-cation interface’. The application only needs toauthenticate itself to the framework if the appli-cation is of “distrusted” type to the framework.In order to authenticate the application, theframework calls ‘authenticate’ (4) on the appli-cation’s ‘Authentication interface’.

The application is now in an authenticated ses-sion with the OSA framework and it may accessthe SCFs that the framework supports. Note, thatthe authentication process (1-4) only needs totake place at initial contact with the frameworkor after an application have called ‘endAccess’,which ends the authenticated session. In order tobe able to access the SCFs, the application needsa reference to the ‘OSA Access interface’, whichit obtains when it calls ‘accessInterface(OSA_Access)’(5).

The application is now able to discover andselect the CC SCF, and calls first ‘obtainInter-face(Discovery)’(6), which returns a reference tothe ‘Discovery interface’. The application thencalls ‘listServiceTypes’(7), in order to get a listof the SCFs supported by the framework. Next,the application gets a full description of the CCSCF, including all the properties it supportswhen calling ‘describeServiceType’ (8). Theproperties typically deal with non-functional andnon-computational aspects of the SCFs Theapplication specifies the desired properties for

the CC and calls ‘discoverService’ (9). Inresponse, the application gets a unique ‘ServiceID’ for the CC SCF and a list of the associatedproperty values. The result from (9) is used asparameters for the application’s call for ‘select-Service’ (10). The application’s call for ‘select-Service’ makes the framework to ask the appli-cation to digitally sign the service agreementtext, ‘signServiceAgreement’ (11), which in turnmakes the application to ask the framework todigitally sign the service agreement text,‘signServiceAgreement’ (12). If the frameworkagrees to sign the agreement text, the frameworkcontacts the CC Service Factory and asks it tomake a new Call Manager (CM) object, ‘getSer-viceManager’ (12-2). The Service Factorymakes a new CM object (12-3), and the refer-ence to the object are then returned to the appli-cation via the framework. The application hasnow obtained a reference to a Call ManagerObject and is able to perform call related issues.The application has access to the object until the

Figure 3-5 OSA communication

Figure 3-6 OSA applicationscenario

CallControlServiceFactory

CallControl

ManagerObject

CallObject

Appli-cation

12-3 “new”

OSAFrame-

work

14 “new”

Service CapabilityServer(s)

12-2 13 15 16

12

3 & (4)56789

101112


‘terminateServiceAgreement’ method is called,which destroys the CM object or the ‘end-Access’ method is called, which ends the currentauthenticated session with the framework.

The application wish to get notified when callarrives for a certain number ‘x’, thus the applica-tion calls ‘callEventNotify(x)’ (13) on the CMobject’s interface. When a call arrives for x, thecall processing in the network is paused; the CMspawns a new Call Object (14) and returns thereference to the Call object to the application, inthe call for ‘eventNotification(x)’ (15). Theapplication decides to route the call to ‘y’ andcalls the ‘routeReq(y)’ (16) method on the Callobject’s interface. The call processing in the net-work resumes and the call is routed to ‘y’.

4 How can OSA Benefit MobileVirtual Network Operators?

Mobile Virtual Network Operators (MVNOs)do not own a radio infrastructure but rely onthe Mobile Network Operator (MNO)’s one. Inorder to be able to compete and survive it is cru-cial for them to offer attractive value-added ser-vices. To develop services and application inclose collaboration with an MNO is a complexand slow process. It is also difficult for theMVNOs to adapt applications and servicesaccording to the customers’ demands.

With OSA, the development and deployment ofvalue-added services is much easier for theMVNOs. The applications can run on applica-tion servers inside the MVNOs domains usingSCFs through OSA (see Figure 4-1). TheMVNOs can use their own software developersto develop and deploy the value-added servicesin an appropriate speed, without being depen-dent of work done by the developers working forthe network operators. It will also be easier tointegrate the value-added services with other ser-vices and applications, e.g. databases, directoryservices, Web Services, Enterprise Services, run-ning in the MVNOs’ domains.

It will be much easier for the MVNOs to providevalue-added services according to the customers’demands. For example, by combining Web ser-vices with the customer databases, the value-added can be personified, e.g. personal forward-ing service, personal ‘buddy lists’, etc. The cus-tomers could for example specify how they wanta service to operate through a web page, e.g.where to forward the call, who to put on the‘buddy list’, etc. The customers profile could besaved in some databases/directory services andeasily be used by the value-added services.

Implementation of Value-addedServices using OSAThere are many alternatives of implementingvalue-added services using OSA. We suggest anapproach that enables fast implementation ofvalue-added services using OSA and easy inte-gration with other services (see Figure 4-2).

Each value-added service could itself be a clientto the OSA API, using call control, user status,etc, but in that case each application has to knowthe specific CORBA code (client stubs) to theOSA API and an OSA client has to be imple-mented in each value added service. To improvethis, the MVNOs could use Java’s componentmodel, Java Enterprise Beans (EJB) [5]. It isthen possible to construct, “implement once runeverywhere”, software components that work asclients for the OSA APIs, e.g. a Call ControlEJB. Hence the Call Control EJB may be usedby any value-added developed by a MVNO and

Figure 4-1 A MVNO that usesOSA to develop value-addedservices

Figure 4-2 Implementation ofOSA applications

IN/SCP

HLR

SMSc

(Virtual)Mobile

NetworkOperatorDomain

MVNODomainValue-

addedService

CustomerDatabases

EnterpriseServices

OSAGateway


Customer

WebServices

OSA CCClient

Value-addedService

OSA CCClient

Value-addedService

OSAGateway

WebService

OSAGateway

Value-added

Service

Value-added

Service

EJB CCClient


also easily integrated with other services runningin the MVNO’s domain. An alternative to useEJB is to implement the OSA client as aCORBA object that also acts as a server to otherservices in the MVNO’s domain. For example,value-added services that wish to use Call Con-trol just act as a client to the Call control serverobject. Some advantages of using componentmodels like EJB or CORBA are that the OSAclients only need to be implemented once(reusable code), and the developers are able tocustomise the interface against the services run-ning in the MVNO’s domain. This will probablylead to faster creation of new value-added ser-vice and easier service integration.

5 ConclusionNowadays, development and deployment ofvalue-added services in telecommunication is acomplex and time-consuming task. Third partyservice providers and MVNOs have to collabo-rate with a network operator in order to providecustomers with value-added services and the ser-vices have to run inside the network operator’sdomain. By using 3GPP’s OSA, network opera-tors, 3rd party service providers and MVNOs isable to access network capabilities in a simpleway through standardised APIs. The applicationscan run on application servers inside the 3rdparty/MVNO’s domain, which will lead to betterservice integration using object oriented tech-niques like CORBA and EJB and more controlof the applications. Using OSA will make it pos-sible for network operators and third parties toprovide value-added services according to theconstantly changing demands of the customers,like e.g. personified services.

3GPP has only specified OSA to enable applica-tions to get access to network capabilities inwireless networks (GSM/UMTS). However, net-work operators may also operate a PSTN net-work and an IP-based network and may wantto provide OSA APIs also for those networks.Value-added services may hence be accessiblefrom heterogeneous networks. 3rd party/MVNOswill therefore be able to provide services for awider range of customers. It is necessary toinvestigate how it is feasible to provide OSA onheterogeneous networks such as PSTN and IP.For example, should it be only one CC SCF forthe whole system or should it be a separate CCSCF for each underlying network. Another issueis whether each network domain should have itsown OSA implementation and the third partieshave to deal with many different OSAs or thereshould be a unique OSA providing applicationswith SCF across network domain boundaries.How the different approaches will impact theapplications and how each approach is imple-mented on each heterogeneous network shouldbe studied.

References1 3GPP. 2000. Virtual Home Environment;

Open Service Architecture (Release 1999).Valbonne. (3G TS 23.127, V3.1.0, 2000-06.)

2 3GPP. 2000. Open Service Architecture,Application Programming Interface Part – 1(Release 1999). Valbonne. (3G TS 29.198,V3.0.0, 2000-06.)

3 OMG. CORBA Primer White paper. January28, 2002. [online] – URL: http://www.omg.org/gettingstarted/corbafaq.htm

4 ITU/ISO. 1997. ODP Trading Function –Part 1; Specification. (ISO/IEC IS 13235-1,ITU/T Draft Rec X950-1.)

5 Sun Microsystems. Enterprise JavaBeanstechnology. January 28, 2002 [online] –URL: http://www.java.sun.com/products/ejb/index.html

Telektronikk 4.2001

1 Definition of the Problem

1.1 Mobile NetworksIn this paper, we present a model by which theprice a mobile virtual network operator (MVNO)has to pay for services from a mobile networkoperator (MNO) can be determined. The modelis based on how investment risk should bedivided between the two types of operators inorder to avoid arbitrage. The model is based onthe theory of financial and real option pricing([1], [2], [3], [4]).

A mobile network is as shown in Figure 1. Thenetwork has the same basic structure for GSM,UMTS, and wireless LANs so that the discus-sion that follows is independent of the particulartype of network considered. See [5] and [6] fordetails concerning the structure of differentmobile networks.

A mobile network cooperates with other fixed ormobile networks in order to offer the followingservices to the users:

• Transport of bits between the communicatingparties;

• Routing of calls between communicating par-ties;

• Terminal mobility enabling the mobile termi-nal to move between different base stationsand updating its location;

• Continuity allowing continuity of conversa-tions when the mobile terminal movesbetween base stations (handover);

• Maintaining and managing the user profile,that is, the set of services and capabilities con-tained in the subscription;

• Roaming enabling mobile users to access net-works owned by different operators.

It is not important how these capabilities aredesigned in particular networks. What is impor-tant is to understand that these functions are sup-ported by all mobile networks. They representthe core functionality of mobile communica-tions.

As in all networks, the applications and servicesthe users are receiving exist outside of the net-works supporting mobility and transport of bits.This observation is important since servicesbeyond those listed above are not part of the net-works, fixed or mobile. However, the servicesand applications determine the characteristics ofthe traffic the network has to support. This sepa-ration between network functionality and ser-vices is particularly easy to see in the Internet.In the Internet, the functions listed above are allsupported by the IP protocol (and other proto-cols at the network layer). All applications andservices exist on top of TCP or UDP. TCP andUDP are end-to-end protocols the header andcontent of which are not visible to the network.

Option Pricing of Mobile Virtual NetworkOperatorsJ A N A A U D E S T A D A N D A L E X E I A G A I V O R O N S K I

Alexei A. Gaivoronski obtainedhis PhD from Moscow Inst. forPhysics and Technology in app-lied mathematics specializing inoperations research and opti-mization. He has since worked inacademy (Ukraine, Austria, Italy,Norway) and telecommunica-tions industry (Italy). He is Pro-fessor of Industrial Economicsand Operations Research at theDept of Industrial Economicsand Technology Management ofNorwegian University of Scienceand Technology. His researchinterests are optimization anddecision support systems underuncertainty ranging from theoryto industrial applications to soft-ware and he has publishednumerous papers on these sub-jects. His applied interests arefocused on telecommunicationsand [email protected]

Jan A Audestad (60) is SeniorAdvisor for Telenor CorporateUniversity. He is also AdjunctProfessor of telematics at theNorwegian University of Scienceand Technology (NTNU). Hehas a Master degree in theoreti-cal physics from NTNU in 1965.He joined Telenor in 1971 after 4years in the electronics industry.1971 – 1995 he did researchprimarily in satellite systems,mobile systems, intelligent net-works and information security.Since 1995 he has worked in thearea of business strategy. Hehas chaired a number of inter-national working groups andresearch projects standardisingand developing maritime satel-lite systems, GSM and intelligentnetworks.

[email protected]

Application or service

Mobile access

functionalityOther

networkMobileaccesspoint

Transport of bitsRoutingTerminal mobilityContinuityUser profileRoaming

Mobile networkoperator (MNO)

Figure 1 Mobile network operator (MNO)

106


1.2 Traffic, Cost and Revenue inTelecommunications

The traffic can be defined in terms of a fewparameters such as temporal variation of callintensity, mobility of users and call duration.We know from experience the value of the traf-fic characteristics of speech and we can predictthe performance of networks supporting onlyspeech with good accuracy. We also know thatthe traffic characteristics of the short messageservice (SMS) are completely different: shortercall duration and higher call intensity. The per-formance of pure SMS networks can also becomputed with sufficient accuracy. However, insystems with mixed traffic such as speech andSMS having different values of the traffic char-acteristics, it is much more difficult to assess thenetwork performance. WAP access to the Inter-net makes this task even worse.

The next important observation is that the majorcost of telecommunications networks is associ-ated with investments and operation, both ofwhich are approximately proportional to thepeak traffic the network can carry. Traditionally,the performance of networks is defined in termsof the peak traffic they can support. This iscalled the grade of service of the network. Moreprecisely, the grade of service specifies theamount of traffic, e.g. 2 %, that will be rejectedbecause the network cannot carry more traffic.Therefore as a first approximation, we may statethat the cost of the network is proportional to thegrade of service offered to the customers. If thegrade of service cannot be met because the traf-fic has increased or changed its characteristics,the network has to be increased, requiring moreinvestments.

The revenues of telecommunications are essen-tially proportional to the average number of cus-tomers and the average traffic they generate.We are thus facing a situation where costs areproportional to the peakedness of the traffic

while the revenues are proportional to the aver-age traffic generated by the customers. Thepeakedness corresponding to the stated grade ofservice can be determined, with reasonable accu-racy, from the variance of the statistical distribu-tion of the traffic. In certain cases also higherorder moments are important such as skewnessand kurtosis. However, this is not important forthis discussion. Observe the important fact thatthere is no relationship between the average andvariance of a statistical distribution (except inthe simple case of a one-parameter distributionsuch as a Poisson distribution).

Therefore, we conclude that there is no mathe-matical relationship between cost and revenue intelecommunications since each is derived fromindependent statistical parameters. This meansthat if the statistical distribution of the trafficchanges, the ratio between revenue and cost alsochanges. Let us illustrate this behaviour by twoexamples. The average traffic may increasewhile the variance is constant (or increases lessrapidly). In this case, the revenue increases whilethe cost is constant (or is increasing at a slowerrate). Telephony is an example of such behav-iour where cost is increasing slower than theincome allowing the operator to reduce pricesover time. On the other hand, when we mix traf-fic with different characteristics such as tele-phony, SMS and Internet, the average trafficmay be unchanged while the variance may in-crease requiring more investment in networks.In this case, the price must sometimes be in-creased in order to retain the profit.

This behaviour is often different from produc-tion of goods where the revenue per good is pro-portional to the cost of producing it. This is thesituation when a major part of the cost of pro-ducing the good is associated with each itemproduced (raw material, workforce, handling)and only a small part of the cost is common forthe whole production (administration, sales).

Figure 2 Roaming

MVNO customers

MNO1 customers

MNO2 customers MNO2 network

MNO1 network

Roaming 2 to 1

Roaming 1 to 2

Roaming MVNO to 2 Roaming MVNO to 1


1.3 Virtual Mobile Network OperatorThe MVNO does not own its own mobile net-work but offers roaming and maintenance of theuser profile. All other functions are bought fromordinary mobile network operators (MNOs) butthe customers of the MVNO will view the totaloffer as coming from the MVNO.

The customers of the MVNO can implementtheir own services and applications on top ofthe capabilities offered by the MVNO.

The motives of the MVNO are several: to avoidbinding investments capital, to offer ordinarymobile communications in competition withMNOs, and to offer something entirely differentsuch as worldwide mobile business communica-tions. The motives may not be publicised so thatthe MNOs (and other MVNOs) do not compre-hend the underlying business plan of the MVNO.We will come back to this when looking at thetypes of market games the MVNO is playing.

1.4 RoamingOne of the major problems associated withMVNOs is associated with roaming. Note thatwe are using the term roaming in the precisemeaning that it allows customers of one mobileoperator (MNO or MVNO) to access the net-work of other operators (MNOs). Note that itis only meaningful to access the network of anMNO since roaming is associated with the act ofconnecting the terminal to the physical resourcesof the network. Roaming is shown in Figure 2.

Figure 2 shows how roaming takes placebetween MNOs and MVNOs. For simplicity,only two MNOs and one MVNO are shown.Roaming between MNOs is simple because itis symmetric in the sense that some customersof MNO1 roams to the network of MNO2, andsome customers of MNO2 roams to the networkof MNO1. If the two streams are equal, nochange takes place with regard to the number of

customers accessing each network. If the streamsare not equal, e.g., because of tourism, the num-ber of customers may increase in one networkand decrease in the other network. The situationis nevertheless well understood and it is simpleto settle business disputes by international roam-ing agreements. What is even more important isthat MNO1 and MNO2 cover different geo-graphical regions so that the networks do notcompete for the same customers.

The situation is different for the MVNO. Sincethe MVNO does not own its own network, nocustomers can roam to that operator. However,all the customers of the MVNO have roamed tothe networks of the MNOs. This makes the situ-ation asymmetric, and, what is worse, it is notcovered by the international agreements guaran-teeing MVNO customers the right to accessother networks. The MVNO and MNO are alsocompeting for the same customers.

Below we shall look at the consequences of thisasymmetry and suggest a method to handle it ona fair basis for all parties involved.

2 Roaming andInvestment Risk

Figure 3 shows how investments are done in amobile network. Sometimes the investments aresmall. This is so if the traffic increase can be sat-isfied by adding more channels to the base sta-tions. Sometimes the investments are large. Thisis so if the cell configuration of the system mustbe altered.

The rational behind this evolution is the follow-ing. The frequency spectrum allocated to amobile system is limited and capable of support-ing a certain amount of traffic. This resource isshared between all operators covering the samegeographic area. The amount of base stationequipment required is proportional to the trafficsupported by the base station. A reasonable strat-egy is then to invest in base station equipmentjust enough to meet the traffic demand. Theinvestment of adding more equipment to thebase station is small.

At some stage, the whole frequency spectrumavailable to the operator in a given area has beenused up. The operator can then increase thecapacity by building a new pattern of radio cellsin the area by making each cell smaller. In Fig-ure 4, the number of cells covering a given arehas been increased from 4 to 20, that is by a fac-tor of five. This means that the new cell arrange-ment can serve approximately five times asmuch traffic as the original cell pattern. How-ever, this increase in traffic requires five timesas many base stations, thus requiring muchinvestment.

Figure 3 Investment asfunction of time

Traffic

Investments

Time

Tra

ffic

Cum

ulat

ive

inve

stm

ent


Such development of the cell pattern can bedone in two ways as shown in Figure 5 (note thatthe ordinate is logarithmic). The base stationantenna may be located such that there is freesight between that antenna and the antenna ofthe mobile terminal. The received power thendrops off first at a rate of approximately d–2,where d is the distance between the base stationand the mobile terminal, and then at a rate of d–4

because of interference with the wave projectedalong the surface of the earth. The base stationmay also be hidden behind buildings and otherobstacles such that there is no free sight betweenthe base station antenna and the antenna of themobile terminal. Because of diffraction, some ofthe radio power is propagated into the shadow ofthe obstacle. The received power then drops atthe rate d–4. In general this makes it possible toconstruct larger cells in rural areas than in cities.

Small cells can then be designed by hiding theantenna behind buildings and other obstacleswhich, in practice, means that the antenna isplaced close to street level. This is an efficientmethod of reducing the cell size in cities. An-other method is simply to reduce the power radi-ated by the base station. This does not alter theshape of the propagation curve as shown in thefigure.

The problem the mobile network operator faces isthat the capacity of the network must be increasedfrom time to time if the service demand isincreasing or the variance of the traffic distribu-tion becomes larger. This gives rise to a stochastictime series of investments as shown in Figure 3.The size of the investment depends upon whetherthe demand can be satisfied by a small investmentor by a large one as explained above.

The investment must be done by the operatorowning the network (the MNO). The require-ment for investment is determined by all opera-tors creating traffic in the network. This may bethe operator owning the network, operators own-ing their own network but creating roaming traf-fic, and operators not owning a network but buy-ing traffic capacity for serving their own cus-tomers (MVNOs). In the first case, the invest-ment risk is carried by the operator owning thenetwork. In the second case, roaming is mutualand it is fair to assume that roaming betweennetworks is approximately equal in the twodirections so that the risk carried by each opera-tor is the same. Even if the roaming traffic is notsymmetric, this type of risk sharing is reasonableand can be settled by international roamingagreements. In the third case, one operator car-ries all the investment risk while the MVNO car-ries no risk. This situation may be balanced byincluding the investment risk in the price ofusing the network.

Note that the investment required to support theMVNO is nondecreasing over time, that is, theinvestment may be zero but it is never decreas-ing (we assume that the overall telecommunica-tions market is not decreasing).

3 Option PricingThe idea is to determine the price the MVNOhas to pay for using the network such that theeconomic risk is distributed on both the MVNOsand the MNOs in a fair manner. The MVNO willthen pay a premium for increasing the risk thatthe MNO has to invest in more infrastructure.The premium must then ideally be such that therisk is divided equally between the MNO and theMVNO. This is similar to option pricing in thefinancial market.

The price the MVNO has to pay consists of twocomponents: the price, p, for using the networkassuming that the average traffic and the vari-ance is constant, and a premium ∆p determinedfrom the expected probability of investing inmore infrastructure. The total price is thenp + ∆p. Note that p + ∆p can be the price peraccess, a fixed price for a certain period of time(a month or a year), or a combination of these.How the price p is determined is subject to nego-tiations and has no impact on the determinationof ∆p. What is important is that p and ∆p aredetermined independently of each other.

∆p is a function of the expected traffic increasecaused by the MVNO, the expected change invariance, the ratio between traffic caused by theMVNO and the MNO, and the geographic distri-bution of the traffic (i.e. where investments are

Figure 4 Cell patternscovering the same area

Old pattern

New pattern


required). The third variable reflects the pointthat the MVNO shall not pay for the investmentrisk caused by the MNO alone.

The MVNO may have similar agreements withseveral MNOs depending upon how the MVNOis designed to operate.

The determination of ∆p is not trivial. Themethod can be based on the same ideas as deter-mining the premium of financial options. Thedynamics of the market may also be captured asfollows. The price increment ∆p is settled for a

certain period, say one year. Thereafter the valueof ∆p is renegotiated.

Let us consider two examples:

Example 1. Assume that the traffic intensity gen-erated by the MNO is constant while that gener-ated by the MVNO is on average λ0 during thenegotiation period T (see Figure 6). The addi-tional traffic generated during the period is thenλ0T. The revenue generated by the MVNO dur-ing he time T is (p + ∆p)λ0T, where the last termshould compensate for the new investment ∆I.This gives:

∆p = ∆I / λ0T.

Note that ∆p = 0 if ∆I = 0 as it should be: noinvestment, no premium. Figure 6 shows how∆p is calculated over several periods.

Example 2. In this example, the traffic of boththe MNO and the MVNO increases. Assume thatthe increase of the traffic alone of the MNOrequires new investment at time T1 while theadditional traffic generated by the MVNOrequires that the investment is made at an earliertime T2. Then we require that the discounted dif-ference of the value of the investment betweenthese instances is paid by the MVNO. With dis-count rate r this gives (note that T1 > T2):

∆p = ∆I(e–rT2 – e–rT1) / λ0T

where λ0 is the average traffic rate generated bythe MVNO in the contract period T.

In real cases, we should determine ∆I and ∆psuch that the economic risks are approximatelythe same for the MNO and the MVNO. TheMNO should than save as much on averagebecause the traffic does not increase so muchthat investments are necessary that the MVNOsaves because the traffic increases so much thatmore investments than anticipated are required.This is the difficult task.

Another approach to quantitative modelling ofcompetition and cooperations between networkoperators in the presence of uncertainty is con-sidered in [7]. This approach is based on sto-chastic programming enhanced with selectedideas from games theory.

4 Some Problems Related toGames

The examples above assume that we can predictfuture investments accurately. This is, of course,not the case. We are instead faced with a numberof problems. ∆I must be estimated based on his-torical data, simulations, or guesses concerningthe future development of the market. The inten-

Figure 6 Two periods withaverage incremental traffic λ0

Figure 5 Cell radii andpropagation

Reduced radiated power level

Free space propagation

Hidden antenna

Receiver threshold

d-2d-4

Cell radii

Distance fromantenna (d)

Rad

iate

d ra

dio

pow

erIn

crem

enta

l inv

estm

ent

Tra

ffic

Time

Determines p

Determines ∆p

2λ0

0

4λ0

T T

∆I1 =∆p1λ0 T

∆I2 =∆p2λ0 T


tions and business plan of the MVNO are alsoimportant. ∆I is also a stochastic variable so notonly the average probability but also the vari-ance and perhaps higher moments must be esti-mated.

Questions that are likely to arise are. Are theMNO and the MVNO competing for the sametraffic, or in other words, is this a zero-sumgame? If so, all traffic gained by the MVNOis lost by the MNO. Is the business plan of theMVNO such that most of the traffic is new traf-fic only generated by the presence of the MVNO?If so, are the operators facing a cooperative non-zero-sum game? Is the game something in be-tween? If the game is cooperative, that is, induc-ing traffic that otherwise will not be present,should then more of the risk be allocated to theMNO?

How can we trust the traffic estimates providedby the MVNO? Has the MVNO deliberatelyunderestimated the traffic in order to reduce thepremium? Should there be a penalty premium ifsuch cheating can be proved? And how shouldthis premium be determined, and who should actas arbiter in such conflicts? Has the MVNOoverestimated its traffic increase allowing theMNO to request a higher price ∆p than required?

The uncertainty is similar from the viewpoint ofthe MVNO. How is the investment risk calcu-lated? Is the MNO cheating claiming higherinvestment probability than actually necessary?Is the MNO deliberately underestimating its owntraffic increase? Is the MNO overestimating itstraffic increase so that the risk of the MVNO isreduced?

How often should the contract be renegotiated,once a year or more often? What are the criteriathat one of the parties can request renegotiatingthe contract before it expires?

None of the above questions is simple. And theyare likely to have different answers for differentMNOs and MVNOs.

5 Similar CasesThis paper only studies the case of the relation-ship between MNOs and MVNOs. However,there are several similar cases of cooperationleading to similar solutions. It is likely that thesame principles applies to all virtual networkoperators (VNOs) and not only mobile virtualnetwork operators. It is harder to define the busi-ness of VNOs in general. A special problem isrelated to the fact that IP and TCP puts a definitedistinction between services involving the net-work and only the hosts (terminals). Therefore,a service offered on top of TCP must be treatedin a different way than a service offering IP.

Examples of the former are transaction services,web services, e-mail services and payment ser-vices. Examples of the latter are dark fibre andmobile access.

The models for cooperation between networkoperators and operators only offering servicesabove IP and services involving IP are different.Operators of the latter type are similar to theMVNOs discussed above and the problems thatthe VNOs and the operators owning the network(NOs) are facing are the same as those describedabove.

The problems facing operators offering servicesabove TCP (let us call them TOs for simplicity)and NOs (mobile or fixed) are different. In thiscase, there are at least to components that mustbe studied in order to understand the relationshipbetween them. These are the costs the operationof the TOs are causing in the NOs. Example ofsuch costs are provision of access and transportof bits. On the other hand, the traffic the TOgenerates may not be generated by any othermeans. In such cases it is fair that the NO paysthe TO for providing this traffic since it is a jointeffort to create value in the network.

Note that the problems we have been addressingabove are new. The MVNO concept has existedfor not more than five years we have had ser-vices such as those generated by TOs for lessthan ten years. Therefore, the aspects of coopera-tion between virtual operators and operatorsowning the network infrastructure is by nomeans sufficiently understood.

References1 Dixit, A K, Pindyck, R S. Investment under

Uncertainty. Princeton University Press, 1994.

2 Trigeorgis, L. Real Options: ManagerialFlexibility and Strategy in Resource Alloca-tion. The MIT Press, 1996.

3 Hull, J C. Options, Futures and Other Der-ivatives. Prentice Hall (4th edition), 2000.

4 Voit, J. The Statistical Mechanics of Finan-cial Markets. Springer, 2001.

5 Schiller, J. Mobile Communications. Addi-son-Wesley, 2000.

6 Walker, J (ed.). Advances in Mobile Infor-mation Systems. Artech House, 1998.

7 Audestad, J A, Gaivoronski, A A, Werner,A. Modelling Market Uncertainty and Com-petition in Telecommunication: NetworkProviders and Virtual Operators. Telektron-ikk, 97 (4), 2001, 46–64 (this issue).

113113

Status

Telektronikk 4.2001

115

As the Internet Protocol (IP) suite has becomethe single technology with probably the highestmomentum in today’s telecom world, it is timeto look into how the Internet technology is stan-dardised. Up till a few years ago, the telecomand computer worlds were separate. The Internetwas used basically to connect computers in net-works. Now, IP is evolving to serve also typicaltelecommunications needs, like voice and video,providing reliability and security. There is stillsome way to go before IP can provide the neces-sary level of quality of the previously mentionedaspects, but much effort is put into it.

In this issue of Telektronikk’s Status section, wefocus on the standardisation work of the InternetEngineering Task Force (IETF) by two contribu-tions from highly skilled participants in the pro-cess.

The first paper is given by Zaw-Sing Su andgives an overview of how the work is carriedthrough in IETF and how the standardisationprocess works. One thing worth noting is thatIETF is almost purely driven by personal mem-bers and the organisational context is less impor-tant. The paper also gives some advice on howEuropean telecom operators should positionthemselves towards IETF and which strategiesthey should use to influence the Internet stan-dardisation process.

The second paper is given by Paal Engelstad andgives an overview of the background for IPv6,based on the fact that different ‘dialects’ and ad-hoc solutions made the current IPv4 less effi-cient and transparent. It concludes with threesolutions for the migration towards IPv6: ashort-term or ad-hoc solution, and two differentlong-term models.

IntroductionP E R H J A L M A R L E H N E

Per Hjalmar Lehne (43) obtainedhis MSc from the NorwegianUniversity of Science and Tech-nology in 1988. He has sincebeen with Telenor R&D workingwith different aspects of terres-trial mobile communications.1988 – 1991 he was involved instandardisation of the ERMESpaging system in ETSI as wellas in studies and measurementson EMC. His work since 1993has been in the area of radiopropagation and access tech-nology, especially on smartantennas for GSM and UMTS.He has participated in the RACE2 Mobile Broadband Project(MBS), COST 231, and COST259 and is from 2000 vice chair-man of COST 273. His currentinterests are in 4th generationmobile systems and the use ofMIMO technology in terrestrialmobile networks, where he par-ticipates in the IST projectFLOWS.

[email protected]

Telektronikk 4.2001

Telektronikk 4.2001

Internet Engineering Task Force, an outgrowthof a U.S. Government sponsored project, is astandards body somewhat different from others.In this article, we go beyond the surface to allowsome appreciation of its inner workings andbriefly recommend a way forward for its engage-ment.

1 Historical PerspectiveOpening the Internet for commercial use in thebeginning of ‘90s together with the introductionof World-Wide Web catapulted Internet technol-ogy into focus for worldwide communications.Internet technology is based on packet switch-ing, which was developed to meet robustnessand stringent bandwidth requirements of militarycommunications. Information communicatedover the network is segmented into and transmit-ted in packets. Packets are propagated alongavailable paths toward their destinations, cir-cumventing network failures to achieve robust-ness and filling available slots to make efficientuse of communication bandwidth.

The basis of Internet technology is the InternetProtocol (IP), developed for interconnecting net-works of different transmission media. It thusachieves media independence and unifies com-munications across different media. Variousapplications and services can thus be developedbased on IP, independent of underlying trans-mission media.

Prior to commercialization, Internet was a pro-ject developed in the ‘70s and ‘80s under theauspices of Defense Advanced Research ProjectsAgency of the U.S. Department of Defense, andlater U.S. National Science Foundation. Duringthat period, there were a number of research anddevelopment projects around the world pursuingpacket based network communication. Not theleast was the effort by major vendors and tele-communication operators under the International

Organization for Standardization. Well-knownresults of that effort include the seven-layer Ref-erence Model for Open System Interconnectionand the X series, including X.25, protocols.

During sponsorship of the U.S. Government, theInternet was operated under its authority. TheGovernment contracted the execution of its oper-ations to Information Science Institute of Uni-versity of Southern California, a non-profit orga-nization, and Bolt, Baranek and Newman, aCambridge-based technical consulting firm. Itpurchased hardware from computer vendors,leased point-to-point communication bandwidthfrom telecommunication carriers, and contractedresearch and development organizations for net-working and software design and development.

Established in 1985, Internet Engineering TaskForce (IETF) was the organization for coordinat-ing technical development among participatingcontractors as well as user organizations, such asthe military, other government agencies, aca-demic institutions, as well as research and devel-opment organizations. When Internet technologystarted to mature, vendors specialized in build-ing Internet equipment, such as Cisco, Proteon,3com and Sun Microsystems, started to appear.IETF, in the mean time, became dominated byequipment vendors while operations of the Inter-net remained under government authority. Theuse of Internet was limited to and free of chargewithin government sponsorship.

2 OrganizationUpon worldwide adoption of the Internet, theneed for standardization of its technologybecame obvious. In the mean time, the U.S.Government terminated its sponsorship. Thiscombination of events enabled the commercial-ization of Internet operations and transferred themanagement of continuing development andevolution of Internet technology to a non-profit

IETF and Internet StandardsZ A W - S I N G S U

Zaw-Sing Su is General Man-ager of Telenor R&D (California)in Palo Alto. He began collegeeducation in Chiao-Tong U.,Shanghai, China and completedhis BASc degree in electricalengineering in U. of Ottawa,Canada. He completed his MAin applied mathematics in the U.of Michigan and PhD in com-puter science in U. of Toronto.He worked for Bell-NorthernResearch in Canada and ITTTTC in Connecticut before join-ing UCLA to manage the Mea-surement and Analysis Group ofDARPA Packet Radio Project.Su was transferred to StanfordResearch Institute in 1980 andstarted working on DARPA Inter-net Project. In addition to gen-eral participation in early Inter-net design and development, heinitiated the design of InternetNaming Convention and NameServer System which providedthe foundation for Internet DNS.In 1985 he combined packetRadio and Internet technologiesto pioneer Internet mobility.Before joining Telenor R&D in1998, he ventured to participatein Silicon Valley startups.

[email protected]

IAB(Internet

Architecture Board)

IETF(Internet Engineering

Task Force)

IRTF(Internet Research

Task Force)

Internet Society(Parliament)

116


public entity, the Internet Society. However,organizations and the organizational structureformed under the government remained. Theseorganizations include the Internet ArchitectureBoard, under which The Internet EngineeringTask Force, the Internet Research Task Force,and other related entities.

2.1 Internet Society (ISOC)The Internet Society is a non-profit internationalmembership organization of Internet experts thatcomments on its policies and practices. InternetSociety may be considered the “parliament” ofInternet community. It oversees a number ofboards and task forces dealing with network pol-icy issues and enjoys the support of more than150 organization and 6,000 individual membersin over 170 nations worldwide. Work of theInternet Society focuses on standards, publicpolicies, as well as education and training infor-mation technology leaders around the world.

2.2 Internet Architecture Board (IAB)The IAB is responsible for defining the overallarchitecture of the Internet and providing guid-ance and broad direction to the IETF. It alsoserves as the technology advisory group to theBoard of Trustees and Officers of Internet Soci-ety. It acts as a source of advice and guidance tothe Board of Trustees and Officers of the Inter-net Society concerning technical, architectural,procedural, and (where appropriate) policy mat-ters pertaining to the Internet and its enablingtechnologies.

The IAB oversees a number of critical activitiesin support of the Internet. It appoints the IETFchair and all other IESG (Internet EngineeringSteering Group – see below) candidates from alist provided by a nominating committee. It pro-vides oversight of the process used to createInternet Standards and also serves as an appealboard for complaints of improper execution ofthe standards process.

IAB members are also nominated by the nomi-nating committee, and reviewed and appointedby the Internet Society Board of Trustees. Theyserve two-year terms. Thus, each spring abouthalf of its members need be reviewed.

The nominating committee for filling IAB andIESG vacant positions consists of ten votingmembers: a non-voting chair, and at least threenon-voting liaison members – one liaison fromeach of sitting IAB and IESG, and chair of theprevious year assumes the other. The votingmembers are randomly selected among volun-teers from the Internet community. The non-vot-ing chair is appointed by the Internet SocietyPresident. [1]

2.3 Internet Engineering Task Forceand its Steering Group

The IETF was originally designated the protocolengineering and development arm of the Internetorganization. When the need for standardizingits technology became apparent, the responsibil-ity of Internet standards fell onto it. As a stan-dards body for Internet, a communication tech-nology, it has not, at least not yet, come underthe jurisdiction of International Telecommunica-tions Union (ITU). But, a memorandum ofunderstanding (MoU) was signed between ITUand IETF during the Oslo IETF meeting in 1999.

To carry out its mission, IETF subdivides itsefforts into Work Areas (Areas) and WorkingGroups. Each Area may consist of a number ofWorking Groups under the management of oneor two Area Directors. The Area Directors sit-ting as a body, along with the IETF Chair, com-prise the Internet Engineering Steering Group(IESG). The IETF Executive Director is an ex-officio participant of the IESG, as are the IABChair and a designated IAB liaison.

The IETF meets at least once a year aroundMarch. In practice, IETF meets traditionallythree times each year, twice in the U.S. andanother time elsewhere. More discussion on theWorking Groups and Internet standards processfollow in the next section.

The IESG is responsible for technical manage-ment of IETF activities and the Internet stan-dards process. It administers the standards pro-cess according to the rules and procedures rati-fied by the ISOC Trustees. The IESG is directlyresponsible for the actions associated with entryinto and movement along the Internet “standardstrack”. Although most of the Internet standard-ization processes are carried out within IETF,the IESG is responsible for the approval of spec-ifications as Internet Standards.

2.4 Internet Research Task Force and its Steering Group

The Internet Research Task Force (IRTF) iscomposed of a number of small ResearchGroups. Research Groups are usually focusedand long term, though short-lived “task forces”are possible. They work on topics related toInternet protocols, applications, architecture andtechnology. Research Groups are expected tohave stable long-term membership (with respectto the lifetime of the Research Group) needed topromote the development of research collabora-tion and teamwork in exploring research issues.Participation is by individual contributors, ratherthan organizational representatives.

The IRTF is managed by the IRTF Chair, in con-sultation with the Internet Research Steering


Group (IRSG). The IRSG membership includesthe IRTF Chair, the chairs of the various Re-search Groups and possibly other individuals(“members at large”) from the research commu-nity. The IRTF Chair is appointed by the Inter-net Architecture Board (IAB). The ResearchGroup chairs are appointed in the formation pro-cess of the Research Groups, and the IRSGmembers at large are chosen by the IRTF Chairin consultation with the rest of the IRSG and onapproval of the IAB. In addition to managing theResearch Groups, the IRSG may from time totime hold topical workshops focusing on re-search areas of importance to the evolution ofthe Internet, or more general workshops to, forexample, discuss research priorities.

The guidelines and procedures for the operationsof IRTF Research Groups are described morefully in [2].

3 IETF Working Group and theStandards Process [3, 4, 5]

As mentioned, IETF activities are organized intoWorking Groups. Each Working Group belongsto an IETF Work Area. Currently, there are 8Areas in IETF: applications, Internet, operationsand management, routing, security, sub-IP,transport services, and user services. While theymay come and go as needed, Work Areas arerelatively stable. The number of WorkingGroups in an Area varies. In a “busy” Area,there can be as many as 20. In recent years, thereare close to 100 active Working Groups eachtime IETF meets.

3.1 Working Group OperationsIn Working Groups, protocols, mechanisms andother networking issues are specified, discussedand matured. One or two chairpersons may chaira Working Group. Goals and milestones aredefined for a focused effort to address a specificproblem or to produce specific deliveries. Wher-ever necessary or appropriate, an Internet proto-col or technology would be standardized. Thestandards process, as a part of Internet technol-ogy development, also takes place in the Work-ing Group.

A Working Group is expected to be short-lived,at least short-lived in nature. Work is carried outthrough mailing list discussion and face-to-facemeetings.

There is no formal membership in IETF. Partici-pation is open to all interested parties, by indi-viduals rather than by formal representatives oforganizations. Participation has swelled hundredfold from initially around 20 participants to inexcess of 2,000 in recent meetings.

3.2 Internet DocumentsInternet documents include Internet Drafts (IDs)and Request for Comments (RFCs). Internetdrafts document work in progress, and enjoy nostatus of any kind. They are not archived and aredeleted after six months from the time of sub-mission or last revision. Internet drafts areannounced and disseminated by IETF. Thosepromoted to become Working Group documentsare prefixed with “draft-ietf-(working groupacronym)”.

Request for Comments are of an archival docu-ment series of IETF. Not all RFCs are standardstrack. Besides standards track RFCs, an RFCmay be of “Best Current Practice”, “Informa-tional”, “Experimental” or “Historic”. They areedited, announced and disseminated by the RFCEditor. For a standards track RFC, it can be a“Proposed Standard”, a “Draft Standard”, or a“Standard”.

3.3 Working Group FormationWorking Groups are typically created to addressa specific problem or to produce one or morespecific deliverables, such as a guideline or aspecification for a standard. A Working Groupmay be established at the initiative of an AreaDirector or initiated by an individual or groupof individuals.

Often it is not clear whether an issue merits theformation of a Working Group. A Birds of aFeather (BOF) session at an IETF meetingtogether with an email list for preliminary dis-cussion would allow an assessment of interestsand pinpoint the technical issues. A WorkingGroup may result if there is enough interest andfocus in the subject, and with the approval of theArea Director for the Area in which the WorkingGroup would fall.

A Working Group may have one or two chair-persons. It must have a clear and focused char-ter, established goals and milestones. The charterof a Working Group is like a contract betweenthe IETF and the Working Group, which is com-mitting to meet explicit milestones and deliverspecific deliverables. The charter may be rene-gotiated periodically to reflect the current status,organization, or goals of the Working Group.

Upon completion of its goals and achievementof its objectives, the Working Group is to be ter-minated. When appropriate, a Working Groupmay be extended with a modified charter. Suchan extension must be with the concurrence ofits chair, its participants, the responsible AreaDirector and thus indirectly the IESG.


3.4 Standards ProcessInitiation of a standard is submitted as an Inter-net Draft to a Working Group. In line of achiev-ing its objective, the Working Group may acceptand promote an Internet Draft to become a“Working Group document” to be activelyworked upon. The Working Group, or an indi-vidual, may recommend to the Area Director a“standards action” to promote an Internet draftalong the standards track. The Area Directorrelays the recommendation to the Internet Engi-neering Steering Group for review. IESG thenapproves a standards action. If necessary, it maycommission an independent technical review ofthe recommendation. Upon approval if withmodification, the document is sent back to theWorking Group for “Last Call”, a final reviewby the Internet community. The Last Call mustlast at least two weeks. To advance along thestandards track, an Internet Draft becomes aRequest for Comment (RFC), a historical namefor the archived Internet document series. Itshould be noted that not all RFCs are standardstrack, but they all share the same process toadvance from Internet Drafts.

A standards-track RFC must remain a ProposedStandard for at least six months, and a DraftStandard for 4 months. The Internet standardsprocess calls for no formal voting, but an app-roval process with rough consensus. Disputesare resolved by discussion and system demon-stration. Listed goals of the Internet standardsprocess include technical excellence; priorimplementations and testing; clear, concise, andeasy understood documentation; openness, fair-ness, and timeliness.

4 Operator StrategyThe exponential traffic growth in early ‘90s cata-pulted Internet technology into focus for world-wide communications, and called for its stan-dardization. Exploration also began in earnest toemploy Internet technology to offer a unifiedinfrastructure for both “best effort” as well asreal-time communications. It imposes a chal-lenge to traditional telecommunication operatorswith changing business models, uncertain mar-ket place, and to keep up with the continuingevolution of Internet technology.

In recent years, there is increasing participationof European operators in Internet community.The high threshold for participation calls forcareful evaluation and strategic engagement.Such a strategy may call for:

• Focused effort: Given the growing complexityof Internet technology, efforts of engagementcan ill afford to spread too thin. They shouldbe necessarily focused to closely follow busi-ness strategies. A business offering can bethought as, for example, of infrastructure, ser-vice platform, or applications and services.Where infrastructure offerings are for bearerbandwidth with possible improvements suchas quality of service or virtual private networkwith security considerations. Service platformofferings are to third party providers who maybase on such offerings to build services andapplications; while the operator itself maydirectly offer user services and applications.

• Leveraged engagement: On selected subjectareas with defined objectives, alliance andcollaboration to leverage efforts may be pur-sued. With an appropriate vendor one may, forexample, tackle service platform technologies.Pre-competition collaboration with other oper-ators may explore multi-domain issues onend-to-end communication.

• Community participation: Participation inIETF emphasizes on personal rather thanorganizational involvement. Professionalinvolvement by individuals not only may playa key role in IETF participation, but may alsooffer the best possibility of penetration intoInternet community. Consistent and focusedparticipation in IETF by selected technicalstaff is recommended.

References

1 Gavin, J. 1998. IAB and IESG Selection,Confirmation, and Recall Process: Opera-tion of the Nominating and Recall Commit-tees. (BCP 10, RFC 2282.)

2 Weinrib, A, Postel, J. 1996. IRTF ResearchGroup Guidelines and Procedures. (BCP 8,RFC 2014.)

3 Bradner, S. 1998. IETF Working GroupGuidelines and Procedures. (BCP 25, RFC2418.)

4 Bradner, S (ed.). 1996. The Internet Stan-dards Process – Revision 3. (BCP 9, RFC2026.)

5 Hovey, R, Bradner, S. 1996. The Organiza-tions involved in the IETF Standards Pro-cess. (BCP 11, RFC 2028.)

Telektronikk 4.2001

IntroductionA number of prominent researchers in the Inter-net Engineering Task Force (IETF) have latelywarned that the Internet architecture is about tochange. This creates new complications andproblems, which means that the Internet may notbe able to accommodate new types of communi-cations in the future Internet – an internet wherecomputer- and tele-communication have con-verged and where security, mobility and qualityof service can be taken for granted. This docu-ment uses the time glass model to describe thechanges in the Internet architecture, and showshow the new problems are being addressed inIETF.

IP Described by the TimeGlass ModelTo fully understand how the Internet architectureis about to change, it is necessary to take a closerlook at how IP was originally designed. Thisarticle uses the time glass model for this purpose(Figure 1). The model illustrates that all higherlayer protocols are running over the same Inter-net Protocol (IP), while the IP protocol may runover a number of underlying technologies. It wasan important choice of design to only allow oneInternet protocol. This choice facilitates maxi-mal interoperability between networks and aminimal the number of service interfaces, whichmakes implementations easier.

The width of the time glass illustrates the num-ber of protocols as well as the amount of func-tionality that is present at each layer. The ‘nar-row’ IP protocol incorporates limited functional-ity at the IP layer, while there are much morefunctionality at higher and lower layers. IP wasdesigned as simple as possible, and the function-ality is pushed out to the hosts on the edge of theInternet. This is called the end-to-end principle,which means that all higher layer protocols areimplemented in the hosts. A ‘narrow’ Internetprotocol includes a maximal number of networksby posing a minimal number of requirements tounderlying technologies. The result of pushingmost of the networking logic to the hosts is asimple, functional, efficient, generic, manage-able and scalable Internet.

This was the IP architecture, as we knew it fiveyears back in time. However, this architecture isabout to change.

The Time Glass Model is nolonger ValidIP technology has experienced an enormousgrowth over the past decade in terms of numberof users and in terms of expectations to what IPtechnology should accommodate. It seems thatthe original IP protocol was designed into sucha ‘narrow’ protocol that the time glass finallybroke. Figure 2 illustrates the situation as oftoday.

Towards a new Internet Architecture?– Or how IPv6 Addresses the FutureP A A L E N G E L S T A D

Paal Engelstad (31) is currentlyworking in Telenor’s office in Sili-con Valley, focusing on stan-dard ization work undertaken inIETF. He started work in TelenorR&D August 2000. Paal Engel-stad wrote his Master Thesis inKyoto University (Japan), re-ceived his MSc degree inphysics from NTNU and a bach-elor degree in Computer Sci-ence from the Univ. of Oslo.Apart from working with scienceand technology, Engelstad hasfour years work experience as amanagement trainee in Orklaasa, where he worked as aproduct manager (marketing),technical advisor (engineering),and project manager (businessdevelopment).

[email protected]

CSMA async sonet...

email WWW phone...

SMTP HTTP RTP...

TCP UDP…

IP

ethernet PPP…

copper fiber radio...

CSMA async sonet...

email WWW phone...

SMTP HTTP RTP...

TCP UDP…

IP

ethernet PPP…


Figure 1 The time glass model Figure 2 Something went wrong ...

(Figures from presentation by Steve Deering at 51st IETF)

120


What we are currently trying to do is to “gluetogether the fragmentation of the time glass”by means of different ad hoc solutions, whichunfortunately are not perfect. Therefore, someapplications and higher layer protocols that aredesigned based on the time glass model willcease to function properly. Furthermore, theyhave no opportunity to determine which ad hocsolutions are currently effective at the IP layer,and the network behavior seems random andnon-deterministic.

There is a number of ad hoc solutions that havegained popularity due to missing functionality ofIP. Firewalls, for example, compensate for thelack of security features in IP. Some applicationsand higher layer protocols such as basic MobileIP still cease to work, because firewalls are notpart of the original time glass model.

More important are Network Address Transla-tors (NAT). NATs are increasingly gaining pop-ularity due to the lack of globally routable IPv4addresses. A NAT replaces a private IP addresswith a global address in a packet that is leavingan intranet and entering the Internet. It alsomakes sure that a reply packet destined for theglobal address is forwarded to the original pri-vate IP address. When the communication ses-sion is terminated, the global address can bereused by new sessions and new hosts on theintranet. This ad hoc solution saves globallyroutable address space, because most of thecommunication happens within the intranetwhere the private IP addresses are valid. It isonly the fraction of traffic entering the Internetthat passes through the NAT.

Functionality we are About to LoseThe lack of functionality of IPv4 (e.g. lack ofaddress space) has lead to an enormous successfor different ad hoc solutions in the marketplace.This fact has a major impact on the networkingarchitecture of the Internet, and the simplicityof the time glass model is about to vanish. As aresult, the following functionality of the Internetarchitecture is about to be lost:

• Network transparency for protocols andapplicationsBefore, all applications could transparentlyuse IP. This is now getting dependent on thenetwork configuration, e.g. it there is a NATor a firewall present. The application requiresthat the ad hoc solution is adapted to the appli-cations to be run, e.g. that there is an Applica-tion Layer Gateway (ALG) implemented inthe NAT. The network is no longer transpar-ent for applications and higher layer protocols.

• Dynamic and robust routingSome ad hoc solutions (such as NATs) requirethat state information about a session is storedin a particular router in the network. If the ses-sion relies on this information, all traffic of asession must passes through the router. (ATCP session, for example, will be dependentof state information in a NAT for the sessionnot to break.) Thus, routing is not as dynami-cal and robust as in the original Internet archi-tecture.

• Connectionless servicesSome ad-hoc solutions (such as NATs) requirethat state information about a session is storedin a particular router in the network, and theunderlying communication model that is notlonger connectionless. The result is that it willbe more difficult to implement connectionlessservices at higher layers, too.

• Stable and unique IP addresses – ‘always on’servicesA private IP address is, unlike a global one,not a globally unique id of a network accesspoint. Due to the extensive use of privateaddresses, the IP address can no longer beused in generality for global identification.It also is common to allocate temporary IPaddresses to hosts in order to save addressspace (i.e. not bind addresses that are not inuse). The temporary allocation may beimplicit (e.g. NAT) or explicit (e.g. DHCP).The use of private and temporary addressesmakes it more difficult to provide ‘always on’services.

• The peer-to-peer communication modelNAT is an ad hoc solution that transforms theoriginal peer-to-peer communication modelinto a client-server model. The NAT requiresthat all communication is initiated from withinthe intranet. This works well as long as hostsaccess web servers, ftp servers or mail serveron the global Internet. However, new applica-tions that require peer-to-peer communication(such as Napster-resembling services or cer-tain configurations of SIP) will cease to work.Two hosts that are located behind differentNATs cannot communicate with each otherdirectly, because each host is required to initi-ate the communication.

Solutions to the New ProblemsThe change in the Internet architecture as de-scribed above seems like a serious degenerationof the original time glass model. IETF haveinvested years and years into development of theprotocols (e.g. Mobile IP, IPSec, SIP, Rsvp, 6to4etc.) that were intended to form the basis for thefuture Internet – an internet where computer –and tele-communication have converged and


where security, mobility and quality of servicecan be taken for granted. Instead, the protocols,which are developed under the assumption thatthe time glass model and end-to-end principleare valid, cease to work as intended.

IETF and the Internet society are currently work-ing hard to assure that the Internet architecturedoes not get to a state that excludes furtherdevelopment. There are three different app-roaches to handle the new situation:

1. New temporary ad hoc solution (short-term)In a short-term perspective one can fix short-comings of ad hoc solutions by making new adhoc solutions. For example:

• Firewall traversal: ‘Reverse Tunneling’ is anexample of a solution that makes basic mobileIP work despite the existence of a firewall.Lately we have seen a number of new propos-als on how to adapt existing protocols to thepresence of a firewall.

• NAT traversal: An FTP-ALG implemented ina NAT can make FTP work across the NAT.There are a number of ALGs for other proto-cols and applications. Many NATs on themarket is not possible to configure or areexpensive to configure and maintain. IETFtherefore works intensively with other solu-tions that do not require changes to the NAT.

• IP-in-IP tunneling has become a popular basisfor solutions. Such solutions are often in linewith the original Internet architecture and the

resulting time glass model is maintained asshown in Figure 3. It is unfortunately limitedhow many problems can be solved by meansof IP-in-IP tunneling.

The drawback of the efforts described above isthat each protocol calls for a separate solution.

2. MIDCOM communication (long term)Another approach is to develop protocols that letnodes (e.g. hosts, clients, servers or agents/prox-ies) communicate with a ‘MIDBOX’ (i.e. a logi-cal unit in the network such as a NAT or a fire-wall). One accepts the existence of MIDBOXesas part of a new communication model. Thismay for example mean a farewell to the end-to-end principle in its current form and to theassumption of network transparency. MID-BOXes, such as firewalls, are expected to bearound even after the introduction of IPv6.

3. IPv6 (long term)IPv6 was developed from scratch as a new Inter-net Protocol that restores the simplicity andfunctionality of IP and the validity of the timeglass model, as shown in Figure 4. It will proba-bly take some years before IPv6 is fully adoptedbecause the user base of IPv4 is so big. Mean-while the time glass model must go through a‘mid-life crisis’ with two Internet protocols inuse (Figure 5). During this transition periodthere will be problems with interoperabilitybetween the two Internet protocols. The dou-bling of service interfaces will lead to complex-ity, and changes in higher and lower layers willalso be required.

CSMA async sonet...

email WWW phone...

SMTP HTTP RTP...

TCP UDP…

IP

ethernet PPP…


CSMA async sonet...

email WWW phone...

SMTP HTTP RTP...

TCP UDP…

IP4

ethernet PPP…


IP6

Figure 3 IP-in-IP tunnelling Figure 4 ‘Mid-life crisis’



From ‘Time Glass Model’ to‘Wine Glass Model’?Figure 5 illustrates that IPv6 is less ‘narrow’than IPv4. This reflects that IPv6 has a signifi-cantly larger address space. Furthermore, IPv6is designed for extensibility by means of ‘headerextensions’ in the IP header. This will make itpossible to integrate mobility, security, re-num-bering etc. into the IPv6 protocol.

As IPv6 is a ‘wider’ protocol the IP layer will beable to solve problems that are solved by under-lying technologies today. Many people in theInternet society and in IETF dislike doublingof functionality, and would rather see that redun-dant functionality at lower layers be removed. Ifthis wish comes true we might one day replacethe time glass model with the ‘Wine glassmodel’, as shown in Figure 6.

ConclusionThe Internet architecture is about to change. Thewell-known functionality that it is about to lose,due to this change includes network trans-parency, dynamic and robust routing, connec-tionless services, stability and uniqueness of IPaddresses, always on services and the peer-to-peer communication model. In a short-term per-spective the changes creates new ad hoc solu-tion. In a long-term perspective IPv6 and mid-com communication will restore the originalmodel and solve the problems. In the future, wemay see that unnecessary functionality of under-lying technologies are stripped off, and the timeglass model may be replaced by 'wine glassmodel'.

Figure 5 Putting on weight ... Figure 6 The wine glass model

CSMA async sonet...

email WWW phone...

SMTP HTTP RTP...

TCP UDP…

ethernet PPP…


IPv6 + mcast+ QoS +...

email WWW phone...

SMTP HTTP RTP...

TCP UDP…

IP6

