Omega 35 (2007) 563–577www.elsevier.com/locate/omega

A hierarchical approach to multi-project planning under uncertainty

E.W. Hansa, W. Herroelenb, R. Leusb,∗, G. Wullinka

aSchool of Business, Public Administration & Technology, University of Twente, P.O. Box 217, 7500 AE, The NetherlandsbDepartment of Decision Sciences and Information Management, Katholieke Universiteit Leuven, Naamsestraat 69, B-3000 Leuven, Belgium

Received 24 June 2005; accepted 27 October 2005Available online 15 December 2005

Processed by B. Lev

Abstract

We survey several viewpoints on the management of the planning complexity of multi-project organisations under uncertainty.Based on these viewpoints we propose a positioning framework to distinguish between different types of project-driven organ-isations. This framework is meant to aid project management in the choice between the various existing planning approaches.We also discuss the current state of the art of hierarchical planning approaches both for traditional manufacturing and for projectenvironments. Next, we introduce a generic hierarchical project planning-and-control framework that serves to position planningmethods for multi-project planning under uncertainty. We discuss various techniques for dealing with the uncertainty inherent tothe different hierarchical stages in a multi-project organisation. In the last part of this paper we discuss two cases from practiceand relate these to the positioning framework that is put forward in the paper.� 2005 Elsevier Ltd. All rights reserved.

Keywords: Project management; Multi-project organisations; Hierarchical models; Uncertainty

1. Introduction

We aim at providing an integrated approach to multi-project planning under uncertainty that deals with boththe complexity aspects of the problem and with the un-certainty. Our goal is to present a general guide for us-ing advanced multi-project planning techniques in prac-tice. We propose a positioning framework to distinguishbetween different types of project-driven organisations.This framework is based on the dimensions variabil-ity and complexity of an organisation. The positioningframework enables the selection of appropriate multi-project management methods as a function of the organ-isational characteristics. We also propose a detailed hi-erarchical framework for project planning and control,

∗ Corresponding author. Tel.: +32 16 32 69 67.E-mail address: Roel.Leus@econ.kuleuven.be (R. Leus).

0305-0483/$ - see front matter � 2005 Elsevier Ltd. All rights reserved.doi:10.1016/j.omega.2005.10.004

which distinguishes four hierarchical levels. We discusseach level of the hierarchy with its associated projectplanning-and-control methods in detail.

Project management is a management discipline thatis receiving a continuously growing amount of atten-tion; comprehensive references are Mantel et al. [1],Meredith and Mantel [2], Rand [3], Shtub et al. [4] andTurner [5]. Both in production and in service sectors, anincreasing number of organisations adhere to project-based organisation. These organisations work withina wide variety of applications, such as research anddevelopment, software development, construction, pub-lic infrastructure, process re-engineering, maintenanceoperations, or complex machinery. A project can beinformally defined as a unique undertaking consistingof a complex set of precedence-related activities thathave to be executed using diverse and mostly limitedcompany resources. Project management deals with the

564 E.W. Hans et al. / Omega 35 (2007) 563–577

selection and initiation of projects, as well as with theiroperation, planning and control.

A significant number of international high-profileprojects fail to be delivered on time and on budget[6–8]. One example that immediately springs to mindis the construction of the Channel Tunnel, but undoubt-edly most readers can also recall smaller-scale projectscloser to their work environment, that did not work outas anticipated. A number of undesirable characteristicsare associated with failing projects: budget overruns,compromised project specifications, and missed mile-stones. In other words, the three basic dimensions ofproject success, namely time, cost and quality, are oftenin jeopardy. To avoid these problems, proper projectplanning is in order: a description of the objectives andgeneral approach of the project, its resources and per-sonnel, evaluation methods, and also a project scheduleas well as a description of potential problems that maybe encountered.

Traditionally, research in the area of project planninghas focused on planning for so-called single-project or-ganisations. An increasing number of companies, how-ever, tend towards an organisational structure in whichmultiple projects are performed simultaneously. A num-ber of authors [9–11], explicitly point out that compa-nies mostly run a number of projects in parallel thatshare the same scarce resources. Frequent conflicts ofinterest arise when more than one project requires thesame resource at the same time. In this paper we re-fer to the overall coordination of such multi-project or-ganisations as multi-project management. These multi-project organisations are characterised by a high degreeof complexity and uncertainty about the activities andoperations.

As coherently described in Silver et al. [12], Anthony[13] divides managerial activities into three broad cat-egories, whose names have evolved into strategic plan-ning, tactical planning and operational control. Thesecategories are concerned with different types of deci-sions and objectives, managerial levels, time horizonsand planning frequencies, and also with different mod-elling assumptions and levels of detail. To deal with theplanning complexity in multi-project organisations, theplanning process is broken down into more manage-able parts using a model for hierarchical planning andcontrol based on the three managerial decision levelsdiscerned in the foregoing.

All real-life projects are faced with uncertainty. Un-certainties in the multi-project-driven organisation havetwo main causes: detailed information about the re-quired activities becomes available only gradually (lead-ing to lack of information in the tactical stage), and

operational uncertainties on the shop floor. In the re-mainder of this paper we use the terms ‘uncertainty’ and‘variability’ interchangeably.

We can distinguish between two distinct approachesfor dealing with uncertainty, namely the proactive andthe reactive approach [14]. The proactive method triesto alleviate the consequences of uncertainties prior tothe start of the project, e.g. by allocating the flexibilityin a plan to uncertain activities, or to the periods wherethere are uncertainties. This flexibility usually takes theform of slack in time or in capacity. The reactive ap-proach aims at generating the best possible reaction to adisturbance that cannot be absorbed by the plan withoutchanging it. This can be done by, e.g., a replanning ap-proach, which re-optimises or repairs the complete planafter an unexpected event occurs. Reactive approachesare particularly useful if disturbances cannot be com-pletely foreseen or when they have too much impact tobe absorbed by the slack in a plan.

De Boer [15] points out that in many organisationspart of the work is made up by projects, while the rest isperformed in a ‘traditional manner’. A software house,for instance, may sell standard software applications,for which it has dedicated product development lines.At the same time, the house can provide custom-madesoftware applications for which project managers areresponsible. De Boer [15] uses the term ‘semi-project-driven’ to describe such organisations. In this text, wedo not specifically distinguish between project-drivenand semi-project-driven organisations: the techniqueswe study are applicable to the project-based part oforganisations, whether this constitutes all or only partof those organisations.

This paper is organised as follows. Section 2 surveysthe existing approaches to practical multi-project plan-ning. Section 3 discusses hierarchical planning-and-control frameworks that can be found in the literature,including the framework that we use in this paper. Sec-tions 4 and 5 treat the tactical and operational aspects ofplanning in further detail. It mainly focuses on methodsfor the tactical rough cut capacity planning (RCCP)problem and for the operational resource-constrainedproject scheduling problem (RCPSP). Section 6 setsout a number of requirements such that these two levelscan be integrated, and it discusses in which situationseach of the hierarchical levels deserves the most at-tention. Section 7 discusses two practical examplesof multi-project organisations for which the hierarchi-cal approach combined with the appropriate planningmethods would be beneficial or have been success-fully implemented. We end with some conclusions inSection 8.

E.W. Hans et al. / Omega 35 (2007) 563–577 565

2. Multi-project management

This section is devoted to multi-project management,the broader management discipline that encompassesthe planning function. Project planning is the maintarget of this paper; hence, the two terms ‘multi-projectmanagement’ and ‘multi-project planning’ are inter-changeably used in the remainder of this text. For adiscussion of the debate that has arisen recently inthe product-development literature on whether projectplanning and execution should be more or less struc-tured (the planned project management style providingmanagerial discipline and direction, monitoring mile-stones, formal project review and close managerialinvolvement in project details, feedback and adjust-ments, and the emergent style monitoring understand-ings, information gathering and participative controlfacilitating team members’ improvisation and expe-riential learning), we refer the reader to Eisenhardtand Tabrizi [16] and Lewis et al. [17]. The focus ofSection 2.1 is on the planning aspect of multi-projectmanagement. In Section 2.2 we discuss organisationalaspects of multi-project management. We present a po-sitioning framework for multi-project management inSection 2.3.

2.1. Multi-project management

Adler et al. [18,19] suggest adopting a process view-point to multi-project management. They remark thatmost managers think of multi-project management sim-ply as the management of a list of individual projects,rather than as a complex operation with a given capacityand workload. Their suggestion is compatible with theintroduction of a ‘Management by Projects’ (MBP) ori-entation at the enterprise level, which takes the benefitsof project management with its focus on specific projectgoals and deliverables as a starting point, but buildsit into the needs of the overall organisation. As such,MBP is the integration, prioritisation and continuouscontrol of multiple projects and operational schedulesin an enterprise-wide operating environment [20,21].Various approaches for “multi-project management andplanning” have been proposed in the literature. Realmulti-project approaches that are compatible with anMBP focus, however, are scarce.

Pennypacker and Dye [22] point out that there stillexists a difference between multi-project management(with the same content as what we defined as ‘MBP’)and project portfolio management. The former is gearedtowards operational and tactical decisions on capacityallocation and scheduling, and is the job of project or

resource managers; the latter is concerned with projectselection and prioritisation by executive and seniormanagement, with a focus on strategic medium- andlong-term decisions. Kavadias and Loch [23] developa dynamic model of resource allocation, taking intoaccount multiple interacting factors. The authors alsoderive an optimal admission policy for projects of vary-ing potential reward that arrive at unpredictable pointsof time. Anavi-Isakow and Golany [24] propose newproject control mechanisms that limit the number ofactive projects in multi-project environments. Incomingprojects are staggered into a network of inter-relatedresources in a way that maintains a stable load on thesystem through the concept of constant work-in-process(CONWIP).

Finally, multi-project management should also notbe confused with program management, which is aseparate concept altogether: program management isa special case of multi-project management that hasa single goal or purpose (for instance putting a manon the moon), whereas multi-project management gen-erally treats the case of multiple independent goals[25]. A program can be seen as a family of relatedprojects.

In the project-based part of an organisation, projectscompete for the same scarce resources. Unfortunately,many multi-project approaches do not recognise thisand treat the multi-project planning problem as a setof independent single-project planning problems by di-viding the resources over the projects based on projectpriorities or project status. In this way, the typical ‘re-source conflict’ that emerges when managing multipleconcurrent projects is overlooked. Moreover, many so-called advanced planning systems lack a multi-projectplanning function at the aggregate capacity level. Of-ten this lack is filled with an ‘aggregate scheduling’module, i.e., scheduling large aggregated jobs, and withfixed resource capacity, which is not capable of utilis-ing the capacity flexibility at the tactical level. After all,the tactical planning level has a longer planning hori-zon, which provides flexibility to anticipate uncertain-ties and changes in the project scope.

An aggregate, combined project plan is a good helpfor management to ensure that the organisation doesnot take on more projects than it can complete [26].It also facilitates cross-project analysis and reporting[27]. Maintaining integrated plans is difficult, however,because of the size of the projects, the dynamic natureof the project portfolio, the uncertainty inherent in eachindividual project, and the fact that different projectsusually have different project managers with differingor even conflicting objectives.

566 E.W. Hans et al. / Omega 35 (2007) 563–577

To adequately perform multi-project planning,projects must be considered simultaneously at all plan-ning levels, while taking into account that those differ-ent levels have different objectives, constraints, degreesof aggregation, and capacity flexibility. These objec-tives are, for instance, the optimal timing of operationsfor the operational level, optimal resource managementfor the tactical level, and the robustness or stabilityof plans for all levels. Multi-project management ap-proaches must deal with these objectives hierarchically.The techniques we study are applicable to the project-based part of organisations and can handle the varyingobjectives of complex multi-project organisations.

2.2. Organisational aspects of multi-projectmanagement

Meredith and Mantel [2] suggest that any time aproject is initiated, whether the organisation is onlyconducting a few occasional projects or is rather fullyproject-oriented and carrying on scores of projects, itmust be decided how to tie the project to the parentfirm, especially to its resources. Meredith and Manteldistinguish three major organisational forms commonlyused to house projects within an enterprise. We brieflydiscuss these three methods.

A first alternative for situating the project within theparent organisation is to make it entirely part of one ofthe functional divisions of the firm. This is only possi-ble when the activities particular to the project are allstrongly tied to the function performed by the functionaldivision that embraces it.

At the other end of the organisational spectrum, wefind a pure project organisation. The project is sepa-rated from the rest of the parent system and becomesa self-contained unit with its own dedicated staff andother resources. Single-project management techniquesat the operational level normally suffice for these cases.This structure has the disadvantage of duplication ofeffort in multiple functional areas and may induce sub-optimisation of project goals rather than overall organ-isation objectives. The benefit of pure project organi-sation is that projects can function autonomously withclear focus, and need not worry about conflicts withother projects or with functional departments.

The matrix structure is an intermediate solution be-tween the two extreme organisational models discussedabove, attempting to combine the advantages of bothand to avoid some of the disadvantages of each form.Resources are associated with functional departmentsbut are assigned to different ongoing projects through-out time. The strength of the link of resources between

their functional department and their current project(s)allows a wide range of different organisational choices.Assuming a ‘balanced’ matrix structure (not tilting to-wards any of the extremes), the multi-project organisa-tion can be modelled from a process viewpoint as a jobshop or assembly shop: work is done by functional de-partments that operate as workstations and projects arejobs that flow between the workstations.

2.3. A positioning framework for multi-projectorganisations

To distinguish between various multi-project organ-isations we propose a positioning framework that willallow us to categorise the various forms of multi-projectenvironments based on their characteristics. Earlier inthis paper we cited variability and complexity as twokey concepts that are often used in the literature abouthierarchical project management. Shenhar [28] for in-stance argues that not all projects have the same char-acteristics with respect to technological uncertainty andsystem complexity, and uses these two concepts to de-fine a framework in which he positions several practi-cal projects. His framework is the starting point for adiscussion of managerial styles that are best suited toparticular project environments. Shenhar [28] does notconsider environments in which multiple projects areexecuted simultaneously. Unfortunately, variability andcomplexity are not mutually exclusive: complexity be-ing a far broader concept, it entails variability.

Leus [29] and Herroelen and Leus [30] describe amethodological framework to position project plan-ning methods in which they distinguish two key de-terminants: the degree of general variability in thework environment and the degree of dependency ofthe project. ‘Dependency’ measures to what extent aparticular project is dependent on influences externalto the individual project. These influences can be ac-tors from outside the company (e.g. subcontractorsor material supply), but also dependencies from in-side, such as shared resources with other projects.Dependency forms part of the complexity of the plan-ning of a project-based organisation (as referred to inSection 1), and is the key complexity component wedistinguish. It will strongly determine the organisa-tional structure (see Section 2.2), although this choiceis not always exclusively based on the characteristics ofthe company. Other factors may also play a role, suchas unwillingness to change: choices that have been de-termined historically are sometimes hard to undo, eventhough better alternatives might be available under newcircumstances.

E.W. Hans et al. / Omega 35 (2007) 563–577 567

Table 1A positioning framework for multi-project organisations

Dependency LOW → HIGHVariability

LOW↓

HIGH

LL

HL

LH

HH

The resulting framework is depicted in Table 1. Weconsider the scale of the dimensions to be continuous.For simplicity we discuss the four extreme cases of lowand high variability and low and high dependency. Nev-ertheless, all possible intermediate positions in betweenthe four extreme cases are conceivable. We provide thetable with a case-by-case comment.

LL: Low variability and a low dependency can typ-ically be found in a dedicated single-project organisa-tion. In such organisations, resources are completelydedicated to one particular project and activities havea low degree of variability. An example is an on-sitemaintenance project that is performed on a preven-tive basis. Activities of these projects are often speci-fied in advance and executed routinely; the degree ofvariability is therefore relatively low. Moreover, suchmaintenance projects often have little interaction withother projects, so the degree of dependency is alsolow.

LH : In this project environment many project activ-ities are dependent on external actors. One can think,for instance, of a small furniture manufacturer that pro-duces wooden furniture on a make-to-order (MTO) ba-sis (e.g. chairs, beds, etc.). Most operations in such acompany are executed on universal woodworking ma-chines like drills, saws and lathes. Hence, the manu-facturing process is relatively basic, which results in alow degree of operational variability. Moreover, vari-ability in the process planning stage is relatively lowbecause of the low degree of complexity of the prod-ucts and the production processes. In contrast to the lowvariability in this setting, dependency of projects in thisenvironment is high because many projects may simul-taneously claim the same woodworking machines. ThisLH-setting is most related to the classical job-shop pro-duction environment.

HL: Typical examples of environments with highvariability combined with a low degree of dependencyare large construction projects. Such projects are usu-ally subject to large environmental uncertainties suchas weather conditions and uncertain or frequentlychanging project specifications. The degree of interac-tion with other projects is typically low because the

deployed resources are often dedicated, in view of theconsiderable size of the projects.

HH : A high degree of variability in combinationwith highly dependent projects can be encounteredin engineer-to-order (ETO) environments with severalcomplex projects in parallel. These projects are typi-cally completely new to the company, which resultsin a long engineering trajectory and many disruptionsand adaptations, e.g. because of specification changeson the part of the customer. As an example, we cancite a company that manufactures welding equipmentfor the automotive industry. Every product is designedfor a specific (new) car type. Therefore, every newproduct requires a long and intensive engineering pro-cess. Moreover, the customer may frequently requiremodifications of the design. Combining this with thecomplexity of the product results in a project envi-ronment with an extremely high degree of variability.Furthermore, such manufacturers often produce mul-tiple products simultaneously, which also results in ahigh degree of dependency between the projects.

A project from the HH-category requires planningand control approaches that are able to deal with boththe organisational complexity and the variability, as wellas with the complexity of the planning problem. There-fore, the lower right quadrant of the positioning frame-work is most difficult to manage. This paper provides aplanning-and-control framework and discusses severalplanning techniques that may prove to be useful evenin such environments. We also discuss the interactionbetween the proposed planning techniques on the dif-ferent hierarchical levels.

3. Hierarchical frameworks for planning andcontrol

Various hierarchical planning-and-control frame-works for manufacturing and project environments havebeen proposed. Section 3.1 surveys the existing litera-ture on hierarchical planning-and-control frameworksfor multi-project planning. Section 3.2 investigates therelated subject of hierarchical planning and controlfor manufacturing environments. Finally, in Section3.3, we present the hierarchical planning-and-controlframework that is proposed in this paper.

3.1. Hierarchical planning and control for projectorganisations

Fendley [31] is an early reference that discusses thedevelopment of procedures for the formulation of a

568 E.W. Hans et al. / Omega 35 (2007) 563–577

complete multi-project scheduling system that uses: (1)a method for assigning due dates to incoming projects;and (2) a priority rule for sequencing individual jobssuch that total costs are minimised (heuristically).Fendley points out that, because of the uncertainty ofprocessing times, it is almost impossible to maintainan advance schedule in a multi-project organisation.What can and should be determined in advance, ac-cording to this author, is a delivery date or due date foreach project by which the organisation desires to haveit completed. He remarks that since the performancetimes of the activities are uncertain, the sequencing ofthe individual activities must be handled on a dynamicbasis.

Leachman and Boysen [32] and Hackman and Leach-man [33] describe a two-phase hierarchical approach. Inthe first phase, due dates are selected for new projectsand resources are allocated among projects based on anaggregate analysis. An aggregate model of each projectis developed by aggregating detailed activities with sim-ilar mixes of resource requirements into aggregate ac-tivities. Given actual due dates for committed projectsand trial due dates for proposed projects, the aggregateproject models are then combined in a multi-project re-source allocation model that is formulated as a linearprogram. The linear program minimises the discountedcost of unused resources, i.e. the present value of costoverruns associated with charging ongoing projects forunutilised resources. The authors suggest to iterativelysolve linear programs and revise trial due dates until adesirable resource loading plan is constructed. Both theselected due dates and the computed resource alloca-tions define the single-project scheduling problem thatis addressed in the second phase.

Kim and Leachman [34] describe another hierarchicalmethodology to schedule multi-project environmentswith the objective of minimising total project latenesscosts. In the first stage, target resource profiles are com-puted for each project as convex combinations of theearly and late cumulative resource curves associatedwith the earliest- and latest-start CPM schedules. Thesetarget resource levels then serve as decision aids forregulating the progress speeds of the projects duringdetailed activity scheduling, using a heuristic proce-dure based on the variable-intensity model proposed byLeachman et al. [35].

Speranza and Vercellis [36] remark that little efforthas been devoted to a structured quantitative approachthat addresses the issue of integration between the tac-tical and the operational stages of the project planningprocess. They distinguish between a tactical and an op-erational level with different planning objectives at each

level. On the tactical level due dates are set and re-sources are allocated. On the operational (service) levelthe activity modes are set and the timing of the activitiesis determined. Their approach is based on the assump-tion that a set of aggregated activities forms a macro-activity on the tactical level. If these macro-activitiesare interrelated by means of precedence relations, theyform a program. Hartmann and Sprecher [37] have pro-vided counter-examples to show that the algorithm mayfail to determine the optimum.

Yang and Sum [38,39] propose a dual-level structurefor managing the use of resources in a multi-project en-vironment. A central authority (resource pool manageror director of projects [11] negotiates the project duedates with the customer, determines the allocation ofresources among projects such that resources are allo-cated to the critical projects, and decides on the projectrelease dates. The lower-level decisions of schedulingthe activities within each project are managed by anindependent project manager who schedules the activ-ities of his project using only the resources assignedto him. Yang and Sum [38] examine the performanceof heuristic resource allocation and activity schedulingrules. Yang and Sum [39] investigate the performanceof rules for due-date setting, resource allocation, projectrelease, and activity scheduling in a multi-project envi-ronment, where significant resource transfer times areincurred for moving resources from one project to an-other.

Franck et al. [40] propose a capacity-oriented hierar-chical approach for hierarchical project planning withproject scheduling methods. They distinguish severalplanning problems as for instance lot sizing, capac-ity planning, and shop-floor scheduling. They formu-late optimisation models that resemble the deterministicRCPSP. They do not distinguish between the differentplanning objectives of the various planning levels.

Dey and Tabucanon [41] propose a hierarchical inte-grated approach for project planning. They discuss dif-ferent planning objectives at different planning levelsand they use goal-programming techniques to solve thecorresponding planning problems. Nevertheless, they donot use a multi-project approach.

De Boer [15] proposes a hierarchical planning frame-work for project-driven organisations. In accordancewith several earlier sources in the literature, he arguesthat a hierarchical decomposition is needed to cometo a more manageable planning process. He also men-tions that, especially in project environments, uncer-tainties play an important role. De Boer argues thatif uncertainties are too large, channels in hierarchicalstructures become overloaded with information. Four

E.W. Hans et al. / Omega 35 (2007) 563–577 569

strategies are suggested to prevent this: (a) the creationof slack by lowering output targets; (b) the creation ofself-contained activities, i.e. large tasks that can be ex-ecuted by multi-disciplinary teams; (c) the creation oflateral linkages using e.g. a matrix organisation or spe-cial teams; and (d) investment in vertical informationsystems. De Boer claims that these strategies are an ef-fective way to deal with uncertainty in project-drivenorganisations. In line with many other authors, however,the methods proposed are deterministic planning tech-niques at the separate planning levels and they do notexplicitly account for variability.

Neumann et al. [42] (see also Neumann and Schwindt[43]) present and illustrate a three-level hierarchicalmulti-project planning process under the assumptionthat a portfolio of long-term projects is to be performedwithin a planning horizon of 2–4 years. Each projecthas a given release date, deadline and work breakdownstructure, i.e. it consists of subprojects, which includedifferent work packages, each of which can be de-composed into individual activities. At the first level(long term) all the projects are grouped into a singlemulti-project network that contains all the subprojectsas aggregate activities. The release date and deadlinesare modelled using generalised precedence relations.The aggregate activities are to be scheduled subject toscarce key resources (e.g. experts, research equipment,special-purpose facilities). The estimated duration of anaggregate activity equals the critical-path length of thecorresponding subproject plus a time buffer that antic-ipates the time extension of the aggregate activity thatwill occur due to the scheduling of the disaggregatedprojects at the third planning level. Neumann et al. sug-gest the estimation of the size of the time buffers us-ing queuing theory. The key resource requirement of anaggregate activity is computed as the ratio of the totalworkload of the corresponding subproject and its pre-estimated duration. The capacity of the key resources isfixed by the general business strategy. The financial ob-jective function is the maximisation of the net presentvalue of the project portfolio. The resulting scheduleprovides a maximum duration for every project, and theresulting resource profiles provide the time-dependentresource capacities for the key resources at the secondplanning level. At the second level (medium term) eachproject is condensed by choosing the aggregate activi-ties to be the work packages. The durations, time lagsand resource requirements are determined analogouslyto what happened at the first level. At the second level,Neumann et al. also consider primary resources (tech-nical and administrative staff or machinery) with un-limited availability. The objective is to level the use of

these resources over the project duration. At the thirdplanning level (short term) the condensed projects aredisaggregated into detailed projects with individual ac-tivities. Resource constraints are given for the key andprimary resources as well as for low-cost secondary re-sources (tools, auxiliary resources). The objective is tominimise the project duration.

3.2. Hierarchical planning and control formanufacturing organisations

The majority of the work on hierarchical productionplanning (HPP) focuses on manufacturing environmentsrather than project environments. Some authors arguethat shop-floor planning is a specialisation of multi-project planning. We adhere to this point of view for thediscussion of hierarchical planning-and-control frame-works. Therefore, we also discuss work on hierarchicalplanning-and-control frameworks for shop floor manu-facturing environments. A fundamental study on HPPis that of Hax and Meal [44]. After this, several articleson hierarchical integration of different planning func-tions followed: for instance, Bitran et al. [45], Bitranand Tirupati [46], Bertrand et al. [47], Hax and Caneda[48] and Vollmann et al. [49].

In a review article about intelligent manufacturingand control systems, Zijm [50] remarks that in prac-tice the existing hierarchical planning approaches haveproven to be inadequate for several reasons. The mainreason, he argues, is that the existing planning frame-works are either material-oriented (e.g. MRP/MRP IIsystems) or capacity-oriented (HPP systems). Zijm pro-poses a hierarchical framework that focuses on the inte-gration of technological and logistics/capacity planning,and the integration of capacity planning and materialcoordination. Zijm also mentions that there is a lack ofappropriate aggregate capacity-planning methods at theorder-acceptance level.

To fill this gap, Hans [51] proposes several determin-istic models and techniques to solve the aggregate (tac-tical) capacity planning problem, which he refers to asthe resource loading problem. With these deterministictechniques a planner can quote reliable due dates andestimate the capacity requirements over a time horizonof several weeks to several months. Hans claims thatthese methods can also be used for multi-project capac-ity planning in project environments.

Kolisch [52] proposes a hierarchical framework todistinguish between the managerial processes in MTOmanufacturing. He distinguishes three processes/levels,namely, the order-selection level, the manufacturing-planning level, and the operations-scheduling level.

570 E.W. Hans et al. / Omega 35 (2007) 563–577

Project selection,rough-cut capacity

R&D, knowledge Strategic

Tactical

Operational

Strategic resource planning

planning

Resource-constrainedproject scheduling

Detailed schedulingand resource allocation

Supply chain design,warehouse design

Procurement andpurchasing

Order picking, routing,and order batching

Resource capacityplanning

Materialcoordination

Technological planning

Macro processplanning

Micro processplanning, engineering

management

Fig. 1. Hierarchical framework.

He also proposes deterministic models for the variouslevels.

From this brief review of hierarchical productionplanning-and-control frameworks we can conclude thatseveral frameworks have been proposed for shop-floor-oriented manufacturing environments and for project-driven organisations. Only few, however, actually dealwith different objectives of the planning problems atdifferent managerial levels. Moreover, little effort hasbeen devoted to the aspect of variability in the hierar-chical multi-project planning approach, the integrationof technological planning and logistics planning, andthe integration of material coordination and capacityplanning.

3.3. Hierarchical planning and control formulti-project organisations

We propose a hierarchical project planning-and-control framework that is partly based on the frame-work that was proposed by De Boer [15]. We haveadapted the framework to be able to discern the variousplanning functions with respect to material coordina-tion and technological planning. As shown in Fig. 1,we distinguish three hierarchical levels: (a) the strate-gic level; (b) the tactical level; and (c) the operationallevel. We distinguish three functional planning areas:(a) technological planning; (b) capacity planning; and(c) material coordination.

In this hierarchy we define four capacity-planningfunctions: (a) strategic resource planning; (b) RCCP;(c) the RCPSP; and (d) detailed scheduling. Contrary toDe Boer, we position both the RCPSP and the detailed

scheduling and resource allocation problem at the oper-ational level. Since RCPSP and resource allocation aretwo different problems [29] we treat them separately. InSections 4 and 5 we elaborate on the tactical (RCCP)and operational (RCPSP) planning level.

At each level of the hierarchy, the positioning frame-work of Table 1 can be applied. Some organisationsare characterised by a high degree of variability on theoperational level whilst on the tactical level the uncer-tainties are much more controllable. At the same time,the dependency of projects in some companies maybe considerable on the tactical level while projects arecompletely independent on the operational level. Thesedifferences play an important role in modelling the in-teractions between the hierarchical planning levels; wewill elaborate on this issue of interaction between thelevels in Section 6.

4. Rough-cut capacity planning

In the early project stages, projects may vary signif-icantly with respect to routings, material, tool require-ments, or the work content of activities. In spite of theuncertain project characteristics, project accept/rejectdecisions must be made, and important milestones (suchas the due date) must be set. It is common practice thatcompanies accept as many projects as they can possi-bly acquire, although the impact of a decision on theoperational performance of the production system is ex-tremely hard to estimate. Moreover, to acquire projects,companies tend to promise a delivery date that is as earlyas possible. This is generally done without sufficiently

E.W. Hans et al. / Omega 35 (2007) 563–577 571

assessing the impact of these projects on the resourcecapacity, which typically leads to a serious overload ofresources, thus having a devastating effect on the deliv-ery performance and the profitability of the productionsystem as a whole.

Customers require reliable project due dates as partof the service mix offered by the company during ordernegotiation. Being able to quote tight and reliable duedates is a major competitive advantage. Therefore, atthe negotiation and acceptance stage, adequate RCCPmethods that assess the consequences of decisions forthe production system are essential. Contrary to theoperational planning level, the tactical planning stageis characterised by a high degree of capacity flexibil-ity (e.g. by working in non-regular time or by subcon-tracting). Tactical planning, therefore, requires meth-ods that use more aggregate data, and that can exploitthis flexibility. Ideally, RCCP-methods should use thiscapacity flexibility to support the planner in makinga trade-off between the expected delivery performanceand the expected costs of exploiting flexibility by usingnon-regular capacity.

Adequate deterministic planning approaches for theRCCP-problem have been proposed by De Boer [15],Hans [51], Gademann and Schutten [53], and Kis [54].These tactical planning approaches all use an objectivefunction that minimises the cost of using non-regularcapacity in a period, hiring additional staff and subcon-tracting. De Boer [15], Hans [51] and Gademann andSchutten [53] implicitly claim that for project environ-ments that are in the LL- and LH-categories it suffices tochoose a proper data-aggregation level to cope with thedisturbances that might occur. Although this assumptionmay be justified for environments that are in the areasLL and LH of our positioning framework, restriction ofattention to the choice of a proper data-aggregation leveland deterministic planning approaches is not sufficientfor HL- and HH-projects.

We believe that project planning methods need tobe able to deal with the uncertainties that are typicalfor the particular planning level they work on. Theseuncertainties may range from unexpected operationalevents (e.g. machine breakdowns or operator unavail-ability) to uncertainties that typically result from thelack of information at the concerned project stage.The former category of uncertainties is typically dealtwith at the operational planning level. The latter cate-gory typically arises in the earlier project stages, andis handled at the tactical (RCCP) level. Elmaghraby[55] affirms that the work content of an activity isone of the most important sources of variability. Heclaims that resource-capacity-management methods

that can deal with these uncertainties have a decisiveimpact on the overall performance of a project-drivenorganisation. Uncertainties that can be considered inRCCP-models are, for instance, the work content ofan activity, activity occurrence, resource availabilityor release and due dates. In general, the determinis-tic models for RCCP have been developed under theassumption that the aforementioned uncertainties aredealt with by using a proper level of aggregation and byreserving additional resource capacity. Few planningapproaches explicitly take into account variability at theRCCP-stage.

Wullink et al. [56] propose a proactive approachto deal with the RCCP-problem under uncertainty.They use a scenario-based approach to model uncer-tain work content of activities. With a scenario-basedMILP-model they minimise the expected costs of us-ing non-regular capacity. The scenario-based approachresults in considerable improvements with respectto the expected costs over all scenarios of a plancompared to the previously proposed deterministicapproaches.

In general, mathematical optimisation techniques fo-cus on optimality of a solution. When an optimum isreached, the problem is generally considered to havebeen solved. Alternative solutions with equivalent or al-most equivalent values for the objective function areusually discarded. Nevertheless, such solutions mightconstitute an improvement with respect to other criteriathan the initial objective, such as robustness (an esti-mate of the ability of the plan to absorb disturbances),which is valuable when uncertainty is present. Such anapproach allows a planner to assess several plans onmore practical characteristics that are difficult to quan-tify, but that are generally implicitly taken into consid-eration during the planning process (e.g. whether a planis workable in practice, or the degree in which an activ-ity is spread over the periods). It may also prove use-ful to generate a set of solutions that contains Pareto-optimal solutions on two or more criteria. For instance,the planning process can focus on trade-offs betweenrobustness and the cost of non-regular capacity.

The approaches to deal with variability on the tac-tical level we have discussed so far are all proactiveapproaches: they aim at anticipating uncertain events.Reactive methods for tactical planning are also possi-ble. These methods normally use one or more replan-ning rules that are applied when a disturbance occurs, inorder to generate a new plan. Most companies alreadyapply reactive planning by updating their plans with acertain frequency or when existing plans have becomeinfeasible.

572 E.W. Hans et al. / Omega 35 (2007) 563–577

5. Resource-constrained project scheduling

Our focus in this section is on the simultaneousscheduling of multiple projects. Apart from the hier-archical multi-project planning schemes discussed inSection 3.1, existing research efforts in multi-projectscheduling have mainly assumed a single-level struc-ture where a single manager oversees all projects andwhere the resource transfer times for moving resourcesfrom one project to another are negligible. In a firstapproach, projects are artificially bound together into asingle project by the addition of two dummy activitiesrepresenting the start and end of the single ‘aggregate’project, possibly with different ready (arrival) times andindividual due dates. In such a case, existing exact andsuboptimal procedures for single-project schedulingmay be used for planning the aggregate project.

Alternatively, the individual projects are consideredto be independent, and specific multi-project schedul-ing techniques—mostly heuristic in nature—are used.Kurtulus and Davis [57] report on computational ex-perience obtained with six priority rules under the ob-jective of minimising total project delay. Kurtulus [58]and Kurtulus and Narula [59] analyse the performanceof several priority rules for resource-constrained multi-project scheduling under equal and unequal project de-lay penalties. Lova et al. [10] have developed a multi-criteria heuristic for multi-project scheduling for bothtime-related and time-unrelated criteria. Lova and Tor-mos [60] have developed combined random samplingand backward–forward heuristics for the objectives ofmean project delay and multi-project duration increase.

Several authors have studied the problem of assigningdue dates to the projects in a multi-project environment.Dumond and Mabert [61] evaluated the relative per-formance of four project due-date heuristics and sevenresource allocation heuristics; related research can befound in Dumond [62]. Bock and Patterson [63] inves-tigate several of the resource assignment and due-date-setting rules of Dumond and Mabert [61] to determinethe extent to which their results are generalisable todifferent project data under conditions of activity pre-emption. Lawrence and Morton [64] also studied thedue-date setting problem and performed large-scale test-ing of various heuristic procedures for scheduling mul-tiple projects with weighted-tardiness objective. Severalmodel extensions are discussed in Morton and Pentico[65].

As we mentioned earlier in Section 3.1, in a hi-erarchical project-management system, due dates areusually set on the tactical level. Yang and Sum [39] de-termine due dates on the first level of their suggested

dual-level structure. They reach conclusions that areconsistent with the ones reported in the references listedin this section. Better due dates are obtained by usinginformation that goes beyond critical-path length andnumber of activities, and by taking into account thework content of the projects. Yang and Sum also con-clude that the relative performance ranking of the due-date rules is unaffected by the presence of customers’control over the due dates and by the choice of the otherdecision rules for resource allocation, project releaseand activity scheduling. In our hierarchical frameworkshown in Fig. 1, we assume that due dates are set onthe RCCP level.

All the methods described so far in this section sched-ule the project activities for efficiency in a determinis-tic environment and under the assumption of completeinformation. During execution, however, the project issubject to considerable uncertainty, which may lead tonumerous schedule disruptions—we refer the reader tothe variability-dimension of Table 1. This variabilityfactor in the framework involves a joint impression ofthe uncertainty associated with the size of the vari-ous project parameters (time, cost, quality), uncertaintyabout the basis of the estimates (activity durations, workcontent), uncertainty about the objectives, priorities andavailable trade-offs, and uncertainty about fundamen-tal relationships between the various project parties in-volved. Reliable and effective RCCP will also have astrong beneficial impact on variability at the operationallevel.

When dependency and variability are both low (caseLL), deterministic single-project scheduling methodscan be used to schedule each individual project in amulti-project environment: the project can be plannedand executed with dedicated resources and withoutoutside restrictions. For case HL, with high variabilityand low dependency, a detailed deterministic schedulecovering the entire project will be subject to a high de-gree of uncertainty. Dispatching of individual activitiesaccording to some decision rule (without prior overallschedule) is possible, since the resources are availablealmost 100% to the project. Alternatively, a reactiveapproach can be followed: reactive scheduling revisesor re-optimises the baseline schedule when unexpectedevents occur. Proactive schedules are schedules thatare as well as possible protected against anticipatedschedule disruptions that may occur during project exe-cution. Proactive scheduling techniques can be appliedto enhance the quality of objective-function projectionsin reactive scheduling.

In the high-dependency case (the right column ofTable 1), a large number of resources are shared and/or

E.W. Hans et al. / Omega 35 (2007) 563–577 573

a large number of activities have constrained time win-dows. A stable plan should be set up for these activities,such that small disruptions do not propagate throughoutthe overall plan. Stability is a particular kind of robust-ness that attempts to guarantee an acceptable degree ofinsensitivity of the activity starting times of the bulk ofthe project to local disruptions; for more details on sta-bility in scheduling we refer to Leus [29]. Satisficingmay be required to obtain a feasible plan with a min-imal number of (for instance resource) conflicts. CaseHH is best seen from a process-management viewpoint:the resources are workstations that are visited by (orvisit) work packages and pass these on to the appropri-ate successor resources after completion. A rough ball-park plan can be constructed to come up with interme-diate milestones, which can be used for setting prioritiesfor the resources in choosing the next work package toconsider.

Intermediate cases with moderate dependency maybenefit from an identification of what we refer to as thedrum activities: these are the activities that induce thedependency. Either they are performed by shared inter-nal or external resources, or their start or completiontime is constrained. It may make sense to adopt a two-level scheduling pass, planning the drum activities firstand the remaining activities afterwards. The drum canbe scheduled either efficiently or in a stable manner; theremaining activities can either be scheduled from thestart or be dispatched according to the progress on thedrum.

6. Interaction of the hierarchical levels

Planning approaches on the various hierarchical lev-els cannot operate independently from each other. Infor-mation that is generated by other (planning) functionsin the framework should be exploited to the best possi-ble extent. More specifically, it should be clear whichinformation is passed down form high to low levels andvice versa. Several authors have discussed the interac-tion between the various hierarchical planning levelswith a focus on manufacturing organisations. Krajew-ski and Ritzman [66] give a survey of a disaggregationapproach in manufacturing and service organisations.For a multi-stage system with multiple products andnonlinear assembly trees they state that this problem ishard to solve because of its computational complexity.Therefore, they propose to use MRP for this problem.In general, however, MRP is not suitable for MTO andETO environments. Kolisch [52] remarks that assem-blies and subassemblies should be exploded into indi-

vidual operations with detailed resource requirements,and that resources be differentiated with respect to theirspecific qualifications.

Most authors, however, do not describe the actual in-teraction and the information that is exchanged betweenthe planning levels. We will discuss this interaction be-tween the various hierarchical levels based on our po-sitioning framework. We distinguish between project-driven organisations with low and high dependency.

Project organisations with high dependency (LH andHH) generally adopt a matrix-organisational structure.For this type of multi-project organisations, we suggestthe exchange of information between the tactical andoperational planning levels in the following way. Inthe early stages of the project, when only rough infor-mation about the project content is available, the mostimportant output of RCCP-methods are internal andexternal due dates, milestones and required capacitylevels. This information serves as the basis for ac-quiring additional resources if necessary, ordering rawmaterials and final fixing of due dates. In later projectstages, more information gradually becomes availableas more preparatory work is performed. These dataare combined with information generated by processplanning and design and passed on as input for theoperational planning phase. Operational planning it-self consists of a multi-project RCPSP, as discussed inSection 5.

The other two cases in our framework (LL and HL)correspond to the other end of the organisational spec-trum: the dedicated or pure project organisation. Forthis kind of organisation we propose a different way ofinteraction. Here, resources are dedicated to a specificproject, and so the assignment of resources to projectscan already be done in the tactical stage. Therefore, be-sides the information that was exchanged between thehierarchical levels in the HH and LH cases (i.e. duedate, milestones and capacity levels), resource alloca-tion decisions are also passed down to the operationallevel in cases HL and LL. Consequently, the subsequentoperational planning problem is single-project oriented:multiple separate single-project plans are developed atthe operational level.

For clarity of exposition, the foregoing paragraphshave described two extreme forms of interaction, but inpractice intermediate solutions may be distinguished.We have also focused solely on the capacity plan-ning aspects of the interaction. The exchange of alot of additional information between the hierarchi-cal levels has been left unmentioned, for instance inthe domain of technological planning and materialcoordination.

574 E.W. Hans et al. / Omega 35 (2007) 563–577

7. Practice

In this section we discuss two research projects thattook place at Dutch companies. One case concerns theimplementation of a hierarchical multi-project planningapproach in the ship-repair industry. The second case isa typical example of a project-driven HH environment.

7.1. Royal Netherlands Navy Dockyard

The Royal Netherlands Navy Dockyard (RNND) isa public company that is responsible for the mainte-nance, repair and modification of national-defence ma-rine equipment. Three sorts of maintenance can gen-erally be distinguished, namely appointed incidentalmaintenance, incidental maintenance, and intermedi-ate maintenance. Maintenance of the last category isplanned on a long-term basis. The organisational struc-ture is purely functional: resources are associated witha functional department. This short case is an exam-ple of a successful implementation of a hierarchicaldecision-support system (DSS) at RNND, which is alarge project-driven organisation, and a typical exampleof a multi-project environment with a high degree ofdependency. Variability is mostly limited and is there-fore not explicitly dealt with by advanced quantitativetechniques.

Before the implementation of a hierarchical DSS thesituation was as follows. Before the start of each project,the dockyard made all tactical planning decisions suchas determining due dates, or hiring additional person-nel. At the same time, in order to support these deci-sions, the process-planning function also produced a lotof estimations about, for example, the duration of de-tailed operations. Because of a lack of separation be-tween (the timing of) operational and tactical decisions,too many (operational) data were needed in early stagesof the projects to support order acceptance, capacity-requirements estimates, and the determination of impor-tant milestones for the project. In other words, opera-tional decisions were made relatively early on and basedon numerous and often imprecise estimations, whichresulted in countless adjustments of the detailed planwhen new information became available at later stagesof the project.

In the positioning framework proposed in Section 2.2,we are in case LH for the tactical level. The degree ofuncertainty varies over the projects depending on theproject type, but is mainly of an operational nature.Incidental projects could be much more uncertain thenintermediate projects. There is, however, always a largedegree of dependency between projects.

In his PhD-thesis, De Boer [15] describes the imple-mentation of a hierarchical DSS for multi-project plan-ning at RNND. He argues that if organisations are toolarge for coordination by simple adjustment of data, ahierarchy is needed to deal with the uncertainty. In otherwords, a hierarchical structure facilitates the downwarddelegation of responsibilities, especially in case of alarge number of disruptions and exceptions. The DSScontains two hierarchical levels: a tactical RCCP-leveland an operational RCPSP-level. Applying a hierarchi-cal DSS with at each level the appropriate quantitativeplanning approach has turned out to be successful. De-terministic planning approaches are currently used tosolve both the RCCP and RCPSP. For RCCP, this ap-proach is satisfactory, while for RCPSP the main mo-tivation for this choice was the lack of adequate plan-ning approaches that account for uncertainties. The lastfew years, however, planning techniques that accountfor uncertainty have been emerging. Embedding suchtechniques in the DSS would allow generating robustand stable plans for the operational level, and is a validfuture line of research.

7.2. REPRO

The REPRO project aimed at improving the produc-tivity of several large Dutch ship-repair yards in Rotter-dam and Amsterdam. One of the most important char-acteristics of the ship-repair industry is the uncertaintyof activities in repair projects, making it hard to esti-mate the duration of each project. When a ship has hada collision at sea, for instance, the damage to the shipis often (mainly) located under the water level, so aninspection to establish the extent of the damage is im-possible without dry-docking the ship. Nevertheless, inthe negotiation process with a ship-repair yard, a duedate has to be established. The competition in the repairindustry is fierce so that the shipyard wants to quote acompetitive due date. At the same time, if the quoted duedate is not met, high delivery penalties may be charged.

This case is an example of a multi-project environ-ment with a high degree of variability and a high degreeof dependency (HH). We discuss this case to illustratethe need for multi-project planning approaches that candeal with the organisational complexity and the uncer-tainty that characterises the environment as well as withthe complexity in the planning process.

Once a ship is at the yard for repair, a project leader isassigned and repair starts immediately. At the time of in-vestigation, the organisational structure resembled a ma-trix structure, but there was no overall planning mech-anism for assigning resources to projects. Planning of

E.W. Hans et al. / Omega 35 (2007) 563–577 575

projects was limited to CPM-computations and capac-ity constraints were completely ignored. This resultedin each project leader claiming as much workforce ashe could possibly obtain, in order to make sure that hecould achieve his goal, which is completing his ship intime. At the same time, other project leaders competedfor the same resource capacity. This single-project ap-proach resulted in an unbalanced and inefficient use ofresource capacity. One week the yard needed a hugeamount of external workforce to complete a project intime, the other week a large part of the workforce wassent home because of a lack of work.

At the ship-repair yards considered in this project theoperational planning was considered to be overly un-certain for it to be performed in advance. Rough tac-tical planning (CPM-based) was used to manage theoperations on the shop floor. It was claimed that theprojects were too uncertain to be planned at all, so onlyrough estimates of the lead-time were made. The se-quence of operations was generated on the shop floorusing straightforward dispatching rules based on com-mon sense of the foremen.

As we mentioned before, this is a typical example ofa project organisation that corresponds with case HHin our positioning framework: uncertainty is high andprojects are strongly dependent. Robust multi-projectplanning methods for the tactical level could yield bet-ter balanced resource utilisation and more reliable duedates. For the operational planning, if uncertainty is toohigh, dispatching is indeed in order, but simple deter-ministic scheduling approaches combined with replan-ning rules are preferable for moderate uncertainty. Ifprojects get larger and better documented, more ad-vanced robust scheduling approaches might be useful:a stable initial plan (not overly detailed) can determinethe pace of the project. This is an example that illustratesthat not every environment is suitable for the most ad-vanced planning techniques, at least not from the start.If possible, variability reduction techniques should beapplied, and the operational scheduling level need notreceive too much attention before the broader tacticaldecision-making process has been streamlined.

8. Conclusions

In this article, we have proposed a positioning frame-work for multi-project planning environments, and wehave pointed out that different levels of hierarchicaldecision-making (strategic, tactical and operational) re-quire different methods and should not always be com-bined into one ‘monolithic’ model. Our models should

allow practitioners to better manage and control com-plex multi-project environments with uncertainty. Inparticular, we recommend managers to examine whichof the four cases of the developed positioning frame-work (LL, LH, HL, HH) best applies to each of theirdecision-making levels, and act according to the guide-lines that we have formulated for that setting.

We have also discussed the current state of the artin the research on hierarchical planning approaches,both for ‘traditional’ manufacturing organisations andfor project environments. Two cases from practice havebeen included to illustrate the ideas that were put for-ward in this text.

Acknowledgements

This work was supported by the NWO (The Nether-lands Organisation for Scientific Research). The workwas performed while Roel Leus was a research assistantof the Research Programme of the Fund for ScientificResearch—Flanders (Belgium).

References

[1] Mantel Jr. SJ, Meredith JR, Shafer SM, Sutton MM. Projectmanagement in practice. New York: Wiley; 2005.

[2] Meredith JR, Mantel Jr. SJ. Project management. A managerialapproach. 5th ed., New York: Wiley; 2003.

[3] Rand GK, Project management. In: Gass SI, Harris CM, editors.Encyclopedia of operations research and management science,Dordrecht: Kluwer Academic Publishers; 2001. p. 699–650.

[4] Shtub A, Bard JF, Globerson S. Project management: processes,methodologies and economics. 2nd ed., Englewood Cliffs, NJ:Prentice-Hall; 2004.

[5] Turner JR. The handbook of project management. 2nd ed.,London: McGraw-Hill; 1999.

[6] Feldman EJ. Concorde and dissent—Explaining hightechnology failures in Britain and France. Cambridge:Cambridge University Press; 1985.

[7] Standish Group, The. The CHAOS Report. Report available at〈http://www.standishgroup.com/sample_research/chaos_1994_1.php〉; 1994.

[8] Winch G. Thirty years of project management what havewe learned? Text available at 〈http://bprc.warwick.ac.uk/repwinch.html〉; 1996.

[9] Levy N, Globerson S. Improving multi-project management byusing a queuing theory approach. Project Management Journal1997; December: 40–46.

[10] Lova A, Maroto C, Tormos P. A multicriteria heuristic methodto improve resource allocation in multi-project scheduling.European Journal of Operational Research 2000;127:408–24.

[11] Payne JH. Managing of multiple simultaneous projects: a state-of-the-art review. International Journal of Project Management1995;13(3):163–8.

[12] Silver EA, Pyke DF, Peterson R. Inventory management andproduction planning and scheduling. New York: Wiley; 1998.

576 E.W. Hans et al. / Omega 35 (2007) 563–577

[13] Anthony RN. Planning and control systems: a framework foranalysis. Studies in management control, Graduate School ofBusiness Administration, Harvard University; 1965.

[14] Herroelen W, Leus R. Project scheduling under un-certainty—survey and research potentials. European Journal ofOperational Research 2004;165(2):289–306.

[15] De Boer R. Resource-constrained multi-project management—ahierarchical decision support system. PhD thesis, University ofTwente, Enschede; 1998.

[16] Eisenhardt KM, Tabrizi BN. Accelerating adaptive processes:product innovation in the global computer industry.Administrative Science Quarterly 1995;40:84–110.

[17] Lewis MW, Welsh MA, Dehler GE, Green SG. Productdevelopment tensions: exploring contrasting styles of projectmanagement. Academy of Management Journal 2002;45(3):546–64.

[18] Adler PS, Mandelbaum A, Nguyen V, Schwerer E. From projectto process management: an empirically-based framework foranalyzing product development time. Management Science1995;41(3):458–84.

[19] Adler PS, Mandelbaum A, Nguyen V, Schwerer E. Gettingthe most out of your product development process. HarvardBusiness Review March 1995;March–April:134–52.

[20] Boznak RG. Management of projects: a giant step beyondproject management. PM Network 1996;10(January):27–30.

[21] Advanced Management Solutions. Improving businessperformance through management by projects. Whitepaper, available at <http://www.amsrealtime.com/news/articles/mngmt_by_projects_whitepaper.pdf>, 2003.

[22] Pennypacker JS, Dye LD. Project portfolio management andmanaging multiple projects: two sides of the same coin?. In:Pennypacker JS, Dye LD, editors. Managing multiple projects,Planning, scheduling, and allocating resources for competitiveadvantage. New York: Marcel Dekker; 2002. [Chapter 1].

[23] Kavadias S, Loch CH. Optimal project sequencing with recourseat a scarce resource. Production and Operations Management2003;12(4):433–44.

[24] Anavi-Isakow S, Golany B. Managing multi-projectenvironments through constant work-in-process. InternationalJournal of Project Management 2003;21:9–18.

[25] Wysocki RK, Beck Jr. R, Crane DB. Extensions to multipleprojects. In: Pennypacker JS, Dye LD, editors. Managingmultiple projects. Planning, scheduling, and allocating resourcesfor competitive advantage. New York: Marcel Dekker; 2002.[Chapter 13].

[26] Wheelright SC, Clark KB. Creating project plans tofocus product development. Harvard Business Review 1992;March–April: 70–82.

[27] Kerzner H. Project management. A systems approach toplanning, scheduling and controlling. New York: Wiley; 1998.

[28] Shenhar AJ. One size does not fit all projects: exploring classicalcontingency domains. Management Science 2001;47(3):394–414.

[29] Leus R. The generation of stable project plans. Complexity andexact algorithms. Ph.D. thesis, Katholieke Universiteit Leuven,Belgium; 2003.

[30] Herroelen W, Leus R. Robust and reactive project scheduling:a review and classification of procedures. International Journalof Production Research 2004;42(8):1599–620.

[31] Fendley LG. Toward the development of a completemulti-project scheduling system. The Journal of IndustrialEngineering 1968;19(10):505–15.

[32] Leachman RC, Boysen J. An aggregate model for multi-projectresource allocation. In: Dean BV, editor. Project management:methods and studies. Amsterdam: North Holland; 1985.p. 43–64.

[33] Hackman ST, Leachman RC. An aggregate model of project-oriented production. IEEE Transactions on Man, Systems andCybernetics 1989;19(2):220–31.

[34] Kim SY, Leachman RC. Multi-project scheduling with explicitlateness costs. IIE Transactions 1993;25(2):34–44.

[35] Leachman RC, Dincerler A, Kim S. Resource-constrainedscheduling of projects with variable-intensity activities. IIETransactions 1990;22(1):31–40.

[36] Speranza MG, Vercellis C. Hierarchical models for multi-projectplanning and scheduling. European Journal of OperationalResearch 1993;64:312–25.

[37] Hartmann S, Sprecher A. A note on “Hierarchical models formulti-project planning and scheduling”. European Journal ofOperational Research 1996;94:377–83.

[38] Yang KK, Sum CC. A comparison of resource allocationand activity scheduling rules in a dynamic multi-projectenvironment. Journal of Operations Management 1993;11:207–18.

[39] Yang KK, Sum CC. An evaluation of due date, resourceallocation, project release, and activity scheduling rules in amulti-project environment. European Journal of OperationalResearch 1997;103:139–54.

[40] Franck B, Neumann K, Schwindt C. A capacity-orientedhierarchical approach to single-item and small batch productionplanning using project-scheduling methods. OR Spektrum1997;19:77–85.

[41] Dey PK, Tabucanon MT. Hierarchical approach to projectplanning: the case of a petroleum pipeline construction. AppliedMathematical Modelling 1996;20:683–98.

[42] Neumann K, Scwhindt C, Zimmermann J. Project schedulingwith time windows and scarce resources—temporal resource-constrained project scheduling with regular and nonregularobjective functions. 2nd ed., Berlin: Springer; 2003.

[43] Neumann K, Schwindt C. A capacitated hierarchical approachto make-to-order production. European Journal of Automation1998;32:397–413.

[44] Hax AC, Meal HC. Hierarchical integration of productionplanning and scheduling. In: Giesler MA, editor. Logistics(studies in management sciences). Amsterdam: Elsevier, North-Holland; 1975.

[45] Bitran GR, Haas EA, Hax AC. Hierarchical productionplanning: a two stage system. Operations Research 1982;30:232–51.

[46] Bitran GR, Tirupati D. Hierarchical production planning. In:Graves SC, et al. editor. Logistics of production and inventory.Handbooks in Operations Research and Management Science,vol. 4. Amsterdam: North-Holland; 1993. p. 523–68 [Chapter10].

[47] Bertrand JWM, Wortmann JC, Wijngaard JC. Productionplanning and control: a structured and design oriented approach.Amsterdam: Elsevier; 1990.

[48] Hax AC, Caneda D. Production and inventory management.Englewood Cliffs: Prentice-Hall; 1984.

[49] Vollmann TE, Berry WL, Whybarck DC. Manufacturingplanning and control systems. 4th ed, New York:Irwin/McGraw-Hill; 1997.

[50] Zijm WHM. Towards intelligent manufacturing planning andcontrol systems. OR Spektrum 2000;22:313–45.

E.W. Hans et al. / Omega 35 (2007) 563–577 577

[51] Hans EW. Resource loading by branch-and-price techniques.PhD thesis, University of Twente, Enschede; 2001.

[52] Kolisch R. Make-to-order assembly management. Berlin:Springer; 2001.

[53] Gademann AJRM, Schutten JMJ. Linear-programming-basedheuristics for project capacity planning. IIE Transactions2005;37(2):153–65.

[54] Kis T. A branch-and-cut algorithm for scheduling of projectswith variable-intensity activities. Mathematical programming2005;103(3):515–39.

[55] Elmaghraby SE. Contribution to the round table discussionon new challenges in project scheduling. Paper presented atthe Eighth Workshop on Project Management and Scheduling(PMS 2002), Valencia, Spain, April 3–5; 2002.

[56] Wullink G, Gademann AJRM, Hans EW, van Harten A.Scenario-based approach for flexible resource loading underuncertainty. International Journal of Production Research2005;42(24):5079–98.

[57] Kurtulus I, Davis EW. Multi-project scheduling: categorizationof heuristic rules performance. Management Science1982;28(2):161–72.

[58] Kurtulus I. Multi-project scheduling: analysis of schedulingstrategies under unequal delay penalties. Journal of OperationsManagement 1985;5(3):291–303.

[59] Kurtulus I, Narula SC. Multi-project scheduling: analysis ofproject performance. IIE Transactions 1985;17(1):58–66.

[60] Lova A, Tormos P. Combining random sampling andbackward–forward heuristics for resource-constrained multi-project scheduling. Proceedings of the Eight InternationalWorkshop on Project Management and Scheduling, Valencia,April 3–5; 2002. p. 244–248.

[61] Dumonds J, Mabert VA. Evaluating project scheduling anddue date assignment procedures: an experimental analysis.Management Science 1988;34:101–18.

[62] Dumond J. In a multi-resource environment, how muchis enough. International Journal of Production Research1992;30(2):395–410.

[63] Bock DB, Patterson JH. A comparison of due date setting,resource assignment, and job preemption heuristics forthe multi-project scheduling problem. Decision Sciences1990;21:387–402.

[64] Lawrence SR, Morton TE. Resource-constrained multi-projectscheduling with tardy costs: comparing myopic bottleneck andresource pricing heuristics. European Journal of OperationalResearch 1993;64:168–87.

[65] Morton TE, Pentico DW. Heuristic scheduling systems—withapplications to production systems and project management.New York: Wiley; 1993.

[66] Krajewski LJ, Ritzman LP. Dissagregation in manufacturingand service organisations: survey of problems and research.Decision Sciences 1977;8:1–18.