Forecasting the Future for ProfitUsing forecast models to better understand your market and outperform the competition

1

l Sourcesofmarketdataarewidelyavailableandvaried,butnoneoftheminisolationcanprovideafullpictureofacompany’smarketbothnowandmovingforwards.Amarketmodelmapsouttheinteractionsofmultiplestructuralelementsinthemarket(suchasprices,products,geographies,demographicsetc.)

l Timeseriesmodels(e.g.Box-JenkinsorHolt-Winters)areausefultoolforshort-termforecastingandunderstandingbehaviourinmaturemarkets.However,inafast-growingtechnologymarket,atime-seriesforecastdoesnottakeintoaccountmarketadoptionbehaviourordemographicdatatoshapeandlimitgrowth.Thisultimatelyrestrictsitsaccuracy

l Technologyadoptionhasbeenobservedtofollowlogisticfunction(orpenetration)curvesacrossmanydifferenttechnologiesincludingcellularphones,personalcomputers,broadbandandtheInternet.Combiningmarketvolumetricdatawithdemographics,churnratesandDiffusionofInnovationTheoryoranchormetricsproducesaverygoodmodellingapproachfordiversemarketsacrossthetechnologysectors

l Sensitivityanalysiscanbecarriedoutonamodelinordertoassesstheimpactofvariablechangesontheshapeofthefuturemarket.Thistechniquedetermineswhateffectasmallchangeinanindependentvariablewillhaveonthedependentvariables.Scenarioplanningisamoreinvolvedandpowerfulwayoflookingatfuturepossibilitiesinthemarket.Itusesanunderstandingofwhichmarketdriversarethemostimportantintermsofmarketimpactandalsouncertaintyofoutcome.Byexploringhowdriversmightinteract,arangeofpossiblefuturescanbeexploredandthemostprobableidentified

l Buildingasinglemarketmodelwiththe

purposeofusingitacrossalldepartmentswithinacompanyproducesstrategicandeconomicbenefits.Thedevelopmentofasingleorganisation-widemodelmeanscompanydiscussionscanfocuson“whatarewegoingtodotowininthemarket?”ratherthan“whatishappeninginthemarket?”Thebenefitsofasinglemodelincludecostreduction,timesaving,theabilitytouseexpertisenoteffort,thepotentialtoderivefirst-moveradvantage,pre-plannedreactiontomarketfluctuationsandcleareremployeeandmarketcommunication

Executive Summary

If you really want to understand your market, you need a model. Over the past 30 years, the collection and availability of market data has grown rapidly thanks to technology and, particularly, the internet. As a result the complexity of the information available about the market has grown substantially. However, having ‘Big Data’ and other information about the market is not the same as understanding the market. To truly understand a market and its dynamics, a company needs to know how all of the structural elements interact, both at present and in the future.

This white paper includes these key points:

2

Therehavebeenanumberoffamousforecaststhatprovedtobewrong.In1895,thePresidentoftheRoyalSociety,LordKelvinsaid“heavierthanairflyingmachinesareimpossible.”Eightyearslaterin1903,theWrightBrothersflewtheirfirstaeroplaneatKittyHawk,USA.On16thOctober1987,theBBC’schiefweatherforecaster,MichaelFish,saidtherewasnoimminenthurricane.Thefollowingday,Englandwasravagedbydestructivewindssoseverethatithasbecomeknownsimplyasthe“TheGreatStorm.”Withexampleslikethis,isitreallypossibletoforecastthefuturewithanycertainty?

Usingtherightmarketintelligence,Kelvinmighthaveknownaboutearlyflyingtrialswithkitesandproto-aeroplanes.Theproblemoflifthadbeensolved(withtheaerofoil)andtheremainingissueswereoptimisingdesignandpropulsion.Hadtheweatherforecastershadthemodelsandcomputingpoweroftoday,“TheGreatStorm”of1987wouldhavebeenforecasted(aswasthe2013UKstorm).Thereisagrowingbodyofevidencetoshowthatusingcomputerstocombinedatawithgoodmodelscanprovideforecastsbeyondanythingthatcanbemadepurelyqualitatively.

Thebook“Moneyball”(Lewis,2003)describeshowtheOaklandAthleticsbaseballteamusedananalytical,evidence-basedapproachtoassembleacompetitiveteam,despiteitsdisadvantagedrevenuesituation.Usingdataaboutplayers’activityduringthegame,Oaklandsignedseeminglylowvalueplayersbutthenoutperformedhighwage-billteamsandmadetheUSbaseballplayoffsin2002and2003.Thebank,JPMorgan,isequallycreditedwithandblamedfortheinventionofcreditderivativeswhichlayattheheartofthebankingcrashin2008.Acreditderivativeisacontractbetweentwopartiesinwhichtheselleragreestocompensatethebuyerifaloangoesintodefault.“Fool’sGold”(Tett,2009)describeshowJPMorgan’smodelsquestionedthecreditproductsandsalesstrategiesofcompetitorbanksfrom2005.Thecombinationoftheirmodelsandmarketexperienceledthecompanytochangeitsposition.TheresultwasJPMorganemergedfromthecrashastheworld’slargestcommercialbankbymarketcapitalisation.

Mapping the future and choosing the routeAmarketmodelisamathematicalrepresentationofthemarket,constructedfromhistoricdataaboutelementssuchasvendors,pricepoints,geographicregionsandtechnologiesovertime.Themodelshowshowthesedifferentelementsinteractwitheachotherovertime.Thisisthenusedtoforecastfuturetrendsintheseelementsandtheconsequencesforthemarket.

This insight into the present and the future enables companies to position themselves more effectively in the market, targeting the segments that are growing or have high growth potential and giving themselves a competitive edge.

The modelling processTobuildamodel,acompanyshouldstartwithabaseofreliablehistoricinformation.Thetechnologiesandcompetitorsinthemarketshouldbereviewedandusedforbenchmarking.Thestructuralelementstobeincludedinthemodel(suchaspricebands,products/services,geographies,customersegments,competitors)needtobedecidedonandbuiltininthecorrectorder.

Toforecast,statisticaltechniquesareappliedtothecollecteddatatocreateaninitialforecastline.Thiscanbecreatedusingregressionanalysisandtimeseriesanalysis(includingBox-JenkinsandHolt-Wintersmethods).However,theresultsfromstatisticaltechniquesneedtobealteredtotakeaccountoftheunderlyingmarkettrends.Theartofmodellingliesinunderstandingmarketbehaviour,applyingtheoriessuchasDiffusionofInnovationandconsideringtheimpactofregionalmarketdynamics.Markettheoriessuchaspenetrationcurves,marketlifecyclesanddisruptionoftechnologyshouldbeconsideredandintegratedintotheforecast.Thiswillensurethattheforecastincorporatesthelonger-termmarkettrends.Thecompany’sregionalmarketintelligenceandon-the-groundknowledgeofthemarketshouldbetakenaccountofintheforecast,toensurethemodelisfullyalignedwiththecompany’sownviewofthemarketdynamicsanddirection.

Introduction

3

Themodelcanberefinedovertimeasnewdatabecomesavailable,meaningthecompanycantestouthypothesesaboutthemarket.Measuringforecastaccuracyenablesabetterunderstandingofthemarket,asthecompanycanidentifyhowandwhythemarketbehaviourdifferedfromtheirassumptions(if,indeed,itdid).Techniquessuchassensitivityanalysisandscenarioplanningcanbeappliedtotestouttheconsequencesofpotentialmajorchangesinthemarket,andthereforethemost-probablefuturecanbedecidedonandmodelled.Thesetechniquesalsoallowacompanytopre-plantheirresponseto

differentfuturedevelopments–bothpositiveandnegative.

The modelling process creates one view that can be subscribed to by the whole organisation. This ensures that all areas of the company are using the same information about volumes, prices, values by sector or segment, the company’s position within the market and where it is heading. As a result strategic planning will be aligned across departments and the company will move in the same direction as one.

Figure 1 Insightsthatcanbederivedfromthemarketmodel

Use of technologies

Geographic focus

Product focus

Market share

Price strategy by geography

Aligned to future technology trends

Regional differences

Behaviour by segment

Consumer Behaviour

Competitor Strategy

Strengths and Weaknesses

Share by geography

Share by price category

Market size and growth rate

Changing preferences

over timePrice sensitivity

4

The Role of the Market Model in Strategy and Planning

Strategising with a modelAmarketmodelprovidesinsightsthatareessentialforstrategyandplanning.Mappingmarkettrendsandpositionsofothervendorsallowsbusinessestounderstandtheopportunitiesforgrowthorexpansionandreducestheriskofrunningintounexpectedobstaclesorgivingothersacompetitiveadvantage.

Themarketmodelwillshowmultipleinteractionsbetweenstructuralelements.Examplesinclude:whichcountriesandregionswillgrow,andwheregrowthwillsloworevendecline;howtheaveragesellingpricewillalterandhowthiswillbedifferentfordifferingmarkets;behaviourofdifferentcustomersegments;competitorbehaviour.

Strategising without a modelItispossibletodevelopcorporatestrategywithoutamarketmodel,butthereareproblemswiththeseapproachestostrategy.

Qualitative view: Purelyqualitativedescriptionsofmarketbehaviourrarelycapturealltherelevantinformation.Evenwhenadescriptionisverythorough,itcannotcapturetheinteractionsbetweenvariablesinthewayamarketmodelcan.Apurelyqualitativeviewwillthereforenotproducesucharobustforecast.

Historic view: Whilsthistoricdataprovidesausefulstartingpoint,itisnotalwaysagoodpredictorofthefuture.Lookingatpastinformationdoesnotprovideanysignificantunderstandingaboutthefuturedirectionofthemarket;thisrequiresadifferentmethodology.

Internal view: Acompanymaychooseitsstrategybasedoninternalinformation,suchasrecentsalesorbrandtracking,andthereforesettargetssuchas‘thepreviousquarter’ssales+10%’or‘launchingproductXinthenextyear’.Thisfailstotakeaccountofchangesinthemarket’sgrowthrateortrendstowardscertainproducttypesorfeatures.

Figure 2 Sourcesofmarketdata

Bespoke

Generic

Internal External

Bespoke

Generic

Internal External

Historic (primaryandsecondaryresearch)

Forecasting

Consumer intention surveys

Analyst forecasts

Economic data

Demographics

Govt. policy

Brand trackingSales figures

Price trends

Competitor financials

Analyst reports

5

Gaining market insightTherearevariousmethodsofcollectingrelevantmarketdata,eachofwhichaddressaspecificaspectofabusiness’futureperformance.Figure2showssomemethodsandtheviewsofthemarketthattheyprovide.However,inisolationnoneofthemgiveafullpictureofthelong-termmarketandshouldnot,therefore,beusedfordevelopingstrategy.

Internalforecastingdoesnottakeaccountofwhereothercompaniesarepositioningthemselvesinthemarketplace,orthewidermarketdynamicsandtrends.Primarymarketresearch,suchasconsumerpanels,onlyprovidesaviewofbehaviourinthenearfuture.

Externalinformationfromanalystreportsandmodelsislackinginthreeareas.Firstly;companiesarerestrictedbytheanalyst’sviewofthefuture,whichmaydifferfromthecompany’sownoutlook.Thisisespeciallyimportantbecauseanindividualcompany’sstrategywillaffectfuturemarketdynamics.Secondly;

usingananalyst’smodeltoformulatetheirowngrowthplansgivescompaniesnostrategicadvantageovertheircompetition,whoareabletoaccessthesamereferencematerial.Thirdly;analystreportsrarelymatchexactgeographicsplits,pricingassumptionsorportfoliospreads.

Drawing all the insights togetherAmarketmodelcombinesallavailableinsightstoshowafullpictureofmarketbehaviour.Itwilltakeintoaccountthequalitativeviewofwhatwillhappeninthemarketplaceandbalanceitagainstthequantitativeinformationtoensurethatthemodelisasaccurateaspossible.Alltheinsightsdata(e.g.internalsalesdata,marketresearch,competitorinformation)acompanyholdsshouldbeharnessedtoensureanaccurateunderstandingofthe‘now’,coupledwithappropriateforecastingtechniquestodocumentthefuture.Usingthis,acompanycanformulatetheirstrategyandtestoutwhichapproachwillbemosteffectiveintheirmarket.

% U

S H

ouse

hold

s

Figure 3 ConsumerproductS-curvesSource: North (2012)

Consumption spreads faster today

1900 1915 1930 1945 1960 1975 1990 2005

100

90

80

70

60

50

40

30

20

10

0

6

Types of marketThemarketbehavesdifferentlydependingonitsstageofmaturity.Broadlyspeaking,therearethreetypesofmarket:growingmarkets,maturemarketsanddecliningmarkets(seeFigure4).

Itshouldbetakenintoconsiderationthatonemarketmightbegrowingandmaturingindifferentregionssimultaneously.Thecorrectforecastingmethodologiesneedtobeadopteddependingonthestageofdevelopmentofthemarket.Furthermore,notallmarketsleavethegrowthstage–somefailtoreachmaturitystageandmovestraighttodecline(Moore,1990).

Market Lifecycle – disruption of technologyThereisacontinuouscycleofinnovationanddisruptionacrossmostmarkets.Thelengthoftimethatamarketremainsinmaturitybeforedecliningdependsonthelevelofinnovationpresentinorpertinenttothemarket.“Incumbentslosetheirmarketleadership(i.e.dominantmarketshare)whenfacedwithdisruptivetechnologicalchange”(Danneels,2004).Thisoccurswhencustomersofamaturemarketstopbuyingitsproductsandstarttoadopttheproductsfromanew,substitutemarketinstead.Theshiftfromfilmtodigitalcamerasisonesuchexample.

Figure 4MarketLifecycleSource: Johnson et al(2008)

6

Types of marketThe market behaves differently depending on its stage of maturity. Broadly speaking, there are three types of market: growing markets, mature markets and declining markets (see Figure 4).

It should be taken into consideration that one market might be growing and maturing in different regions simultaneously. The correct forecasting methodologies need to be adopted depending on the stage of development of the market. Furthermore, not all markets leave the growth stage – some fail to reach maturity stage and move straight to decline (Moore (1990)).

Market life cycle – disruption of technologyThere is a continuous cycle of innovation and disruption across most markets. The length of time that a market remains in maturity before declining depends on the level of innovation present in or pertinent to the market. “Incumbents lose their market leadership (i.e. dominant market share) when faced with disruptive technological change.” (Danneels (2004)) This occurs when customers of a mature market stop buying its products and start to adopt the products from a new, substitute market instead. The shift from film to digital cameras is one such example.

Figure 4 Market Life Cycle Source: Johnson, G., Scholes, K. and Whittington, R. (2008)

Development and Growth Maturity Decline

Mar

ket S

ize

Growth rates: Fast Slow/stable Negative

Sales volumes: Lower Higher Declining

Sales source: New/first Churn/repeat Churn/repeat time adopters customers customers

Level of innovation: High Low Negligible

Price pressure: Medium High Very high

Penetration curvesMostmarketgrowthfollowstheformofapenetrationcurve(alsocalledalogisticfunctioncurveorS-curve).Figure3showshowtheseareconsistentacrossarangeofproducts.Intheearlystagesofpenetrationadoptionisslow,asmanyareunawareoftheproductorunabletoaffordit.Asmorepeoplebecomeawareofthenewtechnology,theadoptionrateacceleratesascustomerschoosetomaketheirfirstpurchase.Whenthemarketreaches50%penetrationtheadoptionrateslows,astheproductiswidelyknownbuttherearefewerpeopleleftinthemarkettopurchaseforthefirsttime.Asthemarketapproachesfullpenetration(wherethemajorityofthetargetpopulationhasboughttheproduct)therateofgrowthslowsandthemarketismaintainedbyrepeatpurchases.

Inrealitythesesmoothpenetrationcurvesmaybedisruptedbyexternalinfluencesthattheforecastershouldtakeaccountof.Forexample,aswecanseeinFigure3,duringstronggrowthintheadoptionof‘autos’therewasaperiodwhenpenetrationdeclined.ThiswasduetotheGreatDepressionfollowedbyWorldWarII,whenmetalandfuelwereneededforthewareffortandthesaleofnewcarswasbanned.

Building the Market Model: Market Theory

7

Forecasting approachesThereareanumberofstatisticaltechniquesthatmaybeselectedtoanalyse‘BigData’andforecastfuturetrends,whereappropriatetothestageofmarketdevelopment.

Time series models:Thetimeseriesmethodsinmarketmodellinganalysehistoricdata(foranumberofmonths/quarters/years)toidentifyseasonalandcyclicalfluctuations,orlong-termunderlyingtrendsthatcanbeextrapolatedtoprovideaforecastoffutureconsumerbehaviour.Inparticular,ARIMAmodels(e.g.Box-JenkinsorHolt-Winters)findthebestfitofpastvaluesofatimeseriesanduseseasonalmovingaveragesinordertomakeforecasts.

Timeseriesmodels areausefultoolforshort-termforecastingandunderstandingbehaviourinmaturemarkets.However,inagrowingmarket,atime-seriesforecastdoesnottakeintoaccountconsumeradoptionbehaviourorthewayinwhichdemographicdatashapesandlimitsgrowth,whichlimitsitsrelevance.

Regression analysis:Multivariateregressionanalysisisatoolthatexamineshistoricdatatodeterminetheindividualeffectsofdifferentdriversofconsumerresponse.Itexaminesthepasteffectsofarangeofindependentvariables(suchasprice,advertising,etc.)onadependentvariable(suchastotalsales)todeterminewhichfactorshavesignificantimpactandthemagnitudeofthese.

Usingthistechnique,aforecastlineiscreatedbysimultaneouslycorrelatingtheindependentvariablesagainstthechosendependentvariable(e.g.sales),toidentifywhichfactorsarestatisticaldriversofthedependentvariable.Thiscanbedescribedusingaformula:

y=c+β1x1+β2x2+β3x3+β4x4+β5x5+…+βnxn+u

wherey=salesandxn=includesvariablessuchasprice,distribution,advertising,promotion,competitoradvertising,weatheretc.

ThelinethatcreatestheOrdinaryLeastSquares(OLS)ischosenbecauseitprovidesthebestfittopastmarketbehaviour.

Inprinciplethisisausefulapproach,howeveritisnotfrequentlyusedformodellingmarkets,principallybecausethegranularityofdataforkeydrivers(suchasmarketingspendetc.)israrelyavailableforallcompetitorsandalsobecausethepotentialforerrorisgreaterusingthisstatisticalapproach.

Forecasting hazardsWhenforecastingitisnecessarytobeawareofpotentialpitfallsinstatisticaltechniques.Thisiswhereaforecastermustapplytheirownskillandexperience.Someoftheseare:

1. Overfitting: Complexequationscanbedevelopedthatverycloselydescribehistoricmarketbehaviourandhaveveryhighcorrelationwiththeactuals(i.e.R2scoresofover0.9).Thereisariskhowever,thattheequationcanbeoverlydependenton,andgivetoomuchweightto,short-termmarketfluctuationsratherthantheunderlyingtrends.Thisproblemisknownas‘overfitting’.InFigure5,theoverfittedforecast(aregressionanalysisbasedonOLSforperiods1-20)trackedthehistoricvariationswellandforecastasharpmarketdropforperiod21.Theactualoutcomeforperiod21wasgrowth.Theproblemwasthatthestatisticalmodelsimplyusedthebestfitthroughthehistoricdatatoforecastthefutureandthisoverfittedmodelfailedtotakeintoaccounttheunderlyingtake-offgrowth.

Building the Market Model: Model Theory

Figure 5Overfittingdoesnotforecastthefuture

0

0.5

1

1.5

2

2.5

3

3.5

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

Historic sales (periods 1-20)

Sales period 21

Overfitted forecast

y = -4E-06x6 + 0.0002x5 - 0.004x4 + 0.0341x3 - 0.1171x2 + 0.1614x + 0.8162

8

2. Correlation vs causation:Apparentcorrelationbetweentwovariablesdoesnotmeanthatthereisacausallinkbetweenthem.Forinstance,inthetabletPCmarketintheUKtheretendstobeaspikeinsalesinQ4whentemperaturesarelowerandpeoplespendmoretimeindoors.SoisweatherapredictoroftabletsalesoristhiscorrelationdrivenbyChristmasgiftingandtheautumnreleaseofnewmodels?Temperaturecouldhaveaneffectontabletsalesbutashortinvestigationwilldemonstratewhetheritiscausalorsimplycorrelated.

3. Overreliance on demographic data: Over-relianceoneconomicorpopulationdatacouldleadtomisjudgementofconsumerbehaviour.Forinstance,whenforecastingthetake-upofphotovoltaiccellsintheUK,relyingondemographicdataaboutnumberofhouseholdswillunderestimatethedemandforthecells.Thisisduetogovernmentsubsidiesandactivitiesthathavealteredthespeedofadoption.

Overlaying art to the science of modellingStatistics-basedmodelscanprovideaccurateshort-termforecastsinmaturemarkets.However,researchintechnologymarketsshowsthatpurelystatistical(quantitative)modellingtechniquescannotbereliedupontoprovideaccurateforecastswithouttheintroductionofothertechniques.Onereasonforthisisthatadoptionratesofnewtechnologycangrowanddeclinemorequicklythanstatisticaltechniquescanrespondto.Whatisrequiredisanunderstandingofthecausesofgrowthatdifferentstagesofamarketlifecycle.

OneveryusefultoolisDiffusionofInnovationtheory,whichallowsthemarketmodellertocombinedemographicdataaboutconsumersinamarketwithhistoricadoptionandpenetrationlevelstounderstandshortandlongertermmarketgrowthpatterns.

Diffusion of Innovation: TheDiffusionofInnovationtheory(Rogers,2003)claimsthatadoptionisdrivenbysocialinfluencesthataffectvariousconsumers(inB2Cmarkets)andenterprises(inB2Bmarket)withorwithouttheirexplicitknowledge.

Rogersdescribesfivegroupsofadopterswithadifferentattitudetowardsinnovation.Innovatorsarethefirsttoadoptnewtechnologyandrepresent2.5%ofthetotalmarket.Thisgroupiswillingtoriskfailureofanewproductbecausetheybelieveithas,orwillhave,ahighutility.EarlyAdopterscomprisealargerproportion(13.5%)ofthemarketandunderstandtheutilityofnewtechnologybetterbecauseoftheexperiencesoftheInnovators.Innovationcontinuestodiffusethroughsubsequentgroupsinthepopulation(theEarlyMajority(34%),theLateMajority(34%),andtheLaggards(16%))witheachgrouptakingitsleadfromthepreviousgroup,andfromthechangessuppliersmaketoproductstomakethemattractivetothetotalpopulation.Figure6illustrateshowprogressthroughthesestagessignificantlyaffectsthemarketgrowthrate.Thegroupshavedifferentcharacteristics,sodifferentmarketapproachesareneededforeach.

Figure 6 DiffusionofInnovationSource: Rogers (2003)

0

20

40

60

80

100 Number of new adopters

Penetration level

Pen

etra

tion

leve

l %

Num

ber

of n

ew a

dopt

ers

Innovators (2.5%)

Early Adopters (13.5%)

Early Majority (34%)

Late Majority (34%)

Laggards (16%)

9

DiffusionofInnovationhasbeenobservedtofollowlogisticfunction(orpenetration)curvesacrossmanydifferenttechnologiesincludingradio,television,VCR,cable,refrigerators,dishwasher,electrificationofhouseholds,telephone,cordlessphone,cellularphone,personalcomputers,broadbandandtheInternet(MooreandSimon,1999).CombiningmarketvolumedatawithdemographicsandDiffusionofInnovationTheoryproducesagoodmodellingapproachinthetechnologysectors.

Churn cycle:Intheearlystagesofamarket,allsaleswillbetonewusers.Howeverasamarketbeginstomature,existinguserswillpurchasereplacementproducts.Thiscanbedrivenbyproductfailure,lossordamageorbytechnologyadvancementsmakingtheoriginalpurchaseoutdatedorobsolete.Thereforeinordertoforecasttotalmarketsales,dataneedstobegatheredtomakeahypothesisabouttheupgradeorchurncycleofusers.Thisinformationaboutthemarketchurnrate,expressedasatimeperiod,allowsnewsalesderivedfromDiffusionofInnovationandchurnsales(basedonthenumberofsalesfromahistorictimeperiodthatreachtheendoftheirlife)tobecombinedtocalculatetotalmarketsales.

White noise: Theterm“‘whitenoise’[is]usedtodescribefunctionssoerraticthatnoinformationiscontainedinx(time)aboutthevaluex(time+δ),nomatterhowsmallδis”(RamsayandSilverman,2005).Thismeansthatthevalueoftheindependentvariableintimeperiod1doesnothelpuspredictthevalueofthedependentvariableintimeperiod2,i.e.thereisverylittlecorrelationbetweenthevariables.Aforecastermustusetheirjudgementtodecidewhichindependentvariableshaveasignificanteffectontheforecastandwhichareuncorrelatedandonlyproducewhitenoise.Notallpotentialvariableswillberelevanttotheforecastandtheymaymakeitmoredifficulttoseewhatthemarketdriversare.ForexampleintheUKtabletmarketfrom1992-2012,yearlymeantemperaturemayhaveverylowcorrelationwithannualtabletsalesandshouldthereforenotbeincludedasanexplanatoryvariableintheanalysis.

Anchor metrics: Inmatureandstablemarkets,alterationsinmarketdemandaredrivenbychangesinunderlyingcustomerbehaviourinthetargetsectors.Aneffectivewaytoforecasttotalmarketsalesinthistypeofmarketistoutiliseanchormetrics.Anchormetricsaredatapointsthatprovideinformationabouttheproductionoutputofaparticularsector(e.g.numberofcarsmanufactured),whicharecorrelatedtosectordemandforaproduct(e.g.numberofcarexhaustssold).Amultiplyingfactorneedstobederivedtounderstandtherelationshipbetweentheanchormetric(cars)andtheproduct(exhausts)inordertocalculatetotaldemand.Themultiplyingfactorneedstotakeintoaccountbothnewsales(measuredbytheanchormetric)andchurnedsales(replacementparts/upgrades).Byforecastingchangesinboththeanchormetricandmultiplyingfactor,thefutureproductdemandcanbecalculated.

Regional market intelligence: Ifdifferentregionsinamodelbehavedifferently,theyneedtobemodelledseparately.Thecorrectgranularityofregionneedstobeselectedtoaccuratelyjudgetheeffectsofthedifferingmarketdrivers.Onaninternationalscale,governmentimportrestrictions,distributionproblemsandcurrencyfluctuationscanallhaveamajorshortorlongtermeffectonadoptionandsalesrates.Quantitativetechniquescannotforeseetheseeventsandmarketintelligencemustbeusedtoadapttheforecast.

Itcouldbearguedthatintroducingseeminglysubjectivefactorsincreasesthepotentialforbiastointrudeupontheforecast.Thisargumentisrelevantwhenforecastsareprovidedbyfinancialandbankinginstitutionsorlargeindustryanalystswhomighthaveunderlyinginterestsorprefertoavoidproducingaforecastthatwoulddifferentiatethemtoomuchfromtheestimateofotherconsultinggroups.Theseorganisationsmayoptforamoreconservativeapproachtoforecasting.However,theroleofaskilledforecasteristoproduceaforecastthatisfreeofpersonal,politicalorsocialbias.

10

Defining the marketAcompany’stargetmarketisasectionofitslarger‘addressable’market,whichconsistsofallthesalesofaproductorservicebyallthesuppliers(i.e.thecompanyanditscompetitors)toallthecustomersinadefinedgeographyoveraspecifictimeperiod.Identifyingtoonarrowanicheinthemarketmeansacompanymaymissthewideropportunitiesavailable,lackawarenessofpotentialthreats,andhaveadistortedviewofthecompany’sactualperformancewithinthemarket.Conversely,definingtoowideamarketwillnotenableacompanytoseethefinerdetailofwhattheirtargetmarketisdoing.

Onecompellingreasonforacompanytobuildorcommissionitsownmodelofthemarketistodefineitsmarketmoreprecisely.Often,itisnotpossibletofindorbuyaforecastoftheexacttargetmarketinwhichacompanyisoperating.Forexample,acompanywantingthesmartmeteringvolumesbyUKregionmayonlybeofferedtheUKtotal.AnothercompanywantingthequarterlyvolumesoftheGermansmartphonemarketmayonlybeabletofindtheannualmarketvalue.

Identifying the structural elements to track within the marketThecoreelementsofanymarketmodelaretheproductorservicevolumes.Tocreateausefulmodel,thevolumescanbegroupedintocustomersegments,pricebands,technologygroups,diseasetreatmentcategories,geographiesoranystructuralelement.

Themorefactorsthemodelincludes,thelargerthesizeandcomplexity.Table1takesasimpleexampleofaforecastmodeloftheUKPCmarketwithvolumesdividedintotwopricebands–under£400and£400andover.Ifthetwopricegroupsarethendividedintodesktopsandlaptops,thesizeofthedataisdoubled.IfthedataisthensplitintoEngland,Scotland,WalesandNorthernIrelandthemarketmodelbecomesfourtimeslarger.Furtherelementscanbeaddedtolookatsegments(students,householdsorbusiness),andtheshareof5othercompetitors(PCsuppliers).Ifalloftheabovefactorswereincludedover12quartersthiswouldresultina3,456cellmodel,beforeanysummarisation.

Despite the challenge of building and maintaining a model with lots of elements, the benefit is a forecast that exactly matches a company’s needs over its time frame of interest. It is clear that even with a small number of elements it is not possible to see all the data interactions without a model.

Maximum penetration levelsInmosttechnologymarkets,theoverallgrowthwillfollowthepenetrationcurve.Tostructurethemodel,itisimportanttoworkoutthehistorictrend,themaximumsizeofthemarketatfullpenetrationandthetimeitwilltaketoreachthatlevel.Whethertheproductorserviceisforanewmarketordisruptinganexistingmarket,itisimportanttodefinepreciselywhichpopulationisbeingpenetratedandwhat100%penetrationwilllooklike.Forexample,forasmartTVcompany100%marketpenetrationmightbeallTVownersadoptingasmartTV(whichmaynotbe100%ofthepopulation),whileformobilephones100%penetrationmightinfactbe120%ofthepopulation,assomepeopleownmorethanonedevice.Thepenetrationlevelofthechosenmarketwillsetthesizeofthemarketopportunityavailable,andhelptodeterminetherateofgrowthovertime.

Structuring a Market Model

Structural Elements

PC Market Example

Price bands <£400 and £400+

Products/Services Laptops and desktops

Technologies Touchscreen, number of cores, etc.

Geography England, Ireland, Scotland, Wales

Customer Segments Students, households, business

Competitors PC suppliers

Table 1 Workedexampleofpossibleelementsinaforecast(PCmarket)

11

Refining the Model

11

‘Wildcard’

Scenario

Probable

Future

Getting more value from the modelOnceamodelisconstructed,itprovidesaninitialforecastandexplanationofhowthevariablesinthemarketinteract.However,thereisagreatdealmorevalueandcompetitiveadvantagetobeextractedbyrefiningthemodel.

Agreeing and modelling the most probable futureDifferentassumptionsaboutthetotalmarketatfullpenetrationorthetimetakentoreachthatlevelcanhavesubstantialeffectsonvolumesinagivenquarter.Otherassumptionsaboutpricingorproductavailabilityinagivengeographycanalsochangethemodelforecasts.Themodelshouldforecastthemostprobablefutureforthebusinessanditstargetmarket.Therearetoolstohelpacompanyunderstandandagreeonthemostprobablefuture,suchassensitivityanalysisandscenarioplanning.

Sensitivity analysis – “what ifs”Sensitivityanalysiscanbecarriedoutinordertoassesstheimpactofvariablechangesontheshapeofthefuturemarket.Thisisatechniquethatdetermineswhateffectasmallchangeinanindependentvariablewillhaveonthedependentvariables(e.g.priceeffectsonvolumes,replacementratesonmarketvalue,segmentbehaviouronmarketshare).Inamarketmodel,sensitivityanalysiscanbeusedtoassessthemagnitudeofpossiblechangesinmarketdynamicscausedby(forexample)changesinvendorstrategy,economicslowdown,resourceshortages,etc.

Scenario planningScenarioplanningisamoreinvolvedandpowerfulwayoflookingatfuturepossibilitiesinthemarket.

Scenarioplanningisaformalprocesstoidentifyandunderstandtheinteractionofexternalvariablesandtheireffectsonthebehaviourofthemarket.Itisparticularlyusefulfordescribingtheattributesofthemostprobableoutcome(whichwillbethebasisofthemodel)butitcanalsogenerateaseriesoflesslikelyfuturescenarioswhich,iftheyoccurred,wouldhaveverydifferenteffectsonsales.

Thisactivitywillenableunderstandingofthemarketdriversandassesswhichdriversarethemostimportant,intermsofmarketimpactandalsouncertaintyofoutcome.Byexploringdifferentinteractionsbetweenthesevariables,businessescanstarttounderstandwhatallthepossiblefuturesmightlooklike;thenthemostprobablefuturecanbeagreeduponandmodelled.

ScenarioplanningwasfirstusedinindustrybyRoyalDutchShellinthe1970stohelpunderstand,modelandplanforchangesintheoilmarket(SchoemakerandvanderHeijden,1992).Itiscreditedasoneofthemostimportantfactorsinthecompany’srisefrom5thintheoilmarketto2ndduringatimeofmarketvolatility.

PL

AU

SI

BL

E

(T

RE

ND

S)

PO

SS

IB

LE

(D

EE

P

DR

IV

ER

S)

Past Present

Figure 7 Identifyingpossiblefuturesandtheirlikelihood

Preferable

TIME

12

Forecasting accuracyForecastingaccuracycanbeimprovedbyaddingnewdataasitbecomesavailable.Thescientificmethod(hypothesis,experiment,knowledgeandrefinehypothesis)canbeappliedtomarketmodels(forecast,collectfuturemarketdata,understand,re-forecast).Regularlyupdatingamodelmeanstheforecastercanrefinetheirunderstandingofthemarket.Thismayleadtonewvariablesbeingadded(e.g.movingfromcountrylevelforecasttoindividualsectorforecastswithinthecountry).Measuringforecasting

accuracyenablestheforecastertoidentifywhereandhowthemarketdynamicsdifferedfromtheforecast.Inbothsuccessandfailureitisimportanttoaskquestionsaboutwhythemarketbehavedasitdidandfeedthisbackintothemodel.Thiswillincreasetheaccuracyofthenextperiod’sforecastandhelpacompanytounderstandcurrentandfuturecustomerbehaviourbetter.Newhypothesesaboutconsumeradoptionratesandcompetitorbehaviourcanthenbescientificallytestedinthemarketplace,resultingingreaterongoingmarketunderstanding.

Applications of a Market Model

Theapplicationsofthedatacontainedwithinamarketmodelarenumerousandarerelevantinalldepartmentsofacompany(seeFigure8).

Herethreekeyapplicationsarelookedatindividually:

Example 1: Supporting an investment decisionAdetailedunderstandingofthesizeofthemarketanditsgrowthrate,coupledwithinsightsaboutthestrengthofthecompetition(marketshare),shouldunderpinanyinvestmentdecision.Thisisbecausesizeandgrowtharekeymeasuresofanindustry’sattractiveness(Johnsonetal,2008).Themarketsizeshowstheaggregatepurchasingbehaviouroftargetcustomersinagiventimeframe.Thiscanbemeasuredbythenumberofunitsbought(unitsales)andthetotalmoneyspent(marketvalue).Evaluatingthemarketwillde-risktheinvestmentprocess,aslikelyreturnscanbecalculatedbasedonaquantifiedmarketunderstanding.Thisisrelevantforbothinternalinvestmentandprivateinvestorsorgrantbodies.

Example 2: Developing the product roadmapAmarketforecastmodeldemonstrateshowthebuyingbehaviourofdifferentmarketelementssuchascustomersectors,geographies,pricingsegmentsandsaleschannelswillchangeovertime.Thesemarketinsightscanbeusedtodeveloprigorousandrobustproductroadmapplans.Forinstanceifacompany

wantedtoassessthePCmarket,itcoulddevelopamarketmodelthatforecastsdemandacrossdifferentelements;productforms(Laptops,HybridsorDesktopPCs),screensizesandmemoryrequirements,overafiveyearperiod.Thisdatawouldhelpthecompanymanageitsproductportfolioandplanitsproductdevelopmenttoensureithadtherightproductsavailableforthechangingneedsofcustomers,attherighttime.

Example 3: Identifying new revenue opportunitiesThein-depthbreakdownofchangingcustomerdemandacrossdifferentmarketelements(suchassectors,geographiesandpricingsegments)providesaninvaluabletooltoidentifynewsourcesofrevenue.Oncethemarketsizehasbeenderived,thecompany’smarketsharecanbecalculatedacrossarangeofelements.Thisidentifiesareasofthemarketwherethecompanyisstrongandlikewiseareaswherethereisroomfordevelopment.Itwillalsoidentifysegmentsofthemarketthatareunattractivenow,butwhichwillgrowovertheforecastperiod,presentinganattractiveopportunityinthefuture.Along-termstrategyandshorter-termsalesplancanbedevelopedbasedonthisinformation,whichwillallowinternalresourcestobeassignedtoopportunitieswiththegreatestpotentialreturns.

13

Figure 8Themarketmodelhasmanyusesacrossthecompany

1414

Havingideasaboutthemarketarenotthesameashavingaformalmarketmodelusedacrossalldepartments.Revenueandsalestargetsarebasedinpartonideas(whichmaybepersonalorinformal)aboutwhatthemarketisdoingorabouttodo.Havingonemodelprovidingaunifiedviewaboutthemarketisthebasisforacompanytoderivesignificantstrategicandeconomicbenefits.Thesebenefitsareasvalidforsmallcompaniesasforverylargecompanies.

Reduce costsCostsavingscanbemadebyhavingonemodelofthemarket.Often,thecostofhavingindividualmodelswithinthebudgetforeachdepartmentisnottakenintoconsiderationwhenlookingatthe‘biggerpicture’ofthecompanyfinances.Asinglemodelbuiltwiththeneeds,supportanddataofalldepartmentswillreducecosts.

Save timeMarketdataisneededbydifferentdepartmentsatdifferenttimes.LongtermforecastsmayberequiredforBoardstrategy,InvestorrelationsorR&D,whilstshortertermforecastsareusefulformarketingpromotionsandtemporarypersonnelrecruitmentcontracts.Ifeachplanorbusinesscaserequiresatrustedmarketforecast,theprocessisslowed.Asinglemodelthereforesavestime.

Align the whole companyWhenacompanyusesonemodel,itallowsalignmentfromtoptobottom.AmodelcommissionedbytheMarketingdepartmentcanprovidetheconcretedataandinsightand‘OneTruth’(acompany-wideacceptedview)aboutthemarket.ThiscanbeusedbytheBoard,Finance,Marketing,Sales,ProductManagementandDevelopment,Operations,InvestorRelations,andHRdepartments(SeeFigure8).Thismeanseachinteractionbetweeneachdepartmentcanstartwith“whatarewegoingtodo?”insteadof“whatisgoingon?”Themodelallowsdepartmentalfigurestobecheckedagainstthemodelforfeasibility,asthemodelisbasedonthecompany’sownstrategyandfedintobyalldepartments.

Use the best experts, not best effortsThiswhitepaperoutlinesthecomplexityandknowledgerequiredtobuildamarketforecast.Giventheimportanceoftheforecastanditsability

togeneratefinancialbenefits,companiesshoulduseexpertsinsideoroutsidethecompanyandnotrelyonthebesteffortsofindividualswithindifferentdepartments.

First mover market advantageDiffusionofInnovationtheorydescribesdifferenttypesofconsumerswithdifferentattitudestorisk,productmaturity,productfeatures,priceandotherfactors.Awelldesignedmodelwillhelppredictwhenthemarketisenteringdifferentstagesandallowthecompanytochangesales,marketing,productanddistributionplansformaximumeconomicbenefit.Firstmoveradvantagecancatchcompetitorsoff-guardandincreasemarketshare.

React faster to the unexpectedSomechangesinthemarket(Political,Economic,Social,Technological,EnvironmentalorLegal)arenotunderthecompany’scontrol.Amarketmodelcanbeusedtoforecastthescaleoffutureopportunitiesandthreatsinconsiderabledetail,consideringthepossibleeffectsofthesechanges.Pre-organisingresponseplans(e.g.expectedpricechangesbyacompetitororavailabilityofnewtechnologies)willallowthecompanytoreactquicklyandefficiently.

Investors like market modelsTheinformationprovidedbymarketmodelsisusefulforinvestorsincompaniesofallsizes.Insmall,privately-ownedcompanies,marketmodelsshowthecompanyisthinkingbeyondproductdevelopment,anddemonstratethesizeandscopeoftheirmarketopportunitytoinvestors.Whenprivatecompaniesaresold,informationfromthemarketmodelisveryimportantfortheInformationMemorandum.WhencompaniesfloatorIPOonmainstockmarkets(orlowercapitalmarketssuchasAIMintheUK),datafromthemarketmodelisusedtodescribethetargetmarket,givemarketsizeandshareinformationandotherkeydisclosures.Forpubliclyquotedcompanies,datafromthemarketmodelisusedtoprovidemarketcontexttothecompany’sfinancialperformancebytheInvestorRelationsteam.

Share the output of the modelItisnotnecessaryforallcompanyemployeestodirectlyaccessorseethesensitiveinformationcontainedinthecompletemodel.Permissionsneed

Benefits of the Market Model

15

tobesetonhowmuchindividualscanseeofthemodelandwhohasaccess.However,itisimportanttoallowandencourageemployeeswithusefulmarketinformationtosharetheirintelligenceinatimelywaysothatitcanbeincludedinthenextiterationofthemodeltoimproveforecastingaccuracy.

TheleadershipteamcanuseextractsfromthemodelintheformofPowerPointpresentationstosharewithteamsofemployees.Detailedreports(sometimes

calledDeepDives)onpartofthemodelcanbeproducedtolookatindividualcountriesorsalesterritories,specificcompetitorsorproductsectors.TheseDeepDivereportscanbesharedwithrelevantteamswithoutexposingthedetailofthewholemodel.

Sharingsomeofthedatawithallemployeesandwiththemarket(viainterviews,articlesandpresentations)isnecessary;thisincreasesclarityofunderstandingandhelpstobuildcompanyreputation.

16

It is possible to forecast markets accuratelyThereisgoodevidencetoshowthatitispossibletoforecastmarketswithahighdegreeofaccuracyusingamarketmodel.Inmaturemarketswithhighlevelsofmarketpenetrationandsloworlittlegrowth,thereislotsofdataandprovenmathematicaltechniquestobuildgoodforecasts.Involatileorfastgrowingmarkets,accuratemarketforecastingisstillpossible.Fewcompaniessharetheirforecastsoraccuracymeasurementbecauseitisasourceofcompetitiveadvantage,butauthorswritingaboutpoliticalelections(Silver,2012),sport(Lewis,2003)andbanking(Tett,2009)showthatitispossibletoforecastthefutureforprofit.

Markets are complex and multidimensionalInaworldof‘BigData’itcanbedifficulttoincorporateallformsofinsightintoonecomprehensiveandquantifiedviewofthemarketwithoutusingamodel.Theinteractionsbetweenthemultiplestructuralelementsofthemarketarenotalwaysimmediatelyobvious.Understandinghowelementssuchasproducts/servicesinteractwithprices,technologies,geographies,customerdemographicsandcompetitorsisacomplexprocesssoitisbesttouseamarketmodelforbetterclarity.

Eachadditionalstructuralelementthatisincludedinamodelmultipliesitscomplexity.Forexample,amodelwithtwopricepointsandtwoproductswillmeanthereare4interactionstotrack;mappingtheseagainstfourterritorieswillproduce16interactionstotrack;modellingtwocompetitorsforthesameelementswillproduce48interactionstotrack,andsoon.

A market model is necessary to understand the complexityAmarketmodelistheonlywaytofullymaptheinteractionsbetweenstructuralelementsinthemarket,providingafullpictureofmarketdynamics.Inmuchthesamewayasafive-dimensionalshapeisnear-impossibletovisualise,itisequallydifficulttopictureallofthemarketelementsandinteractionswithoutusingamodel.

Select appropriately from the tools, techniques and modelsInordertoforecastaccurately,aforecasterneedstotakeintoaccountthemarket’sstageofdevelopmentandselectappropriatelyfromtherangeoftools,techniquesandmodelsavailable.Applyingaone-size-fits-allapproachwillnotleadtoanaccurateforecast.

Therearemanystatisticaltechniquesforbuildingaforecastmodel,suchastimeseriesmethods,ARIMAmodels(e.g.Box-JenkinsorHolt-Winters)andmultivariateregressionanalysis(e.g.OrdinaryLeastSquares).Theseneedtobebalancedagainsttoolssuchaspenetrationcurves,marketlifecycles,DiffusionofInnovationtheory,anchormetrics,churnrates,scenarioplanningandsensitivityanalysis.Forecastersneedtousetheirjudgementandmarketintelligencetoensuretheforecastisrealisticanddepictsthemost-probablefuture.

A market model brings significant demonstrable benefitsCompanieshavefoundthatbuildingandusingamarketmodelhasbroughtthemstrategicandeconomicbenefitsthroughouttheirorganisation.

These benefits include: lreducing costsl saving timel aligning the whole companyl using best experts not best effortsl first mover market advantage l planning for the unexpectedl improved visibility for investors

Conclusions

AuthorsJonathan DavenportistheHeadofMarketAnalysisatMilnerStrategicMarketingLtd.HeleadstheMarketAnalysisteam,directingtheconstructionofmarketmodels,forecasting,customeranalysisandcompetitoranalysis.JonathanisaCharteredMarketer,withastrongbusinesstobusinesssalesandmarketingbackground,developedoveranumberofyearsworkingintheenergyandtelecomssectors.

Nick Milner,PhDistheManagingDirectoratMilnerStrategicMarketingLtd.Nickhashadsignificantstrategyandmarketingexperienceacrossarangeofenterprisesfromsmallstart-upstocompanieslistedontheLondonandIndianStockExchanges.Hehas25yearsofexperienceinbuildingmarketmodelsandforecasting.NickisaFellowoftheCharteredInstituteofMarketing,aFreemanoftheWorshipfulCompanyofMarketors,aCharteredMarketerandaCharteredPsychologist.

Kay Sharpington isaMarketAnalystatMilnerStrategicMarketingLtd.AspartoftheMarketAnalysisteamsheassistswithbuildingmarketmodels,forecasting,customeranalysisandcompetitoranalysis.KayisaCambridgeUniversitygraduatewithexperienceinstatisticalanalysis,regressionanalysisandeconometricmodelling.

About Milner Strategic Marketing LtdMilnerStrategicMarketingLtdisabusinessconsultancy.Wedeliverarangeofservicesfromstrategicmanagementconsultancythroughtospecificmarketingprogrammes.OurclientscometousbecausetheyneedhelpwithMarketAnalysis,StrategyFormulationandMarketingProgrammes.

Ourclientsrangeinsizefromsmall,venture-backedstart-upstolarge,quotedinternationalcompanies.ClientsareusuallyB2Bfocussedandcomefromthreetechnologysectors:High-tech(ICT);Clean-tech(Renewables,SmartGridandEnergyRetail);Bio-tech(Bio-MedicalandHealthcare).

MilneroffersanumberofMarketAnalysisservicesincludingcustomeranalysis,competitoranalysisandmarketmodellingandforecasting.Wehaveextensiveexperienceofbuildingmarketmodelsonaglobalscale.InconjunctionwiththisMilnercanalsoprovidestrategyworkshops,productportfolioreviewandproductmanagementtraining,marketingcommunications/PR,collateralredesign,andarangeofdigitalmarketmarketingservices(includingwebdesign,e-newslettersandsocialmedia).

Bibliography Choong, J. (2012)‘Powerful Forecasting With MS Excel’McGraw-Hill

Firth, M. (1977)‘Forecasting Methods in Business and Management’London:EdwardArnold,

Lind, D., Marchal, W. and Mason, R. (2002)‘Statistical Techniques in Business and Economics’Europe:McGraw-Hill

References Danneels, E.(2004).‘Disruptive technology reconsidered: A critique and research agenda’ Journalofproductinnovationmanagement,21(4),pp246-258

Davis, S. (2009).‘The Shift: The Transformation of Today’s Marketers into Tomorrow’s Growth Leaders’SanFrancisco:JosseyBass

Johnson, G., Scholes, K. and Whittington, R.(2008).‘Exploring Corporate Strategy’8thEdition.PearsonEducationLimited

Lewis, M.(2003).‘Moneyball’W.W.Norton&CompanyInc

Moore, G.(1990) ‘Crossing the Chasm’ HarperCollins

Moore, S. and Simon, J. (1999)‘The Greatest Century That Ever Was: 25 Miraculous Trends of the last 100 Years’ TheCatoInstitute:PolicyAnalysis,No.364

North, S. (2012)‘Speed of Adoption Risk’http://theinnovationofrisk.com/speed-of-adoption-risk/

Ramsay, J. and Silverman, B. (2005)‘Functional Data Analysis’ SpringerScience+BusinessMedia,Inc.

Rogers, E. (2003)‘Diffusion of Innovations’5thEdition.NewYork:FreePress

Schoemaker, P.J.H. and van der Heijden, C.A.J.M.(1992)‘Integrating Scenarios into Strategic Planning at Royal Dutch/Shell’ PlanningReview.Vol.20(3),pp.41-46

Silver, N.(2012)‘The Signal and the Noise’ London:PenguinGroup

Tett, G.(2010)‘Fool’s Gold’ Abacus

About

18

Adastral Park

Email: nick.milner@milnerltd.com

+44 (0) 1473 633124

Milner Strategic Marketing LtdThe Ross BuildingAdastral Park Martlesham IP5 3REUnited Kingdom

St John’s Innovation Centre

Email: david.jenkins@milnerltd.com

+44 (0) 7739 758859

Milner Strategic Marketing LtdSt John’s Innovation CentreCowley RoadCambridgeCB4 0WSUnited Kingdom