forecasting the future for profit - milner strategic · pdf filereaction to market...
Post on 02-Mar-2018
214 Views
Preview:
TRANSCRIPT
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
top related