SawtoothSoftwareRESEARCHPAPERSERIESUsingConjointAnalysisinPricingStudies:IsOnePriceVariableEnough?RichardM.Johnson,SawtoothSoftware,Inc.andKathleenA.Olberts,ChevronChemicalCompany1996©Copyright1996-2002,SawtoothSoftware,Inc.530W.FirSt.Sequim,WA98382(360)681-2300www.sawtoothsoftware.comHowever,it'sconceivablethatthe"shape"ofthepartworthscouldbedifferentforeachbrand.Forexample,whenconsideringBrandA,arespondentcouldbeindifferenttoprice;whenconsideringBrandB,pricecouldbeveryimportant;andtherespondentmightbemorelikelytochooseBrandCatahigherratherthanalowerprice!INCLUDINGINTERACTIONSOfcourse,interactionscanbehandledinconjointanalysis.FinkbeinerandLim(1991)discussedthatsubjectattherecentSawtoothConference.Anotherapproachistousecompoundattributes;onecouldcombinebrandandpricelikethis:BrandAat$1.39BrandAat$1.69BrandAat$1.99BrandBat$1.39BrandBat$1.69BrandBat$1.99BrandCat$1.39BrandCat$1.69BrandCat$1.99(Thiscanonlybedoneifthereareotherattributesagainstwhichtocomparethecompoundattribute.)Sinceninepartworthswouldbedetermined,theycouldassumewhatevershapemightberequiredtoaccountforthedata.ThisapproachisfacilitatedbySawtoothSoftware'sCVASystem.However,therearereasonstobewaryofthisapproach:Many-leveledattributescanresultininterviewsthataretaxingfortherespondent,andiftherearemorethanaveryfewattributesofthiscomplexity,thedataarelikelytobeunusable.TheevidencereportedbyDickWittinkandhiscoauthorsatthisconferenceshowsthatwhensomeattributeshavefewlevelsandothershavemany,themany-leveledattributesarelikelytoemergeasseemingmoreimportantthantheyshould.attributes;oaes;r/P<>>6�/P<<�wlire�i6.63kocany-leveledattr.165TD�[(discussedtha22�>t/P<<yysuneass780.7ewlire�i6.63koi0009eOgtN>t/Po.yaretattruTwJ0.�/MC�/P<5(p)0�f�Tm�(1.39)Tt<dit6rTeOgtN>1rdia)-riC0eergeasseeminy-leveledattr.165TD1�i6Tc�eed,pws522.6v12�ily-leveAdt1ibutesdc>�BDtesdc>�BDtesdc>�BDtesdb.itN>1aretgtiT5ony-ia22�>t/P<<y>1ar0eattributesAMETHODFROMTHE1970'SIn1972,MarketFacts,Inc.publishedapaperbyoneoftheauthorsdescribingatechniqueformeasuringrespondentsensitivitytoprice.ItwasalsodescribedmoreaccessiblybyFrankJonesinJournalofMarketing(1975).Thedatacollectionandanalysisprocedureswereextremelysimple.Considerfourbrands,eachwithfourpossibleprices.Justtoemphasizethatthepricesneedhavenothingincommon,brandsAandBareshownwithlowpricesandbrandsCandDwithhigherprices.Forexample,AandBcouldbeinsix-packs,andCandDin12-packs.BrandA$1.25$1.50$1.75$2.00BrandB$1.33$1.67$2.00$2.33BrandC$2.50$2.99$3.99$4.99BrandD$3.45$4.56$5.67$6.78Ifwehadarespondent'srankorderofpreferenceforthese16combinationsofbrandandprice,thenwecouldpredictwhichheshouldchoose,givenanycompetitivesetofthem.Thedatacollectionprocedurewastoshowtherespondentfourstacksofcards,onestackforeachbrand.Thetopcardoneachstack,andtheonlyonevisible,showedthatbrand'slowestprice,asfoundinthefirstrowofthetableabove.Therespondentwasaskedwhichhewouldpreferfromamongthefour,ifeachbrandwerepricedasshown.Whentherespondentchoseabrand,thatcardwasremovedfromitsstack,andthenextprice"poppedup."Therespondentwasthenaskedasimilarquestionforthemodifiedarrayofproductconcepts,oneofwhosepriceswasnowhigher.Thisprocesscouldcontinueuntilallcardshadbeenselected,orcouldterminateiftherespondentdeclaredhewouldnolongerpurchaseanyoftheproducts,orifallcardsforthebrandsofinteresthadbeenselected.Therespondentmighthavetomakeasm...
