Consider the types of conjoint analyses discussed in the lecture so far. Compare them in terms of what you believe their advantages and disadvantages to be. Specifically, compare full profile ratings to discrete choice conjoints, and then within discrete choice, compare pairwise and three-option choice tasks. What would you gain by using one over another? What problems would they create?

conjointanalysisisactuallysettingupthedatacollection.Andwhilewe'renotgoingtotalkaboutheactualcollectionofdata,you'regoingtoseethat,fromananalyticalperspective,howthedataweresetupisgoingtodetermineinparthowweanalyzethedata.Sounderstandingthesetuptoconjointanalysisisveryimportantevenifyou'rejustconcernedabouttheanalyticsideofit.

Solet'stalkfirstaboutthedifferenttypesofconjointanalyses,andwe'llgetasenseofhowthosearegoingtoaffecttheanalysesthatweconductalittlebitfurtherondowntheroad.AsImentionedearlier,conjointisreallyagroupofstatisticalproceduresthatkindofallmeldtogetherwithdifferentcharacteristics.Andthenicethingaboutitiswecanusetheanalyticalprocedureswe'vealreadylearnedtoactuallyanalyzethedata.

Butthefactthatwecanusetheseexistingproceduresdoesn'tmeanthatwedon'thavetobeverycarefulintermsofsettingupthedatacollectionandhowwewanttheconjointactuallytobesetupandestablishedanddesigned.Solet'sthinkabouttwoimportantdimensionsabouthowwesetupdataaroundconjointtechniquesandhowthey'regroupedmostlyaroundtodifferentdimensions.

Firstofall,they'regroupedaccordingtohowcombinationsofattributesarepresentedtorespondents.Inotherwords,whatdorespondentsseeasthey'reevaluatingthedifferentattributesthatthecompanyistesting?Sowhatcombinations--orhowarewegoingtopresentthingslikecellphonescreensize,cellphoneprice,andsoforth?

Secondly,howrespondentsindicatetheirpreferences.Andinotherwords,howisitthatrespondentsaregoingtoindicateonwhateverquestionnaireinstrumentwehave,howaretheygoingtoindicatewhatitisthattheypreferthemost?Thereareverydifferentwaysofdoingthisthathaveveryimportantimplicationsforhowweanalyzeandsetupthedatadowntheroad.

Now,ifyou'llrememberbacktoexperimentaldesignthatwetalkedaboutafewweeksago,thereisonewayofcollectingdatathatisparalleltowhatweuseinexperimentation,which iscalledthefullfactorialdesign.Andthatis,allrespondentsseeallcombinations.Aproblemthatwehadwiththatinexperimentaldesignistoomanylevelsoftoomanyfactorscreatetoomanygroupsthataredifficulttomanage.Sowithfullfactorialdesigns,youhavetobecareful.

Thisisespeciallytrueinconjointanalysis.Andthat'sbecause,inconjointanalysis,thenumberofcomparisonspeoplemakecanjustballoontoastronomicalproportions,evenfarbeyondwhatwehavewithexperimentation,andcreatetasksforourrespondentsthattheywouldsimplybeunwillingtodo.Itjustbecomestoomuchworktocollectdata.Sousingfullfactorials,whileit'spreferable,requiresthatwekeepthedesignsfairlysmall.

Ifwedon'twanttouseafullfactorialbecausewearetestingabigdesign,youcanusesomethingcalledafractionalfactorial.Andthisiswherethesoftwareadd-onstomodulesinSPSS,forexample,comeinveryhandy.BecauseSPSS'sconjointprocedurethatyouaddontothebasepackagehasaspecialprogramthathelpsyouputtogetherfractionalfactorialdesigns.

Doingfractionalfactorialdesignsbyhandisactuallyverydifficult.It'sveryinvolvedandverylaborious.Andit'sjustastediousascanbe.Soarealvalueofsomeofthesoftwareadd-ons,particularlyfromcompanieslikeSPSSorSawtooth,isthattheyhaveutilitiesinthemthathelpsyoudesignthesefractionalfactorials.

Becausewe'renotgoingtobeusingthem--andfrankly,Ineverhave--Ipreferthefullfactorialbecauseitforcesmetokeepmydesignssimple.AndIthinkthat,whenitcomestoanalyzingdata,thatbecomesarealbenefit,especiallyforpeoplewhoarereallynotsuperhightechnicalwonkswhoreallywanttodigintotheseverysmallnuancedresults.Sokeepingitsimplebecomesveryimportant.

Thesecondmajoreventdimensionofconjointanalysisishowrespondentsactuallyindicatetheirpreferences.Acoupleofwaysofdoingthis.Oneiscalledthefullprofileconjointanalysiswhererespondentsrespondonascaletohowwelltheythinkaparticularcombinationofattributeswouldmeettheirneeds.

Sooftentimes,it'sascaleof1to10orascaleof1to7forbeing,forexample,apoorvalueforacombinationoragoodvalueofcombinations.Whateveritis,weusearatingscaletogofromlowpreferencetohighpreferenceorlowutilityahighutility,howeveritiswe'regoingtowordit.

Butthenthereisalsosomethingcalledadiscretechoiceconjointanalysis.Andthisisonethatisbecomingincreasinglypreferredbymarketersinlargemeasurebecausewe'regettingthecomputerpowerandsoftwarethatcanhandlethesekindsofthings.Sodiscretechoiceconjointsaskcustomers--ortherespondentsactuallytobehaveasiftheywerecustomers--andchoosebetweenpairwisepresentationsofoptionswithdifferentcharacteristicstothem.

Theproblemwithdiscretechoicesisthatthisiswherethenumbersoftasksthatweaskparticipantstodobecomesveryprohibitive.Sowehavetobecarefulaboutthediscretechoice.Let'stakealookathowsomeofthesearesetup.AndthisisanexampleI'mgoingtousejustalittlebitthroughseveralsetsofslides.Andit'sonewherewe'retalkingaboutelectricgolfcarts.

Solet'sthinkaboutthefeaturesthatanelectricgolfcarthas.Maybeyou'reanavidgolfer.Maybeyoutoodlearoundinagolfcart insomeplacethatyouliveoruseittoworkonlandorpropertyorsomethinglikethat.I'm sureyou'veseenthemaround.Solet'sconsiderthefeaturesofelectricgolfcartsifwe'regoingtobuyone.

Theattributemightbeprice,howmucharewewillingtopay.Well,in aconjointanalysis,we'regoingtowanttosetupdifferentlevelsofprice.Andsowewanttosetupacoupleofpricepointsthatwillproducerealdifferencesinpreferenceorutilitysowecangetanideaofhowimportantpriceis.

Sowemightthensetupaconjointanalysiswherewepresentoptionsforgolfcartsthatwouldbeoneat$2,500andoneat$3,500andseewhatdrivespreferencebasedonsomeoftheotherfeatures.Inotherwords,whatarepeoplewillingtopayfor?Well,anotheroneofthosefeaturesmightbebatterylifecycle.Ifyouknowanythingaboutelectricgolfcarts,youknowthatbatteriescanbevery,veryexpensiveinthem.

Sothelifecycleofthebatteryrelativetothecostofthebatterybecomesprettyimportant.Sodoyouhaveathree-yearlifecycleforthebatterywhereitneedstobereplacedabouteverythreeyears?Ordoyouhaveaseven-yearlifecyclewherethebatteriesneedtobereplacedinalittlebitlongerofalifecycle?Andisalongerlifecycleworththeextra$1,000toyou?Well,thatdepends.Andthat'sthepurposeoftheconjointanalysis.

Andthenwhatabouttheconfigurationofthegolfcart?Doyouwanttohaveitsingleuseormultiuse?Inotherwords,doyouwanttobeabletochangethebackseat andwhetherithasagolfcarrackonitorwhetherithasaflatplatformtocarryingthingsorwhetheryoucanaddanextraseattocarrymorepassengers?Ordoyouwantittobesingleusewhereit'sconfiguredinaparticularwayandyoucan'tuseitforanyotherpurpose?Sowhatareyouwillingtotradeoffofthemoney?Thatbecomesanimportantquestionthatliesattheveryheartofconjointanalysis?

Sohowwouldwegoaboutcollectingdataforsomethinglikethis?Well,ifwe'regoingtodoafullprofileratingoftheconjointanalysisofthegolfcart,whatwewoulddoissetupasituationwherewewouldaskpeopletoevaluatethefollowingconfigurationofthegolfcart.Sowewouldsay,onascaleof1to10,frompoorvaluetoexcellentvalue,whatdotheythinkofamulti-useconfigurationwithathree-yearbatterysetfor$3,500?

Now,priortoactuallyprovidingthemwiththisquestion,youmay,earlyoninthequestionnaireadministration,providethemwithmoredetaileddescriptionsaboutwhatthedifferentthingsyou'remeasuringare.Buteventually,it'sgoingtoboildowntothepresentationofdifferentcombinationsofattributesandaskingthemhowpreferablethosecombinationsareonascaleof1to10or1to7orwhateverithappenstobethatyouuse.

Butinthiscase,allrespondentswouldrateallcombinations.Andbecausewehavethreecombinationswithtwolevelseach,itwouldbetwocubedortwobytwobytwooreightratingsperrespondent.Sowewouldaskrespondentstoratealleightdifferentcombinationsoneaftertheother.

Sometimeswemightmixthemup.Wemightaskoroursurveyplatformtorandomizetheorderinwhichthey'represented.Buttheimportantpointisthatwe'regoingtogettheinformationfromalloftherespondentsforalleighttasks.Now,eighttasksandeightcomparisonsthatwayisnotnecessarilyrealhardintermsofaskingrespondentsfortheircooperation.

Butwhenyoustarteddoingdiscretechoice,itbecomesalittledifferent.Let'stakealook.Let'slookat,firstofall,pairwisediscretechoice.Here'swhatitwouldlooklike.Afterexplainingwhatthedifferentconfigurationoptionswouldmeanandtalkingaboutthemalittlebitintheintroductiontothequestionnaire,whatyoudoisyoustartgivingthempairstoselectfrom.

Whichwouldyouchoose,themulti-useconfigurationthree-yearbatteryfor$3,500bucksorasingle-useconfigurationseven-yearbatteryfor$3,500?Whataretheywillingtotradeoffthere?Andasyoupresentthesedifferentcombinations,thenrespondentscanselect.Andyougetanideaofwhattendstodrivetheirchoices.

Howmanyquestionsrespondentsdodependsonthenumberofattributesandlevels.Andthisistheweaknessofdiscretechoiceconjointanalysis.It'sthataswestartaddinglevels,thenumberoftasksthatrespondentshavetodobymakingeverypossiblecomparisonofeverypossiblesetoffeaturesbeginstoabsolutelyskyrockettoapointtherewherethereisnowaytherespondentswouldevenbegintobeco-operativeforit.

Sowecanalsosetupdiscretechoicetoincludemorethantwooptions.In otherwords,itdoesn'thavetobepairwise.Sowecansay,whichconfigurationwouldyouchoose?Andwecanofferthemthree.Theproblemthereisthat,eventhoughthechoicebecomesmorerealisticbecausewehavemoreoptionsandconfigurationsinfrontofusjustaswewouldinthemarketplace,thenumberofpossiblecombinationsandthenumberofcomparisonsjustbeginstobecomeastronomical.

Andthat'swheresoftwarepackagesthatofferyoutheabilitytocreatefractionalfactorialsbecomeveryuseful.Butforme,thesolutiontothatproblemisreallytokeepthingsassimpleaspossibleandfindthatsweetspotbetweensimplicityandcomplexitythatprovidesyouthedatathatyouneedwhile,atthesametime,notoverburdeningyourrespondentswhoarekindenoughtoprovideyouwithsomeinformation.

Next Card