Download - Conjoint Analysis
Conjoint AnalysisConjoint Analysis
What is Conjoint AnalysisWhat is Conjoint Analysis
**A technique for understanding howA technique for understanding how responding responding
develop preferences for products or servicesdevelop preferences for products or services . .
**Also known as “ trade-off analysisAlso known as “ trade-off analysis ” ”
**Premise : consumers evaluate overall utility by Premise : consumers evaluate overall utility by combining values for each attribute of the combining values for each attribute of the
productproduct . . **Subjective preference judgment unique to each Subjective preference judgment unique to each
individualindividual. * . * Encompasses all product or service Encompasses all product or service featuresfeatures . .
**Products/services whit higher utility are more preferred Products/services whit higher utility are more preferred and have a better chance or choiceand have a better chance or choice..
How is Conjoint Analysis How is Conjoint Analysis DoneDone??
11)) DescDescribe product/service in term of its ribe product/service in term of its attribute/characteristics/features attribute/characteristics/features (factors)(factors)..
22)) Select possible values for each factor Select possible values for each factor (levels)(levels)
33)) Construct a set of products / services Construct a set of products / services ( treatments( treatments or or stimulistimuli) by combining levels of ) by combining levels of each factoreach factor. .
44)) Present stimuli to respondents who provide their Present stimuli to respondents who provide their overall evaluationsoverall evaluations..
55)) Determine preference structure – influence of Determine preference structure – influence of each factor and each level on respondent s utility each factor and each level on respondent s utility judgment – individually and collectivelyjudgment – individually and collectively..
Hatco example: Hatco example: Industrial CleanerIndustrial Cleaner
**33 factor: ingredient , form, brandfactor: ingredient , form, brand..
**22 levels for each factor : phosphate- levels for each factor : phosphate- free/phosphate-based ,liquid /powder, Hatco free/phosphate-based ,liquid /powder, Hatco /generic/generic. .
**22**22**22 = =88 stimuli, e.g.Hatco phosphate- free powderstimuli, e.g.Hatco phosphate- free powder..
Hatco customers asked toHatco customers asked to : :**Either rank-order 8 stimuliEither rank-order 8 stimuli . .
Or rate each stimuli on a 1-10 preference scaleOr rate each stimuli on a 1-10 preference scale..
The additive model for The additive model for individualsindividuals
**utility for any stimuli estimated from utility for any stimuli estimated from part- part- worthsworths..
**Utility for product with level I for factor 1 , Utility for product with level I for factor 1 , level j for factor 2 , … , level n for factor Nlevel j for factor 2 , … , level n for factor N
= =part- worth for level I for factor 1part- worth for level I for factor 1 + +part- worth for level j for factor 2part- worth for level j for factor 2… + … +
+ +part- worth for level n for factor Npart- worth for level n for factor N **For example , Hatco phosphate – free powder For example , Hatco phosphate – free powder
utilityutility = =part worth of Hatco brandpart worth of Hatco brand
+ +part worth of phosphate – free ingredientpart worth of phosphate – free ingredient + +part worth of powder formpart worth of powder form
Empirical ExampleEmpirical Example
Two respondent ranked 8 stimuliTwo respondent ranked 8 stimuli::
stimustimulili
forformm
ingredientingredientBranBrandd
D1 D2 D3 D1 D2 D3 Res1 Res1 Res2Res2
S1S1S2S2
S3S3
S4S4
S5S5
S6S6
S7S7
S8S8
LiquidLiquid
LiquidLiquid
LiquidLiquid
LiquidLiquid
PowdePowderr
PowdePowderr
PowdePowderr
powdepowderr
Phosphate- freePhosphate- free
Phosphate-freePhosphate-free
Phosphate- Phosphate- basedbased
Phosphate- Phosphate- basedbased
Phosphate- freePhosphate- free
Phosphate- freePhosphate- free
Phosphate- Phosphate- basedbased
Phosphate- Phosphate- basedbased
HatcoHatco
GeneriGenericc
HatcoHatco
GeneriGenericc
HatcoHatco
GeneriGenericc
HatcoHatco
GeneriGenericc
11 11 11 11 11
11 11- - 11 22 22
11- - 11 11 55 33
11- - 11- - 11 66 44
--11 11 11 33 77
--11 11- - 11 44 55
--11- - 11 11 77 88
--11- - 11- - 11 88 66
Estimating Part - WorthsEstimating Part - Worths
Calculate average rank of each level of each factorCalculate average rank of each level of each factor::
FactorFactorLevelLevel
Respondent 1 Responding 2Respondent 1 Responding 2
Rank Ave. Rank Ave. Part_Part_ wthwth Ranks Ave. Ranks Ave. Part_Part_ WthWth
FormForm
IngredientIngredient
BrandBrand
LiquidLiquid
PowderPowder
Phosphate-freePhosphate-free
Phosphate-basedPhosphate-based
HatcoHatco
GenericGeneric
1,2,5,61,2,5,6 3.53.5 1.01.0 1,2,3,41,2,3,4 2.502.50 2.002.00
3,4,7,83,4,7,8 5.55.5- - 1.01.0 5,6,7,85,6,7,8 6.506.50- - 2.002.00
1,2,3,41,2,3,4 2.52.5 2.002.00 1,2,5,71,2,5,7 3.753.75 0.750.75
5,6,7,85,6,7,8 6.56.5- - 2.02.0 3,4,6,83,4,6,8 5.255.25- - 0.750.75
1,3,5,71,3,5,7 4.04.0 0.50.5 1,3,7,81,3,7,8 4.754.75- - 0.250.25
2,4,6,82,4,6,8 5.05.0- - 0.50.5 2,4,5,62,4,5,6 4.254.25 0.250.25
Predicting RankingsPredicting Rankings
**Sum the part-worths for each stimuli to find Sum the part-worths for each stimuli to find overall utilitiesoverall utilities..
**Predict rankings based on these utilitiesPredict rankings based on these utilities for example, respondent 1 utilities are 3.5, 2.5,-0.5,-1.5for example, respondent 1 utilities are 3.5, 2.5,-0.5,-1.5
, , 1.5,0.51.5,0.5-,-,2.52.5-,-,3.53.5..
Predicted rankings are therefore 1,2,3,6,3,4,7,8Predicted rankings are therefore 1,2,3,6,3,4,7,8,,
perfect predictionperfect prediction!!
Managerial Uses of Managerial Uses of ConjointConjoint
11 ) )Discover object/concept with optimal qualitiesDiscover object/concept with optimal qualities..
22 ) )Establish relative contributions of each attribute Establish relative contributions of each attribute and each level to utilityand each level to utility..
33 ) )Predict utilities for other stimuli not evaluatedPredict utilities for other stimuli not evaluated. .
44 ) )Identify segments of consumers who put Identify segments of consumers who put differing importance on attributesdiffering importance on attributes..
55 ) )Explore market potential for feature Explore market potential for feature combinations currently unavailablecombinations currently unavailable..
Unique Features of Unique Features of ConjointConjoint
Separate models for predicting preference for each Separate models for predicting preference for each respondent( disaggregate)respondent( disaggregate)..
Individual results can be aggregated to calculate Individual results can be aggregated to calculate group utility alsogroup utility also
Handles nonlinear relationships as well as linear Handles nonlinear relationships as well as linear onesones..
Conjoint Decision Conjoint Decision FrameworkFramework
11 ) )Define objectivesDefine objectives..
22 ) )Develop research designDevelop research design..
33 ) )Evaluate assumptionEvaluate assumption..
44 ) )Estimate model and assess fitEstimate model and assess fit..
55 ) )Interpret resultsInterpret results..
66 ) )Validate resultsValidate results..
Conjoint Decision Conjoint Decision FrameworkFramework
11 ) )Define objectivesDefine objectives. .
22 ) )Develop research designDevelop research design..
33 ) )Evaluate assumptionEvaluate assumption..
44 ) )Estimate model and assess fitEstimate model and assess fit..
55 ) )Interpret resultsInterpret results..
66 ) )Validate resultsValidate results..
Define ObjectivesDefine Objectives
**Determine contributions of factors and their levels Determine contributions of factors and their levels to consumer preferenceto consumer preference..
e.g. how much does price contribute to willingness to buy e.g. how much does price contribute to willingness to buy , and, and
which price is bestwhich price is best??
**Find a valid model of consumer judgmentsFind a valid model of consumer judgments.. valid models enable prediction of preference for any valid models enable prediction of preference for any
combinationcombination of factor/levelsof factor/levels..
**What decision criteria do consumers use to make What decision criteria do consumers use to make choices for this type for products/serviceschoices for this type for products/services??
turn these criteria into attributes that give value to the turn these criteria into attributes that give value to the productproduct//
serviceservice..
Conjoint Decision Conjoint Decision FrameworkFramework
11 ) )Define objectivesDefine objectives..
22 ) )Develop research designDevelop research design..
33 ) )Evaluate assumptionEvaluate assumption..
44 ) )Estimate model and assess fitEstimate model and assess fit..
55 ) )Interpret resultsInterpret results..
66 ) )Validate resultsValidate results..
Develop Research DesignDevelop Research DesignSelect a conjoint methodSelect a conjoint method
tradition, adaptive or choice-basedtradition, adaptive or choice-based..
Define factors and levelsDefine factors and levels specify modelspecify model
additive or interactionadditive or interaction linear, quadratic or separate part-worthslinear, quadratic or separate part-worths
Collect dataCollect data full-profile, pairwise , or trade-off comparisonfull-profile, pairwise , or trade-off comparison
presentationpresentation ranking or rating preferencesranking or rating preferences
survey administrationsurvey administration
Conjoint methodConjoint method
traditionaltraditionaladaptiveadaptiveChoice-Choice-basedbased
MaxMax #. #.
FactorFactor99303066
AnalysisAnalysis
LevelLevelIndividual Individual oror
aggregateaggregate
Individual Individual or or aggregateaggregate
aggregateaggregate
ModelModel
formformadditiveadditiveadditiveadditiveAdditive Additive
or or interactiointeractionn
Example of InteractionExample of InteractionstimulistimuliformformingredientingredientbrandbrandRes 3Res 3S1S1
S2S2
S3S3
S4S4
S5S5
S6S6
S7S7
S8S8
LiquidLiquid
LiquidLiquid
PowderPowder
PowderPowder
LiquidLiquid
LiquidLiquid
PowderPowder
PowderPowder
Phosphate-freePhosphate-free
Phosphate-Phosphate-basedbased
Phosphate-freePhosphate-free
Phosphate-Phosphate-basedbased
Phosphate-freePhosphate-free
Phosphate-Phosphate-basedbased
Phosphate-freePhosphate-free
Phosphate-Phosphate-basedbased
HatcoHatco
HatcoHatco
HatcoHatco
HatcoHatco
GenericGeneric
GenericGeneric
GenericGeneric
GenericGeneric
11
33
22
44
77
55
88
66
Part-worth RelationshipPart-worth RelationshipMost Most LeastLeast
Restrictive Restrictive RestrictiveRestrictive
Most efficient Most efficient least efficientleast efficient
Most Estimation Most Estimation least estimationleast estimation
Linear quadratic or Linear quadratic or separate separate
idea pointidea point part-worth part-worth
Data Data collection :presentationcollection :presentation
methodmethod**Full-profile: liquidFull-profile: liquid phosphate-free rate or rank each phosphate-free rate or rank each
stimulistimuli HatcoHatco
**Pairwise : Liquid vs. powderPairwise : Liquid vs. powder phosphate-free phosphate-basedphosphate-free phosphate-based
**trade-off: liquid powder rank eachtrade-off: liquid powder rank each Hatco ? ? combinationHatco ? ? combination
GenericGeneric? ? ? ?
Conjoint Decision Conjoint Decision FrameworkFramework
11 ) )Define objectivesDefine objectives..
22 ) )Develop research designDevelop research design..
33 ) )Evaluate assumptionEvaluate assumption..
44 ) )Estimate model and assess fitEstimate model and assess fit..
55 ) )Interpret resultsInterpret results..
66 ) )Validate resultsValidate results..
Evaluate AssumptionsEvaluate Assumptions
**Few statistical assumptionFew statistical assumption e.g. normaly , homoscedasticalye.g. normaly , homoscedasticaly, ,
independence check not neededindependence check not needed
* *Strong conceptual assumptionsStrong conceptual assumptions e.g. specify model form( additive vs. interaction) before data e.g. specify model form( additive vs. interaction) before data
areare
collectedcollected..
Estimate ModelEstimate Model * *Rated preferences are analyzed using specializedRated preferences are analyzed using specialized
software ,e .g .SAS proc transreg , Conjoint analyser , orsoftware ,e .g .SAS proc transreg , Conjoint analyser , or SPSS Conjoint add-on or even regression with dummySPSS Conjoint add-on or even regression with dummy
variables( effects coding)variables( effects coding)** Rank-ordered preferencesRank-ordered preferences::
can be analyzed using special procedures designed for can be analyzed using special procedures designed for ordinal dataordinal data
or can be analyzed using procedures designed for metric or can be analyzed using procedures designed for metric data ,e. gdata ,e. g
regression (beware unequally spaced preferences however)regression (beware unequally spaced preferences however)
**specify part-worth relationship (separate ,quadratic ,or specify part-worth relationship (separate ,quadratic ,or linear )for each factorlinear )for each factor..
Interpret ResultInterpret Result
* *Interpretation possible at both individual and Interpretation possible at both individual and aggregate levelsaggregate levels..
* *Consider part-worth estimate for each factorConsider part-worth estimate for each factor:: practical relevancepractical relevance
correspondence to theorycorrespondence to theory plot part-worth (y-axis) vs. (x-axis) to identify plot part-worth (y-axis) vs. (x-axis) to identify
patterns :connect pointspatterns :connect points with lines for each respondent or for aggregate resultwith lines for each respondent or for aggregate result..
if population exhibits homogeneous behavior, aggregate if population exhibits homogeneous behavior, aggregate result canresult can
predict market sharepredict market share..
* *Consider conditional relative importance Consider conditional relative importance (CRI) of factors(CRI) of factors..
Plot Result, e.g. formPlot Result, e.g. form
utilityutility22
1.81.81.61.61.41.41.21.21.01.00.80.80.50.50.40.40.20.20.00.0
--0.20.2--0.40.4--0.50.5--0.60.6--0.70.7--0.80.8--1.01.0--1.21.2--1.41.4--1.61.6--1.81.8--2.02.0
liquid powderliquid powder
Validate ResultsValidate Results
* *InternallyInternally use a pre-test study to confirm which composition rule (additive use a pre-test study to confirm which composition rule (additive
oror
interaction) is appropriateinteraction) is appropriate..
use holdout stimuli to assess predictive accuracy individualuse holdout stimuli to assess predictive accuracy individual..
use holdout sample of respondents to assess predictive accuracyuse holdout sample of respondents to assess predictive accuracy
collectivelycollectively..
* *ExternallyExternally does the analysis predict actual choicesdoes the analysis predict actual choices? ?
how representative is the sample of a populationhow representative is the sample of a population??
Self-Explicated ConjointSelf-Explicated Conjoint
* *Respondent directly rates desirability of each Respondent directly rates desirability of each attribute level and relative attribute importanceattribute level and relative attribute importance..
)+( *)+( *more manageable than traditional conjoint for 10+ more manageable than traditional conjoint for 10+ attributesattributes..
)-( * )-( * respondent accuracy often doubtfulrespondent accuracy often doubtful..
)-( * )-( * inter-attribute correlation more problematic than in inter-attribute correlation more problematic than in traditionaltraditional
conjointconjoint..
)-( * )-( * lack of realism since respondent does not perform a choice lack of realism since respondent does not perform a choice tasktask..
Adaptive (or Hybrid) Adaptive (or Hybrid) ConjointConjoint
** Self-explicated ratings used to create a Self-explicated ratings used to create a manageable subset of stimuli , then traditional manageable subset of stimuli , then traditional conjoint used with respondent rating different conjoint used with respondent rating different sets of stimulisets of stimuli
)+(* )+(* more manageable than traditional conjoint for 10+ more manageable than traditional conjoint for 10+ attributesattributes..
)+( * )+( * predictive ability comparable to traditionalpredictive ability comparable to traditional
)-( * )-( * requires specialized software, e.g. sawtooth software s requires specialized software, e.g. sawtooth software s adaptiveadaptive
congoint analysiscongoint analysis..
SourcesSources
Decision pro .bizDecision pro .bizDepartment of consumer student Department of consumer student
university of guelphuniversity of guelphConjoint analysis introduction .htmConjoint analysis introduction .htm
”“”“Thanks for your Thanks for your attention”attention”
negin aghighinegin aghighi0918314671809183146718