conjoint analysis lecture 21 oct iimk

44

Upload: deepak-verma

Post on 23-Nov-2015

10 views

Category:

Documents


0 download

DESCRIPTION

conjoint analysis

TRANSCRIPT

Slide 1

Estimating Part Worth for Industrial CleanerFind average rank for all 8 stumuli= [1+2+3+4+5+6+7+8]/8 = 4.5Find deviation of rank for each stimuli e.g.Phosphate free = (1+2+3+4)/4=2.5 Deviation = 2.5-4.5 =-2Phosphate based = (5+6+7+8)/4=6.5Deviation = 6.5-4.5 = 2

Factor levelRank across stimuliAverage Rank of levelDeviation from overall rankFormLiquid1,2,5,63.5-1Powder3,4,7,85.51IngredientsPhosphate free1,2,3,42.5-2Phosphate based5,6,7,86.52BrandHBAT1,3,5,74-0.5Generic2,4,6,850.5Estimating Part Worth for Industrial CleanerSteps for calculationS1: Sum of deviations from 4.5 are squared =1+1+4+4+.25+.25=10.5S2: Standardizing value=6/10.5=0.571S3: Multiply Squared deviation by standardizing value =0.571 Part worth = Take square root of squared standardized deviationFactor levelRank across stimuliAverage Rank of level(Av Rank=4.5)Deviation from overall rankLiquid1,2,5,63.5-1Powder3,4,7,85.51Phosphate free1,2,3,42.5-2Phosphate based5,6,7,86.52HBAT1,3,5,74-0.5Generic2,4,6,850.5Estimating Part Worth for Industrial CleanerFor phosphate free calc= Sq dev=4, Stand Sq Dev= 4*.571=2.284, PW= (2.284)1/2 = 1.151Factor levelReversed deviationSquared deviationStandardized deviation Part worth Factor ImportanceF-Liquid+11.571+.75628.6 %F- Powder-11-.571-.756I-Phosphate free+242.284+1.51157.1 %I-Phosphate based-24-2.284-1.511B-HBAT+0.5.25.143+.37814.3 %B-Generic-0.5.25-.143-.378Estimating Part Worth for Industrial CleanerFactor Importance is given by the difference in part worths of the levels of the factor. In this case FORM= .756-(-.756)=1.512; % Importance= 1.512/(1.512+3.022+.756) = 28.6 %Conjoint Analysis- Predictive AccuracyThe estimated part worth for each attribute level can be summed up and then rank ordered.This should yield the rank preference of the respondent. The degree to which the predicted rank order preference corresponds to the original respondent rank order is known as the predictive accuracy of the part worth calculationPart worth values are not compared across respondents as they pertain only to the respndents preference structure

Calculate the expected market share for a product -specified level of attribute values and a closed set of competitorsTypes of Conjoint Analysis

Traditional additive

Adaptive or Self-explicated conjoint

Choice Based

Product Design and Market Share Optimization Salem Foods

Antonios brand, which has a thick crust, mozzarella cheese, chunky sauce, and medium flavored sausage.The Kings brand pizza has a thin crust, a cheese blend, smooth sauce, and mild-flavored sausage.For C1: Utility or PW of Antonios =2+6+17+27=52; Kings = 11+7+3+26=47

Salem Problem: Choice of Pizza attributes This can be modeled as an Integer programming problem:Problem def: Salem has to design a Pizza with highest utility for sufficient people so as to justify the design and launch of the new product.Variables : lij = 1 if Salem chooses level i for attribute j, = 0 otherwiseyk =1 if customer chooses Salem Pizza, =0 otherwiseBecause number of customers choosing Salem Pizza has to be maximized, Obj Fn= Max y1+ y2+ y3+ y4+ y5 + y6 + y7+ y8

Thus the problem requires that if C 1 has to switch PW or utility > 52 the current PW for Antonios brand.Since y1 =1 only when customer buys from Salem, so we write this above expression for consumer 1 as

Problem formulation for Salem Foods Thus for all 8 consumers, of which 1,4,6,7,8 prefer Antonios and2,3,5 prefer kingsAdditional constraints to ensure that only one attribute gets chosen

This problem can now be solved using Management Scientist or Lindo/Lingo.Soln: yields the result that Salem must choose Thin crust, Cheese blend, Chunky Sauce and Mild sausage flavor. Consumer types C1, C2, C6 and C7 will choose this pizza type.