Marketing Strategy Conjoint Analysis and Product Design “An economic forecaster is like a cross-eyed javelin thrower: they don't win many accuracy contests, but they keep the crowd's attention.” Anonymous Conjoint • What happens when I ask someone what’s important with respect to retailers? – “In retailers, I like assortment, low price, convenience.” • If I want to know what they REALLY care about, I have to ask them to make trade-offs – “If I have to give up some assortment to get a bit lower price, fine, but I don’t want to give up convenience.” Conjoint Analysis makes these tradeoffs systematic so we can quantify and better understand the results. 1 Why Use Conjoint Analysis? Wendy’s “Big Classic” Nine buns options: Forty special sauces: Lettuce hard, soft sesame, poppy cold, toasted, or warmed square, round (24 options) steak sauce hot sauce mustard salad dressing mayonnaise chopped shredded leaf Tomato Four box styles thin medium thick Conjoint Analysis Features Considered Jointly Conjoint Analysis is a set of methods designed to measure consumer preferences for a multi-attribute product. Steps in Conjoint Analysis Identification of the Determinant Attributes Choice of data collection procedure Data collection Data analysis Strategy Assessment 2 Language of Conjoint • Attributes – any clearly definable feature or characteristic – e.g., cost per flight • Levels of Attributes – range of options available for each attribute that determines the best and worst the attribute can be – must have at least two levels for each attribute – e.g., $4, $6, and $8 Language of Conjoint (cont.) • Preference model – Comparison of overall preferences (utilities) for a set of alternatives for an individual respondent • Importance Weights – the degree to which variation in the level of an attribute affects the overall utility – calculated as the range of part-worth utilities for the attribute levels standardized such that the partworth for the worst level is set to zero 3 Multiattribute Model of Attitudes • Identify relevant attributes • Determine importance weights (W) for those attributes • Determine beliefs about brand on those attributes (a1) • Sum all attributes used in evaluating the brand weighted by the value of each attribute Customer Preferences • Objective: measure customer’s utility function U = Price + CPU + Monitor + Brand • Estimating expected market share Price Process. Monitor Brand Total Importance .4 .3 .2 .1 1 Dell 8 9 10 8 8.7 Compaq 10 6 5 7 7.5 IBM 7 10 7 10 8.2 Share 24.4 4 Another Example Price Brand Horsepower Upholstery Sunroof $23,000 Toyota 220 HP Cloth Yes $25,000 Volkswagen 250 HP Leather No $27,000 Saturn 280 HP $29,000 Kia Computerized Survey 5 Conjoint Output Attribute Level Utility (part-worths) T-value Price $23,000 2.10 14.00 Brand Horsepower Upholstery Sunroof $25,000 1.15 7.67 $27,000 -1.56 10.40 $29,000 -1.69 11.27 5.00 Toyota 0.75 Volkswagen 0.65 4.33 Saturn -0.13 0.87 Kia -1.27 8.47 220 HP -2.24 14.93 250 HP 1.06 7.07 280 HP 1.18 7.87 Cloth -1.60 10.67 Leather 1.60 10.67 Yes 0.68 4.53 No -0.68 4.53 Utility U = Toyota + 280 HP + Leather + No sunroof + $23,000 U= 0.75 + 1.18 + 1.60 – 0.68 + 2.10 = 4.95 U= 0.65 (Volkswagen) = 4.85 10 Attribute Level Utility (part-worths) T-value Price Brand $23,000 2.10 14.00 Toyota 0.75 5.00 Volkswagen 0.65 4.33 Horsepower 280 HP 1.18 7.87 Upholstery Leather 1.60 10.67 Sunroof No -0.68 4.53 6 Conjoint Output Attribute Level Utility (part-worths) T-value Price $23,000 2.10 14.00 Brand Horsepower Upholstery Sunroof $25,000 1.15 7.67 $27,000 -1.56 10.40 $29,000 -1.69 11.27 5.00 Toyota 0.75 Volkswagen 0.65 4.33 Saturn -0.13 0.87 Kia -1.27 8.47 220 HP -2.24 14.93 250 HP 1.06 7.07 280 HP 1.18 7.87 Cloth -1.60 10.67 Leather 1.60 10.67 Yes 0.68 4.53 No -0.68 4.53 If we assume a linear relationship Utility spread 2.10 – 1.36 = 0.74 Dollar spread ($25k and $27k) Utility spread between the two tested price points ($25k and $27k) Attribute Level Utility (part-worths) T-value Price $23,000 2.10 14.00 Sunroof $25,000 1.15 7.67 $27,000 -1.56 10.40 Yes 0.68 No -0.68 1.36 4.53 4.53 7 Share Forecasting Multinomial Logit Model Utilities and Weights 8 Share Forecasting Multinomial Logit Model SAKE = Design of 4 + Max Freq of 34 + Power of 68 + Price of $363 = 0.45 + 0.44 + 0.49 + 0.54 = 1.92 What would its share be? 1. 2. Design of 5 + Max Freq of 50 + Power of 52 + Price of $284 Design of 3 + Max Freq of 26 + Power of 100 + Price of $205 1. 0.45 + 0.48 + 0.13 + 1 = 2.06 2. 0.43 + 0.38 + 0.54 + 0.24 = 1.59 Example – Rank these Beers 9 Example: Fitness Facility Two attribute Lockers & Sauna Lockers come in small, medium and large Sauna exists or it doesn’t There are six different combinations Sauna Locker Yes No Small (always) Large (daily) Medium (always) Rank 2 Rank 4 Rank 1 Rank 3 Large (daily) Rank 5 Rank 6 We can assign “utility” points Sauna Locker Small (always) Large (daily) We can assign points for each ranking. In this example, the top ranked choice gets the most point! Yes No Rank 2 4 Rank 4 2 Average = 3 Medium (always) Rank 1 5 Rank 3 3 Average = 4 Large (daily) Rank 5 1 Rank 6 0 Average = 0.5 Average = 3.33 Average = 1.67 Sauna: Y = 3.33, N = 1.67 Locker: S = 3, M = 4, L = 0.5 10 Conjoint Analysis “Tease” out a “value system” from individual preferences (numerical values from rank orders in this example) The individual is averse to not having a storage locker “always” (medium = 3, large = 0.5) Lockers Sauna Unwilling to trade permanent storage for Sauna decrease = 3 to 0.5 = 2.5 loss increase = 1.67 to 3.33 = 1.66 gain NET LOSS 0.84 s m l Size n y Includes gap Markstrat 11 The individual’s Value System Product Locker medium small medium small large large Value Sauna Rank yes 4 + 3.33 = 7.33 1 yes 3 + 3.33 = 6.33 2 no 4 + 1.67 = 5.67 3 no 3 + 1.67 = 4.67 4 yes .5 + 3.33 = 3.83 5 no .5 + 1.67 = 2.17 6 Original Rank 1 2 3 4 5 6 Importance Weights Product Locker medium small medium small large large Sauna yes 4 + 3.33 = 7.33 yes 3 + 3.33 = 6.33 no 4 + 1.67 = 5.67 no 3 + 1.67 = 4.67 yes .5 + 3.33 = 3.83 no .5 + 1.67 = 2.17 Sauna Range 1.67 – 3.33 = 1.66 Locker Range 0.5 – 4 = 3.5 3.5 + 1.66 = 5.16 Importance of Lockers 1.66/5.16 = 0.32 Importance of Sauna 3.5/5.16 = 0.68 12 Importance of Horsepower Attribute Level Utility (part-worths) T-value Price $23,000 2.10 14.00 $29,000 -1.69 11.27 Brand Toyota 0.75 5.00 Kia -1.27 8.47 Horsepower 220 HP -2.24 14.93 280 HP 1.18 7.87 Upholstery Sunroof Cloth -1.60 10.67 Leather 1.60 10.67 Yes 0.68 4.53 No -0.68 4.53 Utilities and Weights High Design Max Freq Power Price Low 0.45 0.51 0.72 1 Diff 0.35 0.29 0.05 0 SUM 0.10 0.22 0.67 1.00 1.99 Weight 0.05 0.11 0.34 0.50 13 Importance Weights by Segment Widely Used • • • • • • Fujitsu - cellular phones Levi Strauss - blue jeans Marriott Corporation - hotels Sunbeam - food processors condominium design and pricing snowmobiles, credit cards, aircraft, rural health care, technical information services 14