Conjoint Analysis Y. İlker TOPCU, Ph.D. www.ilkertopcu.net www.ilkertopcu.org www.ilkertopcu.info www.facebook.com/yitopcu twitter.com/yitopcu Conjoint Analysis • A statistical method that can be used as an indirect priority determination procedure. • Requires decision makers to rank or rate alternatives and Derives priorities that provide the best fit of the evaluations for alternatives. Conjoint Analysis • A survey research tool that predicts consumer preferences in multi attribute decision making where alternatives are evaluated w.r.t. several attributes (factors) in a wide variety of product and service context. • It became popular in marketing research as it can predict what consumers will buy when they faced with the availability of many brands and a great number of product characteristics. Conjoint Analysis • By systematically varying the characteristics of a product or a service and observing how survey participants react to these product/service profiles, the researcher can statistically deduce the scores for each characteristic (factor) with which participants may have been subconsciously using to evaluate products/services. Steps 1. 2. 3. 4. Determining possible combinations Evaluating possible combinations Computing conjoint utilities Revealing priorities Determining Possible Combinations • Levels within each factor must be developed. • There would be many possible combinations of these factor levels. • By using experimental design principles of independence and balance, some of the combinations are carefully chosen; therefore participants do not have to evaluate all possible combinations. Evaluating Possible Combinations • In the 70’s, survey participants were requested to evaluate each of many combinations that are printed on separate cards one by one by ranking or rating on a scale. • In the 80’s, a computerized version called as Adaptive Conjoint Analysis was utilized, which could effectively gather more attributes and levels by focusing on them that were most relevant to each participant. Evaluating Possible Combinations • In the 90’s, Choice Based Conjoint (CBC) became popular. • With CBC, participants were requested to choose among a certain number of possible combinations instead of ranking them or rating each of them individually. CBC • Nowadays CBC is widely used as consumers in real life do not score each alternative, instead they simply choose among them; which make CBC questions seem more realistic. • Generally, a certain number of possible combinations with an additional “none” choice (that can be chosen if none of the combinations is preferred) are presented to the participants. Computing Conjoint Utilities • By utilizing regression analysis the scores of the factors can be inferred. • Sawtooth software can be used for the necessary calculations required for statistical analysis based on logistic regression. • These scores are useful for determining the relative priorities of each factor. Computing Conjoint Utilities • The scores are scaled to an arbitrary additive constant within each factor. • The arbitrary origin of the scaling within each factor is based on dummy coding. • When using “effects coding”, a specific kind of dummy coding, scores are scaled to sum to zero within each factor. • In this case, the scores can be regarded as conjoint utilities. Revealing Priorities • After finding the range in the utility values of a factor (i.e. the difference between the maximum utility and the minimum utility), the percentages from relative ranges are calculated. • These normalized ranges are the priorities of the factors. Case Study CUSTOMER ORDER SELECTION • Potential profit rate per unit of time • Compatibility of potential order with available capacity • Customer credit of future business opportunity • Negotiability level of production schedule for order • Level of potential future order with higher profit Developing Levels A three-level scale (High-Medium-Low) is used for: • Potential profit rate per unit of time • Compatibility of potential order with available capacity • Customer credit of future business opportunity • Negotiability level of production schedule for order A two-level scale (Exists-Does not exist) is used for • Level of potential future order with higher profit Determining Possible Combinations • By using experimental design, 90 hypothetical customer orders are chosen among 162 of them • 30 conjoint cards having 3 possible orders and an additional “none” alternative are formed. Evaluating Possible Combinations CARD 1 Customer credit of future business opportunity The potential profit rate per unit of time The negotiability level of production schedule for the order The level of potential future order with higher profit The compatibility of potential order with available capacity Alternative 1 Alternative 2 Alternative 3 HIGH MEDIUM LOW MEDIUM HIGH LOW MEDIUM LOW HIGH DOES NOT EXIST EXISTS EXISTS MEDIUM LOW HIGH Alternative 4 NONE Computing Conjoint Utilities DM1 DM2 DM3 Compatibility of potential order with available capacity HIGH 22.60 75.54 72.60 MEDIUM 29.31 43.44 44.42 LOW -51.92 -118.98 -117.01 Potential profit rate per unit of time HIGH 19.41 78.54 40.92 MEDIUM 35.07 17.84 38.01 LOW -54.48 -96.38 -78.93 Customer credit of future business opportunity HIGH 22.13 43.28 53.66 MEDIUM 62.45 -14.78 31.62 LOW -84.59 -28.50 -85.29 Computing Conjoint Utilities DM3 DM2 DM1 Negotiability level of production schedule for the order -3.58 -21.59 HIGH -47.14 16.47 25.89 MEDIUM -1.11 -4.30 LOW 48.25 Level of potential future order with higher profit -5.65 EXISTS 43.40 DOES NOT EXIST -43.40 5.65 -12.89 -11.12 11.12 Revealing Priorities DM1 DM2 DM3 Ave. Compatibility of potential order w. avail. capacity 16.25% 38.90% 37.92% 31.02% Potential profit rate per unit of time 17.91% 34.98% 23.97% 25.62% Customer credit of future business opportunity 29.41% 14.36% 27.79% 23.85% The negotiability level of prod. schedule for order 19.08% 9.50% 5.87% 11.48% Level of potential future order with higher profit 17.36% 2.26% 4.45% 8.02%

1/--страниц