A continuation of our Meet the Experts Series, Steve Gaskin answers questions regarding his experience and expertise in conjoint analysis.
What software do you use to program and analyze the survey?
We use Sawtooth Software, which is the leading conjoint analysis software, to program the survey, estimate the partworths and run scenarios. We also use Matlab for holdout calculations, and Excel and SPSS for miscellaneous analyses.
How do you design a conjoint analysis for a product with a large number of features such as a computer or a smartphone?
In any mathematical modeling application, tradeoffs are made between including every possible feature and the ability of customers to deal with these features during the purchase process and in the survey. The final selection depends upon the objectives of the analysis. For our class action litigation applications, it is not necessary to include each and every attribute. We need to measure the tradeoffs between the presence or absence of the key feature and price. We add in other distractor attributes to ensure that the survey is not leading.Have you run into any problems when explaining or describing conjoint analysis to the courts or to judges?
Not that I can recall. We try to convey what we have done in simple, non-technical language, and write well. The lawyers we work with seem to pick up the underlying concepts of market research and conjoint analysis very well.
What’s the relationship between partworths and utility in a conjoint model?
A partworth is an estimate by the conjoint analysis of the portion of the overall preference or utility associated with each level of an attribute or feature. Partworths are numerical scores that measure how much (on a relative basis) each level of each attribute influences customers’ purchase choices. ‘Utility’ is a term economists use that is a measure of the value consumers derive from the use of a product. The underlying assumption of conjoint analysis is that a consumers’ overall value, or utility, for a product is a weighted sum of the value of each of its parts; or, put plainly, “The whole is the sum of its parts”.
What is the difference between a choice model and a conjoint model?
A choice model is a particular type of conjoint model. They all involve making tradeoffs between alternatives to reveal what drives customer preferences. Decades ago, in a conjoint analysis survey respondents would be presented with a series of hypothetical products, each described on a separate card, and they would rank or rate each card. These days, choice models are most frequently used, because they better match the sort of decision customers make while shopping. In a choice model, survey respondents are presented with several products to choose from, each described in terms of levels of each product attribute. This is called a choice task. The respondent chooses which product he or she would most prefer. The respondent will be presented with a number of choice tasks, during the course of the survey, each generated by an overall experimental design. The characteristics of the products presented in each choice task, and the choices the respondent made, are fed into a choice model to estimate the partworths for each level of each attribute, for each respondent.
How does a conjoint simulator work?
A conjoint simulator is a computer program into which an analyst can enter products, described in terms of the levels of the product attributes, often including price. Then, the simulator uses the partworths to calculate the overall utility for each product for each survey respondent, and a decision rule involving the utilities to predict which product each respondent would be most likely to choose. From this, market shares based on these product utilities are estimated, by summing the probability that each respondent will choose each product.
In our class action applications, we use the simulator to estimate what reduction in market price is necessary to compensate for the absence of a product feature for the class as a whole.
To learn more about conjoint analysis, register for our upcoming webinar, "Conjoint Analysis: Providing a Basis for Damages in Class Action Litigation".