Which Conjoint Method Should You Use?
One subtle (and false) implication of the beautiful chart we presented showing trends in conjoint analysis is that perhaps you should be using discrete choice modeling (CBC, which stands for Choice-Based Conjoint) because CBC is, by far, the most widely used method among those using Sawtooth conjoint tools. But CBC is not the best choice for many types of studies. Its “popularity” as shown in the chart likely reflects a dramatic skew in the types of companies and respondents who do lots and lots of conjoint studies — namely, those doing consumer packaged goods (CPG) research.
A good portion of our conjoint work, in contrast, relies on other methods, including CVA (Conjoint Value Analysis — also known as traditional full-profile conjoint) and ACA (Adaptive Conjoint Analysis). Why? Because there are many situations in which people make decisions that do not involve assessing several competing options from which they select just one.
There are several factors that govern which conjoint method to choose, but the three factors we find most important for our work are:
1. Decision making context. The conjoint method should reflect how buyers actually make decisions for the product or service at hand. For example, do they make a yes-no decision about the product independent of other choices, or do they select one or several from among many? For the former, CVA may be optimal, and for the latter, CBC.
2. Sample size. Often we work with small—even tiny—samples, either because recruiting them is difficult and expensive (consider neurosurgeons) or because we are incorporating conjoint modeling into a qualitative study. CBC is not a good method for small samples because too little information is gathered via each choice set. In contrast, CVA and ACA are information-rich methods, and a robust model can be built even with just a handful of respondents.
3. Design complexity. CBC and CVA work well with just four or five attributes to consider. But if the design gets more complex, respondents get overloaded and data quality suffers. That’s when we need to move into adaptive modeling, either ACA or ACBC.
Beyond these three, there are other factors to consider, including how and whether pricing is central to our research question, how long the survey will be, and what mode and platform we are implementing for the survey.
Need help choosing the right conjoint method? Here is a handy interactive website from the excellent folks at Sawtooth, and here is a useful white paper that reviews the considerations we outlined above, and more. And of course feel free to give us a call at (312) 348-6089 if you need help choosing and implementing a conjoint study of your own.