A Structured Methodology for Choosing Attributes and Levels in Conjoint Analysis

In addition to its advantages, conjoint analysis, like any technique, has a number of drawbacks.

The most prominent are restrictions on the number of attributes and levels that may be included, and the lack of a holistic view (i.e., the whole is purely the sum of its parts).  The chief reason for the number of restrictions on attributes and levels is the burden that measurement of a large number of attributes and levels places on respondents.  The size of the respondent task increases exponentially for many conjoint methods.  While there are types of conjoint analysis that relax this restriction to a greater or lesser degree (e.g., ACA and CaseMap), the most popular type of conjoint analysis, Choice-Based Conjoint (CBC) can deal with, at most, about ten attributes, but six attributes is a much more reasonable number.

Many products dealt with in conjoint analyses have far more than six or ten attributes.  Personal computers and smart phones may have up to 150 attributes.  Helicopters have even more.  Therefore researchers almost always have to make difficult choices about which attributes to include.

So the question is: How do you strike a balance between capturing those attributes most relevant to and important in the product choices of customers, and those attributes important to R&D, engineering, and others involved in product design?

At Applied Marketing Science, we combine state-of-the-art qualitative research methods that identify customer purchase decision processes with The House of Quality from QFD, a classic method for relating customer needs to product characteristics.

The first step in the process is to use Unstructured Direct Elicitation (UDE), a new methodology for understanding purchase decision processes.  Customers are asked to treat interviewers as agents, whom they must train to select the right product from the available alternatives, using a set of decision rules.  The decision rules are coded by analysts as a set of needs.  Respondents are also asked to divide 100 imaginary points between the rules they have mentioned, to get a measure of relative importance for each rule.  Rules are also classified as non-compensatory (“must have” or “must not have”) or compensatory (i.e., doing well on one decision criterion can compensate for doing badly on another).

Through aggregation of these sets of needs across customers, a set of needs which drive purchasing choices can be developed, as well as a rough measure of the importance of each need.  Note that the list of needs developed here is by no means as detailed or exhaustive as the 100 or so detailed needs developed through the Voice of the Customer technique.  For conjoint analysis, we are working at a more aggregate or tactical level, where we might find 15 – 25 “secondary level” needs.

Once the list of drivers of purchasing and their relative importance has been created, they are combined with a set of product characteristics using the House of Quality.  The House of Quality has been used since the 1980’s as part of Quality Function Deployment to help marketers and engineers better communicate about how to meet customer needs through a judicious choice of product features.

The selection of attributes is done by finding the combination of product characteristics which are few enough to fit into the conjoint methodology, yet which best cover the chief drivers of purchase.  This is the debate we always have, but the difference is that here we have a structure and some data upon which to base our reasoning.

To find the levels to use in the conjoint analysis, we use a combination of customer input and our client’s knowledge of the possible range of levels present in competitive and hypothetical products that the client plans to include in the conjoint simulator.

To find out more about this method, along with when to use conjoint analysis, and much more please join AMS at the April 22nd PDMA Webcast, Best Practices in the Design of Conjoint Analysis. http://www.pdma.org/events_detail.cfm?pk_event=550

Summary
This methodology combines a “mini-VOC” with a “mini-QFD,” at a cost appropriate for a conjoint analysis, to provide a structure that connects a prioritized list of needs with a set of product characteristics.  This provides a sound basis for discussions about which product characteristics to use in the conjoint analysis.

<!--Session data-->