As technology advances, new products (e.g., digital cameras, computer tablets) have become increasingly complex. Managers as well as researchers often face considerable challenges in understanding consumers’ preferences for such products. For example, conventional preference elicitation methods such as conjoint analysis often become infeasible, because the number of questions required to obtain accurate estimates increases rapidly with the number of attributes (and/or attribute levels) of the focal product.
In a recent paper with a colleague Lan Luo at USC, we propose an adaptive question design algorithm for complex product preference elicitation. We use a fuzzy support-vector-machine (SVM) based active learning method that in many aspects can meet the product complexity challenge. The proposed method is fast enough to predict the most informative questions for each respondent based on information available from the marketplace and the respondent’s past answers on the fly.
Compared to extant methods, the proposed algorithm is particularly suitable for high-dimensional problems. Our empirical and synthetic studies suggest that the proposed method performs well for product categories equipped with as many as 70 to 100 attribute levels, which is typically considered prohibitive for decompositional preference elicitation methods. Furthermore, we demonstrate that the fuzzy SVM active learning method provides a natural remedy for a long-standing challenge in adaptive question design by gauging the possibility of response errors and incorporating it into the survey design.
Finally, diverging from most adaptive question design methods that only utilize information from the focal respondent for question selection, we show that responses from previous respondents may be used in a live setting to improve active learning of the focal respondent’s product preferences. This research represents one of our attempts to meet the big data challenge from a marketing perspective.