A Least Squares Procedure for Benefit Segmentation with Conjoint Experiments

Abstract

A popular procedure for benefit segmentation based on conjoint experiments has been to estimate individual-level part worths and then form nonoverlapping clusters of consumers with similar estimates. Rather than using these estimates as the criteria for clustering, the least squares procedure discussed in the article attempts to group consumers into homogeneous segments so their stated preferences are explained maximally by their group-level preference functions. This procedure also provides a measure of the expected predictive accuracy that will help the researcher in choosing an adequate aggregation level.