Better preference prediction with individualized sets of relevant attributes

It is concluded that for commercial analyses conjoint analyses should be based on the individualized conjoint approach, based upon an EDP-assisted information handling system, rather than the traditional approach, which is only partially individualized.

Abstract

Conjoint studies are aimed at analysing individual preferences based on product profiles. The findings are used either for exploratory or for predictive purposes. One basic paradigm states that the analysis should have an individual orientation. Therefore, the part-worth utilities are estimated on an individual level. The set of relevant attributes, however, is usually supposed to be unique for all respondents. In contrast to this traditional approach, which is only partially individualized, we developed a completely individualized conjoint approach based upon an EDP-assisted information handling system. In addition, we have tested the quality of this completely individualized analysis against the partially individualized analysis. In regard to predictive validity, the completely individualized analyses were, as far as the correct first choice probability criterion is concerned, significantly better than the partially individualized analyses. Thus, we conclude that for commercial analyses conjoint analyses should be based on the individualized conjoint approach.