Componential Segmentation Based Conjoint Analysis vs Cluster analysis

Abstract

In Componential Segmentation interest focuses on the interaction effect of person and product attribute levels to produce a response (overall evaluation) for various product descriptions. A person’s reaction to a product is broken into the sum of two components: 1) the average part-worth utilities due to the attribute levels of the product and 2) the interactions between the person’s background variables and the attribute levels. In this paper we adopt the dummy-coded parametrization of the model, which provides two baselines. Two segmented methods of performing conjoint analysis, clustered and componential segmentation, are compared with each other. The predictive power of the clustered segmentation model is higher than that of componential segmentation.