Statistical models for market segmentation

Abstract

It is an essential element of market research that customer preferences are considered and the heterogeneity of these preferences is recognized. By segmenting the market into homogeneous clusters the preferences of customers is addressed. Latent class methodology for conjoint analysis, proposed by Green (2000), is one of the several conjoint segmentation procedures that overcome the limitations of aggregate analysis and priori segmentation. This approach proposes the proportional odds model as a proper statistical model for ordinal categorical data in which the item attributes are included in the linear predictor. The likelihood is maximized through the EM algorithm. This paper considers two extensions of this methodology that incorporate individual characteristics into the models.