Market Segmentation Using Bayesian Model Based Clustering

Abstract

This dissertation deals with two basic problems in marketing, that are market segmentation, which is the grouping of persons who share common aspects, and market targeting, which is focusing your marketing efforts on one or more attractive market segments. For the grouping of persons who share common aspects a Bayesian model based clustering approach is proposed such that it can be applied to data sets that are specifically used for market segmentation. The cluster algorithm can handle very large data sets, is able to deal with data that are missing at random and accounts for within cluster item dependencies. Using six criteria of good market segmentation, an information criterion and two conjectures, describing the geometry of model based clustering models, presented by Hoijtink and Notenboom (2004), a procedure reduces the statistically optimal number of clusters to a smaller number, suited for the intended marketing purposes. The Bayesian model based clustering approach is compared with the two model based cluster algorithms implemented in LatentGold and Glimmix. Using simulation studies the performance of each of the approaches is evaluated. For market targeting purposes it is shown how marketing information can be fused to a company's customer database. Using two real marketing application, two traditional data fusion methods, that are, logistic regression and nearest neighbor algorithms, are compared with two model based clustering approaches. Finally, the results are evaluated using internal and external criteria.