Integrating datasets: Segmenting the fashion market using risk aversion

Abstract

A marketing segmentation can often be improved with the addition of variables which are often found on different datasets. Using a classification regression tree (CRT) methodology with predictor variables shared across datasets, the terminal node identification equations can be used to estimate the variables on a different dataset. The use of CRT allows the inclusion of categorical variables, such as marital status and ethnicity, as well as continuous variables, such as age and education. Three datasets were integrated and a chi-square automatic interaction detector (CHAID) tree is then used to segment the women's clothing fashion market by demographic and reward and aversion variables. The analysis suggests possible marketing strategies targeting high-spending segments as well as media strategies.