Using segment level stability to select target segments in data-driven market segmentation studies

Abstract

Market segmentation is widely used by industry to select the most promising target segment. Most organisations are interested in finding one or a small number of target segments to focus on. Yet, traditional criteria used to select a segmentation solution assess the global quality of the segmentation solution. This approach comes at the risk of selecting a segmentation solution with good overall quality criteria which, however, does not contain groups of consumers representing particularly attractive target segments. The approach we propose helps managers to identify segmentation solutions containing attractive individual segments (e.g., more profitable), irrespective of the quality of the global segmentation solution. We demonstrate the functioning of the newly proposed criteria using two empirical data sets. The new criteria prove to be able to identify segmentation solutions containing individual attractive segments which are not detected using traditional quality criteria for the overall segmentation solution.

Appendix: Technical Appendix

1.1 Relabelling algorithm required for the calculation of pertinaciousness

For series of partitions we propose a new relabelling algorithm which makes it possible to track segments over partitions with different numbers of clusters. Let again P ₁, P ₂,…, P _m be a series of m partitions with numbers of clusters k ₁<k ₂<…<k _m .

Note that if k _i+1 = k _i +1, then only one column needs to be inserted in step 4. However, the algorithm also works for the more general case.

1.2 Calculating segment-wise rerun stability