Alternative segmentations can be compared on an “apples-to-apples” basis, by the amount of information they convey about a set of customer attributes. This article presents a simple and effective new metric to compare different segmentations on a common basis—the information they convey about a particular set of relevant attributes.
This turns out to be the same metric that latent class analysis maximizes, yielding new insights into how segmentation algorithms work and how they can be better harnessed in practice.
This article appeared in Marketing Research magazine published by the American Marketing Association in Winter 2010, and was awarded the David K. Hardin prize for best published paper in 2010 by the AMA.
