Treatment classification trees: A method for offer segmentation in direct marketing

Abstract

Direct mail offer designers and copywriters strive to create the offer which will yield the greatest response for a particular product offering. Testing is generally performed to determine which offer of the many candidates has the broadest appeal. Once the test results are compiled, usually a single offer will be chosen and sent to all targets in a bulk mailing. The assumption made here is that all the targets will prefer the offer which was deemed best in the tests. The hypothesis is that consumers are not homogenous in their reactions to offers, and that customers' characteristics can be used to determine which offer is best on an individual, not a list basis. This paper proposes a technique for classifying consumers into offer groups based on test samples of different offers. This new technique is called Treatment Classification Trees, or TaCT. The first application shown illustrates when this method is preferred over a standard logit model. Also presented is a case study using actual test data showing that profits can be increased using TaCT.