Fuzzy target selection using RFM variables

Abstract

An important data mining problem from the world of direct marketing is target selection. The main task in target selection is the determination of potential customers for a product from a client database. Target selection algorithms identify the profiles of customer groups for a particular product, given data about the clients and a test sample of customers known to possess or have interest in the product. In addition to the numerical performance of the target selection models, model transparency is also important for evaluation by experts, for obtaining confidence in the model derived, and for selecting an appropriate marketing channel. Fuzzy models for target selection are interesting in this respect, since they can be used to obtain numerically accurate models, while providing a linguistic description as well. Fuzzy clustering is a suitable method for obtaining fuzzy target selection models. The paper describes how fuzzy target selection models can be developed when using customer features based on recency, frequency and monetary value. Both a segmentation approach and a scoring approach are considered. The two approaches are illustrated on a case by using data from the target selection campaigns of a large charity organization.