How to develop new approaches to RFM segmentation

Recency, frequency and monetary (RFM) is a simple and actionable way that has long driven direct marketing efforts. In many cases, however, this empirical segmentation is encumbered with basic shortcomings evident in two aspects: first, the advantage of simplicity often disappears in terms of statistical significance. Secondly, the three-dimensional measure is less predictive than sophisticated models, such as Chi-Square Automatic Interaction Detection (CHAID) and regression analysis. Using RFM as an entry point, this paper discusses the necessity and reality of upgrading this crude method to advanced approaches, where two options are hereby proposed:—option 1: substituting result-based statistical findings for the traditional intuition-based coding, RFM implementation becomes simple and robust. The transition also advances empirical RFM to CHAID analysis;—option 2: introducing ‘V=M/R’ defined as a customer value, traditional RFM schemes are readily migrated to individual V-scores without in-depth statistic proceedings. The innovation is compatible to logistic regression.