Behavioral Intelligence in Digital Retail: An Extended RFM Framework for Customer Segmentation and Resource Allocation

The proliferation of behavioral data in digital retail has not been matched by equally rigorous frameworks for converting that data into customer intelligence that practitioners can act on. This paper addresses that gap by introducing RFM-B, a behavioral segmentation framework that extends the classical recency–frequency–monetary (RFM) model with four additional indicators derivable from standard e-commerce event logs: conversion rate (CVR), category breadth, average order value (AOV), and brand diversity. Applied to 4,635,837 interaction events from 64,204 purchasing customers observed over 61 days on a large-scale multi-category platform, the framework produces five customer archetypes—Champions, Loyal Customers, Potential Loyalists, At-Risk, and Lost—whose behavioral profiles differ systematically in purchase efficiency, platform embeddedness, and commercial significance. A machine-learning recoverability analysis using a Random Forest classifier achieves 96.99% held-out accuracy (5-fold cross-validated: 96.84% ± 0.10%), confirming that the segments are operationally deployable in real-time marketing automation systems. The central empirical finding is the identification of a Potential Loyalist segment characterized by an average order value of USD 147.2—more than three times the platform-wide mean—combined with low purchase frequency, a profile that standard RFM frameworks would systematically misclassify as low-priority. The results show that enriching the behavioral feature space yields a customer typology that is both analytically coherent and directly actionable, and that interpretability is not a secondary concern but a functional prerequisite for organizational adoption.