This study aims to improve the accuracy of customer equity (CE) estimation in complex retail environments by introducing a hybrid framework that integrates machine learning (ML) with Markov Chain modeling, enabling dynamic prediction of CE in settings characterized by frequent transactions and diverse product offerings.
This study develops a hybrid framework that integrates ML and Markov Chain modeling for dynamic CE prediction. ML models extract customer behavioral features, which are combined with a probabilistic model to enhance retention prediction. A newly introduced “stability” dimension extends the traditional Recency–Frequency–Monetary (RFM) framework, enabling dynamic clustering. These clusters form the basis for Markov states, allowing the modeling of customer transitions and more accurate CE estimation.
The findings demonstrate that integrating ML with Markov Chain modeling improves the accuracy of CE estimation. The approach enables managers to monitor both current and future customer states with precision. Dynamic clustering, enhanced by the stability dimension, identifies behavioral shifts early, supporting proactive churn detection and more targeted retention strategies for high-value customers.
This study contributes originality by introducing a hybrid ML–Markov Chain framework that extends the traditional RFM model with a stability dimension. This integration enables dynamic analysis of customer transitions and provides more accurate assessments of CE. Compared to traditional static models, the approach delivers deeper insights into customer behavior and supports more informed managerial decision-making.
