Dynamic Churn Prediction and Value-Aware Customer Retention in E-Commerce

Customer retention is essential for business profit in the cut-throat world of e-commerce. Nevertheless, classic customer churn prediction techniques tend to implement static batch learning models that require periodic retraining and are concentrated only on prediction accuracy and do not pay enough attention to the financial side of things. In this paper, an intelligent approach to identify potential at-risk customers based on a dynamic and value-centric model is introduced. The suggested method is based on using RFM criteria and customer behavioural patterns to give an insight into customer activity and purchasing behaviour. Incremental learning through stochastic gradient descent (SGD) decreases the need for full retraining by continually adjusting the model parameters as new observations arrive, thus solving the limitations associated with static learning. SHapley Additive exPlanations (SHAP), which illustrates the contribution of each feature in every possible outcome, increases the interpretability of the model. A combination of both probability of attrition and customer value is used by Value-Aware Prioritization in assessing the Expected Financial Loss, thus ensuring that the available retention efforts are focused on customers with more financial impact. Risk levels and customer value are then converted into suitable retention actions via a rule-based decision engine. By including adaptive learning, transparent decision support, and financially informed prioritisation, the suggested framework provides e-commerce companies with a workable solution to turn churn analysis into a continuous customer retention strategy.