Bridging Machine Learning and Strategic Decision-Making: A Framework for Customer Churn in Banking

Abstract

Customer churn management has become increasingly important in modern customer relationship strategies. Among the main applications of machine learning and artificial intelligence in this field are customer churn prediction and model interpretation. Research on prediction models is already well developed, and work on interpretation methods to address specific business problems is steadily growing. However, major challenges remain, such as overly complex models, unverifiable recommendations, and high technical barriers for users. This study examines customer churn in the banking industry as a case example. It introduces the The Simplicity, Goal orientation, and Verifiable solutions (SGV) framework, which combines simplified model construction, goal-oriented interpretation, and verifiable recommendations. The framework is validated using SHAP explanations. The proposed interpretation method analyzes and visualizes the relationship between the target outcome and key variables through simulated data, providing an integrated link between interpretation, findings, and actionable strategies. SGV framework offers a practical toolkit for practitioners across industries. It enables effective use of machine learning and AI for in-depth data analysis, reduces the entry barrier for non-specialists, and ultimately contributes to improved productivity.