Customer attrition within commercial banking operations constitutes an institutional friction point that directly erodes stable liability bases, impairs Net Interest Margins (NIM), and inflates customer acquisition costs (CAC). This comprehensive study establishes a data-driven framework to analyze and predict structural customer churn using empirical asset data from major European retail banking cross-sections. By evaluating multi-layered demographic parameters, capital ledger balances, and active financial product relationships, we construct an analytical ecosystem that isolates high-risk consumer archetypes. Exploratory Data Analysis (EDA) exposes significant macroeconomic risk concentrations: a pronounced age-related attrition corridor between ages 46 and 60, extreme geographic volatility within the German retail banking sector reaching a baseline churn rate of 32.44%, and an unexpected multi-product relationship paradox where customers cross-sold into three or more financial products demonstrate exponential closure velocities. Moving beyond passive descriptive analysis, we design and tune an ensemble machine learning pipeline utilizing a cost-weighted Random Forest architecture. The system successfully separates stable liability accounts from vulnerable profiles, achieving a robust validation Area Under the ROC Curve (ROC-AUC) score of 0.86, thereby offering financial institutions a proactive mechanism to safeguard institutional capital and meet macroprudential liquid stability targets.
Rai Fabian (Mon,) studied this question.