Collection Model Development to Predict Which Customers Will Pay

This article addresses the challenges faced by banks in managing non-performing loans (NPLs) and optimizing collection strategies for delinquent customers. The study presents a systematic approach to optimize collection strategies, aiming to maximize the total probability of payment before litigation for each customer while minimizing the risk of customer churn. The study proposes the development of a collection scorecard framework to predict the effectiveness of various collection actions and their potential impact on customer churn. Using logistic regression models, the authors analyze historical data to identify key predictors of payment probability, incorporating action severity and capacity constraints into the modeling process. Future research directions include the development of a churn scorecard and validation of the proposed approach across different banking environments.