Customer Intelligence Predictive Models Targeting Customer Attrition in Credit Unions

• Activity Number: 228
• Type: Invited
• Date/Time: Monday, August 10, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Marketing
• Author(s): Vyacheslav Lyubchich* and Yulia R. Gel
• Companies: University of Maryland Center for Environmental Science and The University of Texas at Dallas
• Keywords: attrition ; bank ; churn ; customer ; CART ; logistic regression

Abstract

Loss of clients, also known as "churn", is one of the key business issues for variety of companies, from telecommunication providers to insurance firms and banks. As companies pursue new customers through acquisition marketing efforts, existing customer attrition undermines their profitable growth. The rampant amount of information available in each of the companies allows to track individual customers and their peers and study their behavioral patterns to predict a possible churn. We used anonymized data on individual characteristics of bank customers, their account balances and recent account activity to identify the customers who are likely to leave the bank. Particularly, we employed CART (Classification and Regression Trees) and logistic regression to identify the variables, pertaining to individual customers and their peer networks, which exhibit the highest predictive power for assessing the future customer churn. The breaking points from CART analysis were employed to determine the 'risk groups' of customers, i.e. those who has a high propensity to leave the bank.