Early Warning System for Potential Churners among Mortgage Customers

摘要

Retaining existing customer is a major task for many companies because cost to acquire new customers is higher than retaining existing customers. For mortgage business in Bank X, customer relationship management plays a big role to understand their customers' profile and churners so that suitable action can be done to retain their potential churners. Objectives of this study are (1) understanding their customers' profile and churners, (2) modeling potential churners using neural network model and (3) to deploy the model to identify potential churners. Data was divided into two parts: sampling (67,470 cases) and scoring (4,488 cases). Analysis was done using SAS Enterprise Miner. Dependent variable is churner/non churner while independent variables are balance and amount of loan, interest rate offered installment amount, loan performance, months in arrear, vintage, tenure, age, race and gender. Potential churners were identified as Malays, followed by Indian, other races and Chinese. Nonperforming loan and male customers tend to churn compared to performing loan and female customers. Younger customers with small loan amount, balance and monthly instalment, higher interest rate, have many months in arrears, longer vintage and tenure have higher tendency to churn from Bank X. Hence, Bank X should focus on the potential churners for their campaign to minimize the expenses of retaining existing customers by doing an effective campaign with high successful rate.