Customer may abandon the organization, despite high investments made by the organization in their prospection and retention, which demands diagnose and understanding. The objectives of this article are threefold: i) to model the probability of churn, which refers to the rate of customers who leave a relationship with an organization during a given period; ii) to validate the model; and iii) to describe possible variables that explain the abandon of customers. We used historical data from a large retail chain´s dataset, and applied logistic regression in a sample of 70.000 users of a private label credit card to calibrate the model. Sixteen variables (14 individual characteristics and two behavior variables) were used. The model was validated in a different sample of 30.000 customers from the same dataset, through Kolmogorov-Smirnov (KS) test and Receiver Operating Characteristic (ROC) curve, confirming the good prevision power of the model. Managerial implications and suggestion for future research are discussed based on customer relationship management field of research.
Churn; data mining; customer relationship management; consumer lifetime value; logistic regression
