Statistical Methods in Customer Relationship Management

6.3 Chapter Summary

Customer churn can be considered a negative outcome of the customer retention process. The purpose of retention modeling is to understand the effects of marketing variables on the duration of customer lifetime, while the churn modeling is to estimate the possible causes which induce customer defection or the ending of customer lifetime duration. The modeling of churn is as simple as a probability modeling, whether customers will churn or not, and it can be estimated by a logit model. As churn modeling can also be viewed as a binary classification problem, techniques from the machine learning field have been adopted, such as neural networks, bagging and boosting classification trees, and cost-sensitive classifiers. When the churn data are in panel form, time series techniques can be adopted to correct the possible correlation of lifetime duration and customer defection behaviors. Hazard models (as used in the empirical example for this chapter) are also suitable for churn modeling and baseline hazard distribution, such as the exponential, Weibull, and log-normal, among others, have been selected in the model specification. The PHM is often used and it is able to estimate the effect of time-variant covariates on the hazard rate.