Customer Churn Models: A Comparison of Probability and Data Mining Approaches

As markets become increasingly saturated, astute companies acknowledge that their business strategies should focus on identifying those customers who are likely to churn (Hadden, Tiwari, Roy, & Ruta, 2007). Since net returns on investments for retention strategies are generally higher than for acquisitions, it is generally accepted that companies should concentrate their marketing resources to keep existing customers rather than to attract new ones (Colgate & Danaher, 2000). This calls for models capable of making accurate predictions about consumers’ behavior in a future time period. Such models should be able to specify which customers in a dataset have a higher probability to churn in a given future time period. Literature on churn modeling reveals that predictive models fall into one of two categories, namely probability modeling and data mining modeling. Although many studies from both of these streams have focused on developing models to predict and identify customer churn, to the best of our knowledge, none of them have compared the performance of these modeling approaches in terms of accuracy in identifying and predicting customer churn.