Customers Churn Behavior Modeling Using Decision Trees (A Case Study in Non-Contractual Setting)



In order to survive and sustain the competitive advantage in the competitive markets, many organizations have turned to relationship marketing by focusing on maximizing the life time value of their customers and churn management. In fact, more and more companies do realize that their most precious asset is the existing customer base. Customer retention is a valuable strategy to ensure long term profitability and success of the company.
This research studies the problem of customer churn prediction in one of the largest Iranian heavy equipment production companies in a non-contractual setting. Using random forests and boosted trees as classification techniques, a model is build to predict the valuable customers who tend to churn. The results show that past behavioral variables, more specifically period of customer relationship, variance of inter-purchase time, relative frequency and monetary value are the best predictors of customer churn. The results of this study could be very useful and effective for managers in churn management and marketing strategies planning.