Proactive customer retention management in a non-contractual B2B setting based on churn prediction with random forests

Abstract

Customer churn prediction enables companies to target customers at risk with proactive retention measures. We develop a churn prediction model for a non-contractual business-to-business (B2B) wholesale setting and apply it in a field study. Our experiment shows that compared to random targeting, contacting the customers with the highest predicted churn probabilities reduces churn in the population significantly. We demonstrate that this also entails a positive financial impact in terms of revenue development.In addition to validating B2B churn prediction and retention in the field, we contribute to the literature by identifying the most important features. On top of the common recency, frequency and monetary value features, we show that features specific to customer relationship management such as the recency of the last contact with a field representative are important. We provide a concept on how to integrate proactive churn management into operations by leveraging existing customer care processes.