Optimal Churn Rate Reduction Under Information Spillovers

This paper investigates theoretically when it is optimal for a firm to retain old customers or to attract new ones in the presence of information spillovers. Under the former (retention) strategy, the firm makes a special offer to customers who are about to leave. With some probability, other (loyal) customers learn about this offer and hence, may have incentives to leave – this is what we call information spillovers that never emerge when the firm attracts new customers. We characterize the optimal offer and show when the retention strategy is best for the firm. To illustrate how ignoring information spillovers, as it is usually done in the empirical literature on the topic, distorts predicted churn rates, we apply 3 Machine Learning (ML) algorithms to simulated data. As our analysis reveals, ignoring information spillovers worsens the quality of the predicted customer churn rates by 26-41%. Keywords: churn prediction, information spillovers, ML models.