Will I Stay or Will I Churn?

Sammanfattning: Big data is transforming the way we understand and predict user behavior. One of these areas is customer churn prediction in mobile apps. This quantitative case study investigates which variables predict churn in a mobile wellness app, for the first week of use. Also, this study investigates the predictive performance of a neural network, a logistic regression, and a rule-of-thumb. Literature on human psychology and behavior is applied to formulate hypotheses. In addition, literature on prediction models and heuristics is briefly presented. Subsequently, the hypotheses are tested on a sample of over 200,000 app users with over 40 million recorded actions. Additionally, a comparison of the predictive churn performance of a neural network, a logistic regression, and a rule-of-thumb is performed. The main findings propose that frequency of completed sessions is a key variable that predict churn in a mobile wellness app. Also, the neural network marginally outperformed the logistic regression and the rule-of-thumb in predicting churn.