- The document presents the results of a predictive model built to identify customers at high risk (80%) of churning from their services in the next two months. The top three predictors of churn were identified as days since last login, number of logins, and account age.
- A decision tree model was recommended over logistic regression due to its ability to provide a shortlist of high-risk customers with at least 67% accuracy, allowing for more cost-effective targeting. The top customers to focus on engaging have accounts less than 22 months old.
- To reduce churn, the company should prioritize engaging primary target customers identified by the model through phone calls for feedback, and send promotional emails to secondary targets about new features