Predicting churn doesn't work (well)
- Predicting churn should be targeted with the actionable insight, not just binary results (churn or not)
- Churn is hard to predict with high accuracy
- customers may took long time to cancel
- Utility or enjoyment is a fundamentally subjective eperience
- Churn is normally rare in comparisons with retention, false positive predictions are common.
Reference: https://www.fightchurnwithdata.com/