Zansa › Answers › Forecasting and data science
How do I know if a churn prediction model is any good?
Test it on later data and check that the customers it ranks as high-risk really leave more often than others. Useful measures are precision, recall, lift or capture rate in the top-ranked group. Most importantly, check whether acting on the predictions reduces churn compared with a control group.
Accuracy alone is misleading when few customers churn: a model that predicts nobody leaves can look highly accurate.
Practical checks:
- Rank customers by risk and measure churn in each decile.
- Compare with a simple rule such as days since last purchase.
- Check whether predictions are well calibrated, meaning a 30 percent risk really means about 30 percent.
- Review which factors drive predictions for sense.
A model is only valuable if the business can intervene effectively on the flagged customers.
Want this applied to your business? Zansa builds dashboards and forecasts for small and mid-sized companies worldwide. Start with a free intro call.
Book a call