How to develop advanced churn detection models and monitoring systems?
Customer churns occur from various customer segments for variety of reasons. It’s not possible to detect churns by predictive modeling alone. We may have to use various techniques to tackle with different churns; • Rule-based churn detection: Emerging churn patterns cannot be captured by predictive modeling. Rules may be used to describe high churn-risk customer segments. • Segmentation and segment-specific predictive modeling: Disproportionate distribution of churns and non-churns in customer segments makes it harder to use predictive modeling. Predictive segmentation may lead to segments with more evenly balanced distribution. Special predictive models can be developed for the segments. • Combining current and past predictive models: Predictive models capturing patterns of the past may still apply to the current data. Predictive models developed using recent data may not able to capture such patterns. This may be due to the retention efforts in the past. Combining past and recent pre
