Churn prediction: behaviors in the 14 days before cancellation

Analyze the behavioral patterns of subscribers who cancelled in the last 90 days. Compare their activity in the 14 days before cancellation vs. the same subscribers 30 days prior. Identify the top 5 behavioral signals that predict churn (e.g., declining session frequency, shorter watch times, drop in completion rate, browsing without playing, visiting account/billing pages). Rank signals by predictive power and flag which can be detected early enough to intervene.

Click to copy