Churn Prediction Indicators AI Prompt

Identify the leading behavioral indicators that predict user churn before it happens.

User behavior data: {{behavior_data}}
Churn definition: {{churn_definition}} (e.g. no activity for 30 days, subscription cancelled)
Observation window: {{observation_window}} (behavioral features measured in the N days before churn)

1. Feature engineering for churn prediction:
   Compute these behavioral features for each user in the observation window:
   - Login frequency: sessions per week
   - Days since last active
   - Core action completion rate: % of sessions where {{core_action}} was completed
   - Feature breadth: number of distinct features used
   - Engagement trend: comparing last 7 days vs prior 7 days
   - Support contacts: number of support tickets or error events
   - Billing events: failed payments, plan downgrades

2. Univariate analysis:
   For each feature, compare the distribution between:
   - Users who churned within {{horizon}} days
   - Users who did not churn
   Compute: mean, median, and statistical significance of the difference (Mann-Whitney U test)

3. Predictive ranking:
   - Which features show the largest and most statistically significant difference between churners and non-churners?
   - Rank features by predictive power (use AUC of a simple logistic regression per feature)

4. Early warning thresholds:
   - For the top 3 features: what threshold value separates high-churn-risk from low-churn-risk users?
   - Example: users with > 14 days since last login have a 3x higher churn rate than average

5. Churn risk segmentation:
   - Combine the top 3 indicators into a simple churn risk score (Low / Medium / High)
   - What % of users currently fall into each risk tier?
   - What intervention should each tier receive?

Return: feature importance table, threshold analysis, risk tier definitions, and intervention recommendations.