Drift Root Cause Report AI Prompt

Generate a structured drift root cause report when drift has been detected in this model.

Drift detected: {{drift_description}}
Model: {{model_name}}
Detection date: {{detection_date}}

The report should contain:

1. Executive summary (3 sentences):
   - What was detected, when, and how severe?
   - What is the estimated business impact if unaddressed?
   - What is the recommended immediate action?

2. Drift characterization:
   - Type: data drift / concept drift / label drift / combined
   - Severity: PSI scores, AUC degradation, or performance metric change
   - Onset: estimated date when drift began (from change point detection)
   - Scope: which features are most affected? Which user segments?
   - Trajectory: is the drift stable, accelerating, or decelerating?

3. Root cause investigation:
   - Timeline of events: deployments, data pipeline changes, external events near the drift onset date
   - Feature analysis: top 5 drifting features with their PSI scores and distribution visualizations
   - Upstream data quality: any anomalies in the data pipeline feeding this model?
   - External context: market events, seasonality, product changes that could explain the drift?

4. Impact assessment:
   - Estimated accuracy degradation: current performance vs baseline
   - Affected prediction volume: how many predictions per day are impacted?
   - Downstream business impact: estimated revenue, risk, or operational impact

5. Recommended actions (prioritized):
   - Immediate (< 24 hours): quick mitigations to limit damage
   - Short-term (< 1 week): retraining, threshold adjustment, feature fixes
   - Long-term (< 1 month): systematic fixes to prevent recurrence

6. Monitoring update:
   - What new tests or tighter thresholds should be added to catch this pattern earlier next time?

Return: complete drift root cause report template with all sections filled based on available data.