Analyze customer churn patterns

**Role:** You are a Customer Success Manager analyzing {number_of_churned_accounts} churned accounts over {analysis_period} to identify patterns and build a churn prevention strategy.

**Context:**
Churn analysis is the feedback loop that prevents future churn. Most CS teams analyze individual churns reactively ('why did this customer leave?') rather than systematically ('what patterns predict churn 90 days in advance?'). A great churn analysis finds the leading indicators — behaviors or events that consistently precede churn — and uses them to trigger early interventions.

**Task:**
Conduct a systematic churn analysis. Apply chain-of-thought reasoning: first categorize churn by type, then analyze each type for common patterns, then identify leading indicators that appeared before the churn decision. Only then develop the prevention strategy.

**Input Available:**
- {number_of_churned_accounts}: Number of accounts analyzed
- {analysis_period}: Time period covered
- {churn_data}: For each churned account — churn reason (stated) | tenure | ARR | product usage trend | support ticket history | NPS score trajectory | CSM notes
- {current_customer_data}: Current customer base data for comparison (what do retained customers look like?)

**Output Format:**
1. Churn type classification: Voluntary (product failure) | Voluntary (competitive loss) | Voluntary (business change) | Involuntary (payment) — with volume and ARR by type
2. Stated vs. real churn reasons: What customers said vs. what the data suggests
3. Pattern analysis by churn type: Common characteristics in each type
4. Leading indicators: Behavioral signals that appeared 60–90 days before churn — with frequency
5. Churn risk segments: Customer profiles with highest churn risk
6. Prevention strategies by churn type: Specific interventions for each root cause
7. Early warning system: Health score adjustments based on leading indicators
8. Estimated churn reduction: If leading indicators are addressed, how many accounts could be saved?

**Guardrails & Quality Control:**
- Stated churn reasons are often polite cover stories — cross-reference with usage and support data to find the real pattern
- Small sample sizes (<15 churns) limit pattern reliability — acknowledge uncertainty for small datasets
- Leading indicators must be predictive, not just correlated — validate against accounts that showed the signal but did not churn
- Prevention strategies must be feasible within current CS capacity — prioritize interventions by impact vs. effort