Customer Churn Diagnosis

**Role:** Analytics lead at a $20M ARR B2B SaaS. You've debugged 10+ retention curves and you know which patterns mean what.

**Context:** Cohort retention table: [paste table OR describe the curve shape]. Pricing tiers: [list]. ICP: [persona]. Recent product changes in the affected period: [list]. Onboarding flow changes: [list].

**Task:** Walk the data and propose 3 falsifiable hypotheses for the churn pattern.

1. Read the curve. Is it Month-2 cliff, gradual erosion, or shoulder-shape? Each pattern has different likely causes.
2. Propose 3 hypotheses ranked by likelihood given the curve shape + product context.
3. For each hypothesis: 1-paragraph explanation + the SQL query / cohort cut that would test it + what result would confirm vs refute.
4. Distinguish "product gap" from "wrong ICP" from "onboarding leak" — these have very different fixes.
5. Identify the lowest-cost test we could run THIS WEEK to start ruling hypotheses in or out.

**Constraints:**
- Show the math/curve-reading reasoning
- Never propose "improve onboarding" without a specific step in the flow
- No correlation claims without a falsifiable test
- Hypotheses must be ranked, not all equal

**Output format:** 3 hypothesis blocks · ranked · each with explanation + test query + confirm/refute criteria · plus "this-week test" recommendation.