dragoon0x Cohort Retention Analysis
Understand retention by isolating who joined when and tracking what they did after.
How to use
/cohort-retention-analysisApply cohort analysis constraints to this conversation./cohort-retention-analysis <context>Analyze retention for the described product and data.
Constraints
Why Cohorts Matter
- Aggregate retention numbers lie. A growing product masks terrible retention with new users.
- MUST isolate groups by when they joined and track behavior over time
- NEVER report retention as a single aggregate number without cohort breakdown
Cohort Types
- Time-based: weekly signup cohorts (fast products), monthly (most SaaS), quarterly (long sales cycles)
- Behavioral: users who completed onboarding vs. didn't, used Feature X vs. didn't
- Segment: by plan tier, company size, acquisition channel, geography
- SHOULD use at least two cohort types for any retention analysis
Retention Metrics
- User retention: did they come back? (Day 1, 7, 14, 30, 60, 90)
- Revenue retention: did they keep paying? NRR includes expansion; GRR only contraction and churn.
- Activity retention: did they do the core action? (Define what "active" actually means)
- MUST pick the metric that matches your product type. "Active" for a chat tool is different than a tax tool.
Curve Shape Interpretation
- Steep early drop then flattens: normal. Focus on getting more people past the initial drop.
- Gradual continuous decline: dangerous. Product delivers diminishing value over time.
- Flattens then drops again: something triggers later churn. Investigate the inflection point.
- Newer cohorts retain better: product or onboarding is improving.
- Newer cohorts retain worse: product-market fit may be weakening or user quality declining.
Diagnosing Drop-Off Points
- Day 0-1 drop: onboarding too complex? Aha moment not reached?
- Day 1-7 drop: no reason to return? No habit loops?
- Day 7-30 drop: novelty wore off? Hitting plan limitations?
- Day 30+ drop: product not delivering ongoing value? Competitor pulling them away?
- MUST compare cohorts to find what drives differences, not just observe the drops
Anti-Patterns
- Reporting DAU/MAU without cohort context
- Averaging retention across segments that behave completely differently
- Measuring only one retention metric when multiple views tell different stories
- Ignoring that improving one stage can sometimes hurt a later stage