Run a retention analysis and diagnose the leak

Retention Leaks Live in the Funnel, Not the Aggregate

Aggregate retention curves conceal specific funnel drop-offs — the 70% of new users who never complete step 2, the 30% of week-1 activators who never return in week 4. Amplitude's retention research and Reforge's engagement writing both emphasize that retention diagnosis should start with cohort segmentation and end with a specific event-level drop-off — not with aggregate curves.

How the Run a retention analysis and diagnose the leak Prompt Works

The prompt defines the retention metric unambiguously, segments the leak across three cuts, and runs funnel analysis to locate the biggest drop-off. The intervention ranking at the end forces prioritization by expected lift × time-to-ship × evidence cost — which prevents the team from pursuing 10 interventions at once.

When to Use It

• Retention is below bar and needs diagnosis.
• Activation moved but retention didn't.
• Paid acquisition is scaling but retained users aren't.
• A new PM needs a baseline for retention interventions.
• A board is asking "what's the retention story."

Common Pitfalls

• Arguing without a defined retention metric. Daily vs. weekly active, core event vs. login — these choices change the analysis. Define first.
• Pursuing multiple interventions simultaneously. Without sequencing, you'll conflate signal across tests. Run one, measure, next.
• No hypothesis before intervention. Interventions without hypotheses fix symptoms. Name why users drop before shipping a fix.

Sources

• The Activation Metric — Reforge
• Retention, Engagement & Growth: The Silent Killer — Reforge
• Amplitude Blog — Amplitude
• Growth Loops — Reforge