Analyze this funnel data and find the biggest leak.
Inputs:
- Funnel export: {{paste CSV or describe stages + counts}}
- Time period: {{e.g. last 90 days}}
- Comparison period (optional): {{e.g. same 90 days last year, or prior quarter}}
- What you already suspect: {{paste — or "no priors"}}
Output:
## Conversion table
| Stage | Entered | Completed | Conversion % | Drop-off | Change vs. comparison |
|---|---|---|---|---|---|
## Biggest leak
The stage with the highest drop-off in absolute users (not just %). Name it. Quantify.
## 3 hypotheses for the leak
Ranked by likelihood × evidence. For each:
- Hypothesis (one sentence)
- The signal that points to it
- The signal that would rule it out
- The cheapest test to confirm (within 1 week)
## Second-order leaks
The 1–2 other stages that look fine in % terms but where the absolute loss is meaningful. Don't skip these.
## What's normal here vs. unusual
If you have a comparison period: which conversion rates are normal seasonal/weekly variance and which are signal. Distinguish them.
## What I'd watch for the next 30 days
- The leading indicator that would tell us the fix is working
- The leading indicator that would tell us we mis-diagnosed
Hard rules:
- Always show absolute numbers, not just %. A 10% drop on 1M users is different from 10% on 100.
- Flag anything that looks like a data pipeline issue (sudden zeros, identical numbers across stages) before treating it as user behavior
- If the funnel definition is ambiguous (e.g. "active user" could mean many things), say sofunnelconversionproduct-analytics