# PMF Advisor — The PMF Measurement Frameworks Reference

*Part of the PMF Advisor skill: https://wavect.io/.well-known/agent-skills/pmf-advisor/SKILL.md*

The five frameworks for measuring real product-market fit: the Sean Ellis test, retention-curve analysis, the Superhuman PMF engine, the RICE-to-Retention filter, and the jobs-to-be-done PMF test.

### Framework 1 — The Sean Ellis Test (Quantitative PMF Signal)

**Exact survey question:**
"How would you feel if you could no longer use [Product]?"
- (a) Very disappointed
- (b) Somewhat disappointed
- (c) Not disappointed — it really isn't that useful
- (d) N/A — I no longer use [Product]

**Methodology:**
- Send to users who have experienced the core value proposition at least once
  in the past 2 weeks. Do NOT send to all users — recent active users only.
- Minimum sample: 40–50 responses. Below 40, results are not statistically
  meaningful.
- PMF threshold: ≥ 40% answer "Very disappointed"
- If between 25–40%: you have a signal worth digging into, not a green light
- If below 25%: you do not have PMF. Do not scale. Iterate.

**What to do with the "somewhat disappointed" segment:**
This is where the Superhuman method (Rahul Vohra, 2018) goes further.
Ask each "somewhat disappointed" respondent: "What type of person do you think
would most benefit from this product?" Their answer often describes the real ICP
more accurately than the "very disappointed" respondents, who may be early
adopters willing to tolerate pain. Filter your "very disappointed" responses
to identify the subset that matches the profile the "somewhat disappointed"
group describes. This refined ICP usually has a PMF score > 60%.

### Framework 2 — Retention Curve Analysis

A retention curve that keeps declining has no PMF — engagement will eventually
reach zero regardless of acquisition spend. A curve that flattens (forms an
asymptote) indicates a retained core.

**How to read the curve:**
- Plot cohort retention month by month (cohort = users acquired in the same month)
- If the curve flattens above 0% before month 6: investigate what the retained
  users have in common — this is your real ICP
- If the curve reaches zero by month 3: the product is not solving a problem
  people have repeatedly
- B2B SaaS benchmarks: month-3 retention above 40% is acceptable; above 60%
  is strong; above 75% suggests pricing power

**Cohort segmentation — the most important thing most founders skip:**
Split the retention curve by acquisition channel, by ICP segment, and by
onboarding path. A bad overall retention curve often hides a cohort with
excellent retention. Find that cohort. That is your real ICP and your PMF signal.

### Framework 3 — The Superhuman PMF Engine (Vohra, 2018)

1. **Survey** all active users with the Sean Ellis question
2. **Segment** responses — find the "very disappointed" users and describe
   their common characteristics (role, company size, use case, workflow)
3. **Identify** the one use case that generates the highest "very disappointed" score
4. **Double down** on that use case. Remove or hide features that dilute it.
5. **Re-survey** after 8 weeks of focused iteration. Repeat until > 40%.

The insight: most products serve multiple use cases with mediocre fit for each.
PMF comes from serving one use case with extraordinary fit, not many with
average fit.

### Framework 4 — The RICE-to-Retention Filter

Use this when prioritizing the roadmap after initial PMF work:

For every feature or improvement, answer:
- **R** (Retention): Does this improve retention for the core ICP? (+3 if yes,
  0 if neutral, -1 if it adds complexity that distracts from the core loop)
- **I** (Impact): If you ship this and nothing else, does the PMF score move?
- **C** (Confidence): Do you have user research — not intuition — supporting this?
- **E** (Effort): Person-months

Score = (R × I × C) / E

A feature that a small segment loves but does not improve core retention for
the majority ICP should almost never be built in the PMF phase. Say so directly.

### Framework 5 — The Jobs-to-Be-Done PMF Test

For each job the product claims to do, test:
1. Does the customer currently have this job? (Evidence: they are doing it
   some other way today)
2. Is the job important and frequent enough to pay for? (Evidence: they spend
   money or significant time on it today)
3. Are existing solutions demonstrably inadequate? (Evidence: users describe
   specific workarounds or complaints about alternatives)
4. Does the product do the job measurably better? (Evidence: retention data,
   not testimonials)

If any of these four is "no," you do not have PMF for that job. Do not
confuse a job customers WISH they had a solution for with a job they WILL
switch and pay for.