measuring-product-market-fit（衡量产品市场契合度）

Measuring Product-Market Fit

Scope

Covers

• Measuring PMF using a triangulated signal set (survey + behavior + customer evidence)
• Running and interpreting the Sean Ellis "Very Disappointed" survey (overall + by segment)
• Reading retention curves / cohort retention as PMF evidence (and knowing when they mislead)
• Using reference-customer / advocacy signals as an additional PMF proxy
• Detecting PMF drift (market shifts, rising expectations, competitive resets) and setting a re-measurement cadence
• Special handling for marketplaces (measure PMF per side; focus on the "hard side" first)

When to use

• "Do we have PMF? For which segment?"
• "Run a Sean Ellis PMF survey and tell me what it means."
• "Build a PMF scorecard with retention + survey + references."
• "Our market shifted—did we lose PMF?"
• "We want a go/no-go signal for scaling growth spend or launching publicly."

When NOT to use

• You haven’t defined the problem/ICP yet (use problem-definition).
• You only need a survey instrument, not a full PMF measurement system (use designing-surveys).
• You’re deciding whether/how to pivot (use startup-pivoting) rather than measuring PMF signals.
• You need a product vision/strategy doc as the primary output (use defining-product-vision / ai-product-strategy).
• You already have PMF and need to optimize retention or engagement (use retention-engagement); this skill measures PMF, not post-PMF growth levers.
• You need to brainstorm or validate new startup ideas (use startup-ideation); this skill assumes a product already exists with real users.
• You want to define or refine a north-star metric for an established product (use writing-north-star-metrics); this skill uses metrics as PMF evidence, not as a metric-design exercise.

Inputs

Minimum required

• Product + category + current stage (pre-PMF / early PMF / growth / mature)
• Business model: B2B / B2C / marketplace (and, for marketplaces, which side you’re focusing on)
• Your current best guess at the target segment/ICP (and any meaningful segments)
• Definition of active user and the core value moment (the action that indicates value received)
• What data you can access: survey channels, product analytics, retention cohorts, revenue, qualitative feedback, reference customers/testimonials
• Time horizon and constraints (deadline, privacy/PII constraints, internal-only vs shareable)

Missing-info strategy

• Ask up to 5 questions from references/INTAKE.md, then proceed.
• If key inputs are missing, proceed with explicit assumptions and label confidence.
• Do not request secrets. If data includes PII, ask for redacted excerpts or aggregated fields.

Outputs (deliverables)

Produce a PMF Measurement Pack (Markdown in-chat; or as files if requested) containing:

1. Context snapshot (product, stage, decision, timebox, segments, constraints)
2. PMF measurement model (core value moment, active user definition, signal set, thresholds as heuristics)
3. Sean Ellis survey plan + results (sample definition, questions, response counts, "very disappointed" % overall + by segment, top benefits)
4. Behavioral evidence (retention/cohort summary + engagement frequency; instrumentation gaps + how they affect confidence)
5. Reference-customer / advocacy evidence (who is willing to vouch; quotes; counts vs heuristic targets)
6. PMF Scorecard (signals, targets, current state, confidence, evidence links/notes)
7. Diagnosis + action plan (PMF status by segment; top drivers; prioritized next actions/experiments)
8. Risks / Open questions / Next steps (always included)

Templates and checklists:

• references/TEMPLATES.md
• references/CHECKLISTS.md
• references/RUBRIC.md

Workflow (7 steps)

1) Intake + decision framing

• Inputs: User context; references/INTAKE.md.
• Actions: Confirm the decision (scale spend, launch, refocus ICP, pricing), the timebox, and the audience. Define "what will we do differently based on this?"
• Outputs: Context snapshot + measurement constraints.
• Checks: A stakeholder can answer: "What decision will this change by?"

2) Define the PMF measurement model (and segments)

• Inputs: Product + segment hypotheses; data availability.
• Actions: Define:

- The core value moment and active user definition - The segment(s) to evaluate (ICP + meaningful slices) - The signal set (survey + behavior + customer evidence) and what "good" looks like (as heuristics)

• Outputs: PMF measurement model + segment plan.
• Checks: Each signal has (a) a metric definition, (b) a data source, and (c) a limitation note.

3) Run the Sean Ellis PMF survey (must-have test)

• Inputs: Target population list (active users); distribution channel; references/TEMPLATES.md (PMF block).
• Actions: Draft and run:

- "How would you feel if you could no longer use?" (Very / Somewhat / Not disappointed) - Follow-up: "What is the primary benefit you receive?" (text) - Segment respondents (persona/ICP, use case, tenure) to find the "must-have" cohort

• Outputs: Survey plan + results table (overall + by segment) + top benefit themes.
• Checks: Sample definition is explicit; results include counts (n), not only percentages; major bias risks are listed.

4) Analyze behavioral evidence (retention + engagement)

• Inputs: Product usage data or best-available proxy; activation definition.
• Actions: Build a minimal behavioral picture:

- Cohort retention (or repeat usage/purchase) by segment and tenure - Retention curve shape (improving/flat/decaying) and interpretation - Engagement frequency vs the product’s natural cadence (daily/weekly/monthly)

• Outputs: Retention/engagement summary + confidence notes + instrumentation gaps.
• Checks: Retention is measured from a clear cohort start; analysis separates activation from retention.

5) Collect reference-customer / advocacy evidence

• Inputs: Customer list; CS/sales notes; reviews; testimonials.
• Actions: Identify users willing to vouch publicly/privately:

- B2B heuristic target: 6–8 reference customers - B2C heuristic target: 15–25 strong references/advocates - Capture the "why" (benefit) and the segment they represent

• Outputs: Reference evidence log + gaps by segment.
• Checks: References map to the intended ICP/segment; evidence is current (not from a different market era).

6) Synthesize into a PMF scorecard + diagnosis (by segment)

• Inputs: Survey + behavior + reference evidence.
• Actions: Triangulate signals to answer:

- Do we have PMF for any segment? Which one is strongest? - What are the top drivers of "must-have" value? - What’s blocking PMF for adjacent segments? - Are we at risk of PMF drift (market shift, expectations rising)?

• Outputs: PMF Scorecard + diagnosis narrative + confidence rating.
• Checks: Diagnosis is segment-specific and evidence-backed; "unknowns" are explicit.

7) Quality gate + action plan + cadence

• Inputs: Draft pack; references/CHECKLISTS.md and references/RUBRIC.md.
• Actions: Run the checklist + score with rubric. Produce:

- Prioritized next actions/experiments (what to change, how to measure impact) - A PMF re-measurement cadence + drift triggers - Risks / Open questions / Next steps

• Outputs: Final PMF Measurement Pack.
• Checks: Actions are concrete enough to execute next sprint/quarter; measurement plan includes owners and dates (if known).

Anti-patterns

1. Single-signal overreliance — Declaring PMF based solely on one metric (e.g., "40% said very disappointed, so we have PMF"). A Sean Ellis score without retention evidence and reference-customer signals is a partial reading. Always triangulate survey + behavior + advocacy.
2. Whole-company PMF fallacy — Reporting PMF as a company-wide yes/no instead of measuring per segment. You may have strong PMF with mid-market sales teams and zero PMF with enterprise IT. Segment-level conclusions are mandatory.
3. Biased survey population — Sending the Sean Ellis survey only to power users or recent sign-ups, then generalizing. The sample must reflect the target segment, and bias risks (survivorship, recency, self-selection) must be explicitly stated.
4. Confusing retention with PMF — Treating a flat retention curve as proof of PMF without examining whether users are getting the core value or are merely locked in (contractual, switching costs, habit without satisfaction). Separate activation from retention and check for satisfaction signals.
5. Ignoring PMF drift — Measuring PMF once and treating it as permanent. Markets shift, competitors launch, expectations rise. The pack must include a re-measurement cadence and drift triggers.

Quality gate (required)

• Use references/CHECKLISTS.md and references/RUBRIC.md.
• Always include: Risks, Open questions, Next steps.

Examples

Example 1 (B2B SaaS, early growth): "Use measuring-product-market-fit. Product: AI meeting notes for account executives. Segments: mid-market sales teams vs SMB founders. Data: 90-day cohorts + in-app survey. Decision: whether to scale paid acquisition next quarter. Output: a PMF Measurement Pack."

Example 2 (Marketplace, supply-first): "We’re building a caregiver marketplace. We have early demand, but supply is thin. Measure PMF for the supply side first using a PMF survey + retention proxies. Output a scorecard and a plan to strengthen the core value exchange."

Boundary example (insufficient inputs): "Do we have PMF?" Response: ask up to 5 intake questions (segment, active user definition, data sources, survey channel, decision), then produce a minimal PMF Measurement Pack with explicit assumptions and confidence limits.

Boundary example (redirect to retention-engagement): "We confirmed PMF last quarter with 48% very-disappointed and strong retention. Now we need to reduce churn in our freemium tier." Response: This is a post-PMF retention optimization problem. Use retention-engagement to diagnose churn drivers and design interventions. Re-run PMF measurement only if you suspect PMF drift (e.g., new competitor, market shift).