pmf
This skill should be used when the user asks about product-market fit — measuring it, testing it, or diagnosing weak traction — e.g. "do we have PMF", "product-market fit check", "design a Sean Ellis
SKILL.mdventure/skills/pmf/SKILL.md
GitHubPMF Assessment — Measure, Score, Decide
Answer the scariest question honestly: does anyone need this enough? Two jobs depending on what exists: design measurement (no data yet) or assess and decide (data exists). The output is a decision aid — strong/weak/no fit, why, and what to do about it.
Shared conventions (venture suite)
- Workspace: locate
ventures/in the current working directory (ask once and create if missing). All outputs go toventures/<slug>/<NN-module>/. - Templates are read-only. Files under a skill's
references/live inside the installed plugin and are shared by every venture. Never write to them. "Fill template X" always means: copy it intoventures/<slug>/<NN-module>/under a dated name and fill the copy. - Read
VENTURE.mdfirst. When done: update its Status table and append key findings to "Key facts & numbers" (format:- [module, YYYY-MM-DD] finding (source, confidence)). - Language: converse in the user's language. Write formal deliverables (reports, canvases, battlecards) in the
report_languageset in VENTURE.md. Persian reports keep standard English business terms as-is (TAM, SAM, SOM, CAC, LTV, GTM, churn, retention...). - Research integrity: cite every external claim (source + date). Mark every estimate with confidence (high/med/low). Never fabricate numbers — write "no data found" instead.
- Chain: end by updating VENTURE.md and suggesting the logical next module.
Workflow
1. Establish what data exists
Ask the user what they have (don't assume): user counts, retention by cohort, usage frequency, churn, NPS, Sean Ellis survey results, revenue/renewals, qualitative feedback, support logs. Also read 01-discovery/ and "Key facts" for prior signals.
- Little/no data, or pre-launch → Mode A: design measurement.
- Real usage data → Mode B: assess.
2A. Mode A — Design the measurement plan
Produce 04-pmf/pmf-measurement-plan-YYYY-MM-DD.md containing:
- Sean Ellis survey customized to the product from
references/sean-ellis-survey.md(translated to the users' language — survey language = customers' language, regardless of report language), with delivery and sampling instructions (who to ask: people who used the product ≥2× recently; minimum n≈30–40). - Instrumentation shortlist: the 3–5 metrics worth tracking for this business type (from
references/signal-benchmarks.md) and how to capture them cheaply. - Decision date: when to come back and run Mode B.
Pre-launch with zero users → say plainly that PMF can't be measured yet and route to /venture:customers (problem validation comes first).
2B. Mode B — Assess
- Score the scorecard — copy
references/pmf-scorecard.mdto04-pmf/pmf-scorecard-YYYY-MM-DD.mdand fill the copy, usingreferences/signal-benchmarks.mdfor thresholds by business type (B2B SaaS retention ≠ consumer app retention). Every cell: value, benchmark, verdict, confidence. Missing metric → "not measured" (that's signal too). - Read the qualitative texture: what do the most-retained users have in common (segment, use case)? Sean Ellis "very disappointed" users — who are they? Pull quotes.
- Stress-test before verdict — mandatory: launch the
venture:devils-advocateagent with the draft scorecard + your tentative verdict. Prompt it to attack: data quality, survivorship bias, vanity metrics, segment-mixing, wishful benchmarks. Address every objection in the report (accept, rebut with evidence, or downgrade confidence). - Verdict & diagnosis in
04-pmf/pmf-assessment-YYYY-MM-DD.md:- Strong fit → protect it: identify the fit segment precisely, route to
/venture:gtmto pour fuel. - Weak/partial fit → diagnose the gap with this ladder, in order (each step cheaper to fix than the next): ① wrong segment being acquired → retarget; ② right segment, wrong value prop/messaging → reposition; ③ right promise, product under-delivers → fix activation/core loop; ④ problem isn't painful enough → pivot territory. Recommend the specific experiment to confirm the diagnosis.
- No fit → say it kindly but plainly; recommend pivot directions grounded in any positive signal found (who did retain?), or a structured stop. Route to
/venture:reviewfor the kill/pivot call.
- Strong fit → protect it: identify the fit segment precisely, route to
3. Close the loop
Update VENTURE.md (Status, key facts: PMF score + verdict + fit segment; riskiest assumptions revised). If verdict is weak/none, mark GTM module stale if it exists — scaling plans built on weak PMF are invalid.
Quality bar
- Never inflate. The user's hopes are not data. A kind, evidenced "not yet" saves them months.
- Benchmarks must match business type AND geography reality (emerging-market consumer benchmarks differ; note it).
- Retention curve flattening > any single number. NPS alone is never sufficient evidence.