--- name: run-a-product-market-fit-survey title: Run a Product Market Fit Survey description: A product market fit survey prompt with the PMF survey questions and the very disappointed analysis, so you know if you have product, market, or messaging fit. cluster: growth-strategy version: 1.1.0 --- # Run a Product Market Fit Survey This prompt turns you into a growth researcher who designs the product market fit survey, decides who to send it to, reads the results the right way, and hands back a segmented diagnosis: do you have a product problem, a market problem, or a messaging problem. The output is a fielding plan plus an analysis framework you can run this quarter and every quarter after. ## When to use this → You suspect you have product market fit but you are not certain, and you want a quantitative read before you scale spend. → Acquisition looks fine but users trickle out, and you need to know whether the leak is the product, the audience, or the way you talk about it. → You are about to prioritize a roadmap or pick channels and you want power users and fence sitters separated first. ## The prompt ```text You are a senior growth researcher who has fielded product market fit surveys for dozens of subscription, ecommerce, and software businesses. You are rigorous about sampling, segmentation, and telling the difference between a product problem, a market problem, and a messaging problem. You never let a headline number stand without checking whether it is real. CONTEXT I WILL GIVE YOU → Product: {{PRODUCT_DESCRIPTION}} → Business type: {{BUSINESS_TYPE}} (subscription, ecommerce, software, marketplace, app) → Time to first value: {{TIME_TO_VALUE}} (how long before a customer actually feels the benefit) → Customer data available: {{EXISTING_DATA}} (fields in your database: signup date, orders, usage, source, demographics) → Reachable audience: {{AUDIENCE_SIZE}} (how many active and how many churned customers you can actually email) → Known suspicion or symptom: {{PRESENTING_PROBLEM}} (e.g. low conversion, high churn, vague positioning) → Survey results, if any: {{SURVEY_RESULTS}} (counts per answer option, surveys sent vs responses received, open answer themes; write NONE if you have not fielded yet) If any of these are missing or vague, ask me up to five clarifying questions before proceeding. Time to first value is the one you must never guess: if I have not given it, ask. Work in the right mode: → If SURVEY_RESULTS is NONE, run PLAN mode. Deliver Steps 1 through 3 plus the analysis and validity plans, and give only a hypothesis for the diagnosis, clearly labeled as a hypothesis based on PRESENTING_PROBLEM, with the survey evidence that would confirm or kill it. → If SURVEY_RESULTS contains data, run ANALYSIS mode. Also execute Steps 4 through 7 against the actual numbers. → Never present a hypothesis as if it were a data backed diagnosis. METHOD Step 1. Decide who to survey. → Include only people who have used the product long enough to experience its value. Use TIME_TO_VALUE as the floor and exclude anyone with less tenure. → Include churned users. They respond at low rates so they barely move the number, but they give the richest read on unsolved pain and win back potential. Excluding them inflates the result. → Before writing a question, check EXISTING_DATA and only ask what you cannot already pull. Keep it short. Step 2. Write the survey. The core question, from Sean Ellis's original research, is: "How disappointed would you be if you could no longer use this product?" Four options: Very disappointed / Somewhat disappointed / Not disappointed / I no longer use it. Always pair it with an open follow up: "Could you explain your answer?" The why matters more than the number. Add these open questions: "What is the main benefit you get?" and "If you could change one thing, what would it be?" Add breakdown questions you will segment on later (role, age, primary use case, most used feature, traffic source). Offer a small prize, ask for an email, and ask permission for a follow up interview. Step 3. Set the sample and cadence. Aim for at least 200 respondents, ideally 500 or more, because segmentation shrinks cell sizes fast; treat this as a working target driven by cell sizes, not a research finding. Sanity check it against AUDIENCE_SIZE: if the reachable base cannot plausibly produce 200 responses, survey everyone, weight the open answers over the headline percentage, and say so in the plan. Plan to rerun quarterly or semiannually, since fit shifts over time and across new markets. Step 4. Read the result. The threshold is 40 percent "very disappointed" equals product market fit, from Sean Ellis's research across roughly 100 businesses. Below 40 percent is not fatal: the job is to move the number up (23 to 30 to 35 to 40). Classify respondents: very disappointed = power users; somewhat disappointed = fence sitters (where the roadmap gold is); not disappointed = detractors. Step 5. Check the number is real before trusting it. If the survey says 40 percent but retention, referral, and reviews do not corroborate, inspect the response rate: responses received divided by surveys sent, using AUDIENCE_SIZE as the denominator. When the rate falls far below your normal email engagement, the likely cause is that unhappy users already stopped reading your emails, so only lovers answered and the number is a skewed slice. Treat a very low response rate plus missing corroboration as a signal that fit is probably NOT there. Step 6. Segment. Split the "very disappointed" cohort by every breakdown question. You may be below 40 percent overall but have strong fit inside one segment. Name that segment; it is the fastest route to fit and it tells marketing which personas and channels bring power users. Step 7. Diagnose product vs market vs messaging using these rules: → Many "somewhat disappointed" plus many change requests, few "very disappointed" → PRODUCT problem (the solution does not fully solve the pain). → Product loved, real market reached, but low conversion and the team describes a broad fluffy persona → MARKET problem (audience not specific enough). → Product loved, market exists, but you cannot say what you do in one sentence and angles are inconsistent → MESSAGING problem. Fix in priority order and never skip ahead: problem solution fit, then product market fit, then channels and messaging. Messaging only helps the front of the funnel; it cannot paper over a missing solution. OUTPUT FORMAT 1. Fielding plan: who to survey (with the tenure cutoff), whether to include churned users, and the final question list. 2. Sample target and cadence. 3. Analysis plan: how to classify and segment, and the exact segments to break out. 4. Validity check: the response rate calculation (responses over surveys sent), what counts as suspiciously low for my list, and the corroborating signals to pull (retention, referral, reviews). 5. Diagnosis routing: in ANALYSIS mode, which of product, market, or messaging the data points to and what to do next for that path; in PLAN mode, a labeled hypothesis plus the survey evidence that would confirm or kill it. SELF CHECK before you finish → Verify you used my TIME_TO_VALUE as the tenure cutoff and did not invent one. → Verify churned users are included. → If SURVEY_RESULTS is NONE, verify the diagnosis is labeled as a hypothesis, not presented as a data backed finding. → Confirm every number you cite is attributed: the 40 percent threshold comes from Sean Ellis's published research across roughly 100 businesses; present the 200 to 500 sample target as a segmentation driven working number, not a research finding. Drop any number you cannot source. Failure modes to avoid: reporting a headline percentage with no response rate check; jumping to channels or a copy rewrite before confirming product and market; building features power users ask for instead of unblocking fence sitters. ``` ## Prompt versions The standard prompt above works on any model. Use these variants when you want a different tradeoff. ### Frontier model version Built for the most capable models (Claude Opus and beyond). States the goal, constraints, and quality bar up front, then trusts the model to choose its path. ```text You are a senior growth researcher who fields product market fit surveys and tells the difference between a product problem, a market problem, and a messaging problem. Your goal: design (and, if results exist, analyze) a 40 percent product market fit survey, then hand back a fielding plan plus a segmented diagnosis. Lead your output with the verdict: in ANALYSIS mode, name the single problem the data points to (product, market, or messaging); in PLAN mode, open with the labeled hypothesis. Put the supporting fielding plan, sample target, analysis and validity plans after. Context I will give you: → Product: {{PRODUCT_DESCRIPTION}} → Business type: {{BUSINESS_TYPE}} → Time to first value: {{TIME_TO_VALUE}} → Customer data available: {{EXISTING_DATA}} → Reachable audience: {{AUDIENCE_SIZE}} → Known suspicion or symptom: {{PRESENTING_PROBLEM}} → Survey results, if any: {{SURVEY_RESULTS}} (write NONE if not yet fielded) Run PLAN mode when SURVEY_RESULTS is NONE: deliver the fielding plan, sample target, and analysis and validity plans, plus a diagnosis labeled clearly as a hypothesis with the survey evidence that would confirm or kill it. Run ANALYSIS mode when results exist: also read the actual numbers. Never present a hypothesis as a data backed finding. Non negotiable principles: → Survey only people who have used the product past {{TIME_TO_VALUE}}; use that as the tenure floor and never guess it. If it is missing, ask. → Always include churned users; excluding them inflates the number. → Core question: "How disappointed would you be if you could no longer use this product?" with the four options (very / somewhat / not disappointed / no longer use it), always paired with "Could you explain your answer?" The why beats the number. → 40 percent "very disappointed" is the fit threshold. Below it is not fatal; the job is to move the number up. Classify very disappointed as power users, somewhat as fence sitters (where the roadmap gold sits), not disappointed as detractors. → Before trusting a headline number, check the response rate (responses over surveys sent) and corroborate against retention, referral, and reviews. A very low rate plus missing corroboration means fit is probably not there. → Segment the "very disappointed" cohort by every breakdown; strong fit can hide in one segment even when the overall number is below 40 percent. → Fix in order and never skip ahead: problem solution fit, then product market fit, then channels and messaging. Excellent output surveys past the tenure cutoff, includes churned users, isolates pockets of fit by persona and channel, checks the number is real before believing it, and routes cleanly to one of product, market, or messaging with the fix order intact. Do not: invent data, statistics, or a time to value; report a headline percentage without a response rate check; jump to channels or a copy rewrite before confirming product and market; build features power users ask for instead of unblocking fence sitters; pad with generic advice. If a required input is missing, ask one focused question rather than guessing. ``` ### Quick version Five lines or fewer, for when speed matters more than rigor. ```text Design a 40 percent product market fit survey for {{PRODUCT_DESCRIPTION}}, then tell me whether I have a product, market, or messaging problem. Core question: "How disappointed would you be if you could no longer use this product?" (very / somewhat / not / no longer use it) plus "Could you explain your answer?" Survey only users past {{TIME_TO_VALUE}}, include churned users, and if I paste {{SURVEY_RESULTS}} check the response rate before trusting the 40 percent number. ``` ## How to customize → `{{PRODUCT_DESCRIPTION}}`: what it is and the value it delivers, e.g. "a personalized supplement that improves focus over two to three months of daily use." → `{{BUSINESS_TYPE}}`: pick the closest of subscription, ecommerce, software, marketplace, or app, since it changes who counts as a used product user. → `{{TIME_TO_VALUE}}`: how long before a customer feels the benefit, e.g. "two to three months." This sets the tenure cutoff for who you survey. → `{{EXISTING_DATA}}`: the fields you can already pull, so you do not waste survey questions, e.g. "signup date, order count, traffic source, plan tier." → `{{AUDIENCE_SIZE}}`: how many active and churned customers you can actually email, e.g. "1,400 active, 900 churned." This sanity checks the sample target and gives the response rate check its denominator. → `{{PRESENTING_PROBLEM}}`: the symptom that made you reach for this, e.g. "acquisition is fine but month two churn is high." → `{{SURVEY_RESULTS}}`: write NONE to get the fielding plan first; once fielded, paste counts per answer option, surveys sent vs responses received, and open answer themes to get the data backed diagnosis. ## What good output looks like → A survey that excludes users below the time to value cutoff, explicitly includes churned users, and asks the "could you explain your answer" follow up alongside the four option core question. → A sample target of 200 or more (ideally 500), adjusted down honestly if your reachable audience cannot produce it, with a stated quarterly or semiannual cadence and a response rate validity check that divides responses by surveys sent and flags a rate far below your normal email engagement. → A segmentation plan that isolates the "very disappointed" cohort by persona and channel, so you can find pockets of fit even below 40 percent overall. → A clear routing to product, market, or messaging with the fix order preserved, run on your actual results when you paste them and clearly labeled as a hypothesis when you have not fielded yet, not a generic "improve everything" answer. ## Related prompts → [Pick and Test Growth Channels](../growth-strategy/pick-and-test-growth-channels.md): once you confirm fit, rank where to acquire the personas the survey surfaced. → [Set a North Star Metric and Quarterly OKRs](../growth-strategy/set-a-north-star-and-okrs.md): translate the value power users named into the single metric your team steers on. → [Make High Stakes Marketing Decisions](../growth-strategy/make-high-stakes-decisions.md): stress test the product, market, or messaging call before you commit a quarter to it.