Before a percent counts as fit
Product-Market Fit Survey — Write the Tripwire Before You Count “Very Disappointed”
A “very disappointed” percent is not a success score. The Sean Ellis product-market-fit survey is useful only when you write the tripwire first — the signal, the threshold you will honor, the calendar window, and the named action if you miss. Then you ask recent real users one question. The number does not predict your startup. Matching the page you dated before the first send does.
Last updated: October 4, 2026
Direct answer
What is the Sean Ellis product-market-fit survey?
It is one question, asked of people who have recently done real usage — not people who only signed up. Sean Ellis’s 5 April 2019 article “Using Product/Market Fit to Drive Sustainable Growth” writes the question as: How would you feel if you could no longer use [ProductName]? The options are Very disappointed, Somewhat disappointed, Not disappointed, and N/A I no longer use [ProductName]. The score is the percent who say very disappointed among valid current users — you exclude N/A. Ellis’s experience: around 40 percent very disappointed is when sustainable growth becomes possible. Treat 40 percent as a default tripwire threshold you may adopt or rewrite before the survey, not a law of nature and not a personal forecast. About 30 responses are directionally useful; 100+ made him more confident. This page writes that survey as a tripwire so Sunday cannot invent a kinder bar after the results arrive.
Key takeaways
What to remember
- Write signal + threshold + calendar window + named action before the first send. A percent invented after the results is a recap, not a test.
- Survey only people with recent real usage. Ellis’s example: took a ride, not just signed up. Define real usage for your product. Ideally last couple of weeks.
- Score = percent very disappointed among valid current users. Exclude N/A. Somewhat disappointed does not fire the tripwire.
- Ellis’s experience bar is around 40 percent. Adopt it or rewrite it before you survey. Do not treat it as a law, and do not invent other N thresholds.
- A Yibud score is not this percent. Scores do not prove PMF. Matching a pre-written tripwire does.
Why it matters
Waiting for retention cohorts, or scaling on a vibe, both waste the same months
I have watched founders treat a warm signup list as product-market fit, then buy ads because “people like it.” Nobody had used the product in the last couple of weeks. Nobody would have been very disappointed if it disappeared. Ellis wrote the survey so you do not have to wait for retention cohorts to mature — that wait can delay a scale decision by several months — and so you do not scale before fit. He is blunt: scaling growth before product/market fit is the fastest way to kill your startup. The survey is a leading indicator of whether early customers treat the product as a must-have, not a verdict that you will survive. CB Insights’ March 2026 analysis of 431 VC-backed companies that shut down since 2023 is useful here as a warning about pattern frequency — not as a personal forecast. “Ran out of capital” leads their list at 70 percent, and they are explicit that this is often the final cause of death. Among the 385 companies with identifiable reasons, more telling patterns include poor product-market fit (43 percent), bad timing (29 percent), and unsustainable unit economics (19 percent). Many shutdowns cited more than one reason, so those shares can add to more than 100 percent. A missing must-have is a fit problem. It does not show up in a Yibud score, a polite interview, or a smoke-test click. This page does not invent other percentages.
Vocabulary
What the four tripwire fields mean on this survey
The survey is not “ask anyone if they would miss us.” It is a dated page: who counts, what the signal is, where the bar sits, when the window closes, and what you will do if you miss. Each field needs a log a stranger could read.
1. Signal — percent very disappointed among valid current users
Ask Ellis’s exact question. Count only Very disappointed. Exclude N/A I no longer use [ProductName]. Somewhat disappointed and Not disappointed stay in the denominator of valid current users; they do not add to the numerator. Signups who never did real usage should not be in the sample.
2. Threshold — the bar you will honor
Ellis’s experience: around 40 percent very disappointed is when sustainable growth becomes possible. You may adopt 40 percent or rewrite it before the first send. Write the number on the page. A kinder bar after a 22 percent result is a recap.
3. Calendar window — start date and end date
The window is when you collect the sample you already named. Fourteen or twenty-one days from first send is a window. “We’ll keep surveying this quarter” is not. Silence inside the window does not raise the percent.
4. Named action — what happens if you miss
Stop paid scale, rewrite targeting / positioning / onboarding, or Continue to the next cheapest test. Hope is not an action. Ellis’s own example: one business moved from 7 percent to 40 percent by improving targeting, positioning, and onboarding — cite his article; do not invent other case numbers.
5. Very disappointed percent is not a Yibud score
Yibud scores come from a deterministic rule engine on a one-line idea plus five questions. This percent comes from recent real users answering Ellis’s question. Do not convert one into the other. Do not treat either as a success prediction.
The tripwire
Signal, threshold, window, action — written before the first send
The question tells you what you are asking. The tripwire tells you what happens if the must-have share does not clear the bar in time. Do not skip a field. Do not invent an N this page does not publish.
Field 1 — Observable signal
Percent very disappointed among valid current users (exclude N/A). A stranger could recompute it from the response log. “People seemed to love it” is not a signal.
Field 2 — Threshold
The percent you will treat as a pass. Default you may adopt: Ellis’s experience bar of around 40 percent. Or write a different bar before the send and say why. This page does not publish a law of nature.
Field 3 — Calendar window
Start date and end date. Ideally you only count people who used the product in the last couple of weeks — Ellis’s sampling rule. Stretching the window because the percent is low is a new page with a reason.
Field 4 — Named action
If the percent misses the bar by the end date: stop this scale bet, rewrite targeting / positioning / onboarding, or run a named cheaper test. Ellis used the open-ended answers from very disappointed users to learn who benefits and what the key benefit is.
Field 0 — Written before the first send
If the four fields appear after the first batch of answers, you wrote a recap. Date the page in the same sitting you define real usage. A kinder bar later is a new page with a reason — not an edit after 22 percent.
Not the same page
PMF survey vs Yibud scores, decision contracts, WTP, earlyvangelist, interviews, and reach tests
Several Yibud pages sit next to this one. Mixing them up turns a rule-engine score, a polite interview, or a click into “we have product-market fit.”
Yibud scores — not the Sean Ellis percent
The validation-score and score-calculator pages explain Startup MRI numbers from a deterministic engine. Those scores are flashlights on assumptions. They are not the percent of recent real users who would be very disappointed. Do not read a mid-band Yibud score as 40 percent, and do not read 40 percent as a Yibud GO.
Kill, pivot, pre-mortem, assumption, decision — contracts
Those pages write Stop, rewrite, rank failure modes, name one claim, or read a week of evidence you already have. This page supplies one evidence instrument those contracts can later read: a dated very-disappointed percent against a bar you wrote first. It is not a Stop / Pivot / Continue sitting.
Willingness-to-pay — money moving
That page’s signal is a deposit, pre-order, paid concierge, or priced checkout. This page’s signal is stated must-have among recent real users. A very-disappointed percent is not money moving. A checkout is not this survey.
Willingness-to-pay test →Earlyvangelist — who-filter, not a percent
That page’s signal is matching Steve Blank’s five characteristics. This page is whether recent real users would be very disappointed without the product. An earlyvangelist can later sit this survey. Matching the five is not a PMF percent.
Earlyvangelist validation →Design partner — scarce co-dev, not a survey
That page’s signal is staff time, data, workflow access, and weekly use as Customer Validation evidence. A survey response is not a signed partner. A design partner can later answer Ellis’s question. The labels are not interchangeable.
Design partner validation →Beta testers — usage testers, not a PMF percent
That page is how to find people who will run the unfinished product and tell you what broke. A beta tester can be quiet, unpaid, and still useful. This page is a usage-based survey against a pre-written bar. Do not recruit twenty testers and call two compliments 40 percent.
How to find beta testers →Mom Test, discovery, solution interview — talks, not this survey
Those pages are conversations: last-week behavior with no pitch, or a scripted artifact show after the problem is evidenced. This page starts when recent real usage already exists and you need a leading must-have read. An interview compliment is not very disappointed. A completed walkthrough is not this percent.
Smoke, fake-door, landing page — reach and intent before real usage
Those pages measure whether anyone reaches or clicks. They usually sit before real usage. Ellis’s sample is people who already did the real thing — took the ride, not downloaded the app. A waitlist is not a very-disappointed percent.
The sitting
How to write and run the PMF-survey tripwire in one sitting
Do this after real usage already exists — not after you have only a landing-page waitlist. Set a timer. Leave the ad account closed. If you have a co-founder, each of you writes a draft alone, then you keep the stricter bar.
- 1
Write the four fields before anyone is invited
Signal: percent very disappointed among valid current users (exclude N/A). Threshold: adopt Ellis’s around-40-percent experience bar, or write a different bar and the reason. Window: start date and end date. Action: stop scale, rewrite targeting / positioning / onboarding, or run a named cheaper test. If any field is blank, you are not ready to send.
- 2
Define real usage for this product
Ellis’s Uber example: took a ride, not just signed up. Write one sentence: the event that counts. “Opened the app” is usually too thin. “Published a digest a teammate opened,” “closed a job the customer accepted,” “ran the weekly report for a live account” — name the event. People who have not done it are not in the sample.
- 3
Limit the sample to recent real usage
Ellis: if they never used the product or have not used it in several months, you already know the answer is Not disappointed or N/A. Ideally survey people who used it in the last couple of weeks. Do not pad the list with signups to chase 30 responses.
- 4
Ask the exact question — then the open-ended follow-ups
How would you feel if you could no longer use [ProductName]? Options: Very disappointed / Somewhat disappointed / Not disappointed / N/A I no longer use [ProductName]. Then, from the very disappointed group, learn who benefits and what the key benefit is. Ellis used those answers to retarget, reposition, and redo onboarding.
- 5
Score only valid current users
Numerator: very disappointed. Denominator: everyone who is still a current user — exclude N/A. Ellis’s sample guidance only: about 30 responses are directionally useful; 100+ made him more confident. Do not invent other N thresholds or conversion rates. If you cannot reach 30 people with recent real usage, you may have a reach problem — that is a different page, not a reason to survey strangers.
- 6
Honor the action on the end date
If the percent misses the bar you wrote, do the named action the same day. Ellis’s article example: one business moved from 7 percent to 40 percent by improving targeting, positioning, and onboarding. That is his case, not a promise about your week. If you later want a kinder bar, that is a new page with a reason — not an edit after the send.
What you leave with
The PMF-survey contract (copy this)
If the sitting produced a mood, you did not finish. The page should have one filled tripwire and a real-usage definition a stranger could apply. Fill the brackets. Leave nothing as “users,” “soon,” or “we’ll know it when we see it.”
The tripwire
Product: [name]. Real usage (must have happened): [one event — not signup]. Recent window for “current user”: last [couple of weeks / your dates]. Question: How would you feel if you could no longer use [ProductName]? Signal: % Very disappointed among valid current users (exclude N/A). Threshold: [40% Ellis default, or your pre-written bar]. Sample guidance you will honor: directional at ~30; more confident at 100+ (Ellis). Window: [start]–[end]. Action if miss: [stop this scale bet / rewrite targeting-positioning-onboarding / run a named cheaper test].
If any bracket still says “users,” “traction,” or “later,” the contract is not written. Signups who never did real usage do not count toward the sample. Somewhat disappointed does not count toward the numerator.
What does not count as the signal
Not this signal: a Yibud score, a compliment in an interview, “I’d use that,” a waitlist, a smoke or fake-door click, a beta signup, a deposit or checkout, matching Blank’s five, a signed design partner, a completed solution-interview walkthrough. Optional later: those instruments have their own pages.
You may still take notes on praise or money. You may not let them fire this tripwire. Very disappointed among valid current users fires this tripwire.
Example tripwire (template — the counts are blanks or Ellis’s published guidance)
If fewer than 40% of valid current users (recent real usage; exclude N/A) are Very disappointed by day 21 → stop paid acquisition the same day and rewrite targeting / positioning / onboarding from the very-disappointed follow-ups. Directional read at ~30 responses; more confident at 100+ (Ellis). Do not add signups to chase N.
40 percent, ~30, and 100+ are from Ellis’s 2019 article, labeled as his experience. They are not Yibud results and not a claim about what usually happens in your category. This page invents no other N.
The fork
When the PMF survey is the expensive unknown — and when it is not
This page decides only whether “would recent real users be very disappointed without this?” is the unknown you should buy evidence for now. Interviews, money, scarce commitment, reach, and decision contracts have their own pages. Do not survey people who have never done the real thing.
Run the PMF survey first
Recent real usage exists. You can name the event. The expensive unknown is whether enough current users treat the product as a must-have. Write the tripwire. Send Ellis’s question. Do not start with ads.
Run WTP or a design-partner ask first
The unknown is whether they will pay this price — or put time, data, and workflow at risk. Use the WTP page or the design-partner page. A very-disappointed percent is not a charge and not a scarce-commitment pass.
Run interviews or a reach test first
Nobody has done real usage yet. Interviews are cheaper than a survey of signups. Smoke, fake-door, and landing pages measure reach and intent. Come back here once a named event has happened in the last couple of weeks.
When a PMF-survey tripwire is the expensive unknown
A real-usage event is named. People have done it recently. You do not know whether they would be very disappointed if it disappeared, and you do not want to wait for retention cohorts to mature. That is this page. Ellis built the survey as that leading indicator. A weak “users will miss this” line on a Yibud report sits near this fork — as a flashlight, not as a percent.
When a PMF-survey tripwire is the wrong next test
You cannot name real usage. You are still asking “would you use this?” You only needed reach, a problem interview, a solution show, a usage tester, a scarce-commitment partner, or a charge. Run those pages first. Surveying a waitlist and calling the result product-market fit is theater.
Worked example
One PMF-survey contract you can copy (illustrative)
The week below is illustrative — names, product, and counts are invented so you can see the setup. It is not a study, not a conversion benchmark, and not a Yibud report. Copy the method, not the numbers. Ellis’s 7-to-40 example stays in his article; it is not this story.
The must-have bet, written before the first send
Illustrative names: founder Priya Shah at Standupkit, a SaaS that turns a team’s standup replies into a daily digest. Real usage, written before the send: “published a digest that at least one teammate opened.” Signup and “connected Slack” do not count. The tempting next step is a paid acquisition week. The current bet assumes engineering managers who already published a digest would be very disappointed without it.
The tripwire, written the same sitting
Signal: percent Very disappointed among valid current users (exclude N/A). Threshold: 40 percent — Ellis’s experience bar, adopted before the send, not invented after. Sample: directional at about 30; more confident at 100+ (Ellis only). Window: first send through day 21. Recent-usage filter: used in the last couple of weeks. Action if miss: stop the paid-acquisition week and rewrite targeting / positioning / onboarding from the very-disappointed follow-ups. Compliments and “somewhat disappointed” do not count toward the numerator.
What a miss looks like on day 21
Illustrative close: 34 valid current users (N/A excluded). 8 say very disappointed — about 24 percent. Twelve compliments in sales calls. A waitlist of 200. That is a PMF-survey miss you already named — stop the ad spend and rewrite who you target, how you describe the digest, and how a new team reaches first real usage. It is not a reason to scale “because the waitlist is busy.” Ellis’s published case of 7 percent moving to 40 percent after targeting, positioning, and onboarding work is a separate, cited example — not this fictional week.
What must not happen in the window
Do not add signups who never published a digest just to reach 30. Do not count somewhat disappointed as very disappointed. Do not treat a Yibud score as this percent. Do not ask “how much would you pay?” and treat a yes as this page’s signal — that is the WTP page. Do not stretch day 21 because the percent is low — silence and a miss are a miss. If you cannot find ~30 people with recent real usage, you may have a reach problem — smoke, landing, or discovery, not a fake pass.
Write the tripwire before the first send. If day 21 arrives and the very-disappointed percent misses the bar you wrote, honor the action. If it cleared the bar, Continue to the next cheapest test — often money or a design-partner ask, not a factory. The method is the dated page, not the story you tell after.
Where Yibud fits
Use the free validator as a flashlight, then write the PMF-survey contract yourself
Yibud is a free, no-signup startup idea validator. Scores come from a deterministic rule engine, not from a language model guessing success. Optional AI text, when it is used, only polishes prose. It does not invent the number. A Startup MRI report can name a weak “users will miss this” or monetization assumption, then a Day-7 Continue / Refine / Re-test / Stop signal. PMF-survey work sits next to that assumption — write the tripwire, do not hope a mid-band score means “we already have 40 percent.” If you already have a report, start the contract with the named assumption — not with the overall score. Then run the window yourself. The method stands alone if you never open the analyzer.
Analyze my idea →Common mistakes
What usually wastes the PMF-survey contract
Surveying signups instead of recent real usage
Ellis’s Uber line is the one this page keeps: took a ride, not downloaded the app. If they never used it, or have not used it in several months, you already know the answer. Padding the sample with signups to hit 30 is how a waitlist becomes “fit.”
Treating 40 percent as a law — or rewriting it after the results
Ellis labeled around 40 percent as experience: when sustainable growth becomes possible. Adopt it or rewrite it before the send. A bar invented after 22 percent is a recap. This page does not call 40 percent a personal forecast.
Counting somewhat disappointed, compliments, or a Yibud score as the signal
The numerator is very disappointed. Exclude N/A. An interview “looks great,” a smoke-test click, and a mid-band engine score do not add to that percent. They have other pages.
Scaling before the tripwire fires — or skipping the follow-ups
Ellis: scaling growth before product/market fit is the fastest way to kill your startup. The open-ended answers from very disappointed users are how you learn who benefits and what the key benefit is. His 7-to-40 example ran through targeting, positioning, and onboarding — not through a bigger ad budget.
Inventing other N thresholds or a typical response rate
The only sample guidance on this page is Ellis’s: about 30 directionally useful, 100+ more confident. Do not publish a “usual” conversion rate. Do not import numbers from other products. Fill the rest of the page yourself.
Sources
Where these ideas come from
- Sean Ellis, “Using Product/Market Fit to Drive Sustainable Growth” (April 5, 2019) — Used for the exact question and options; surveying recent real usage (last couple of weeks; took a ride, not just signed up); score as percent very disappointed among valid current users; around 40 percent as Ellis’s experience threshold for when sustainable growth becomes possible; about 30 responses directionally useful and 100+ more confident; the survey as a leading indicator versus waiting for retention cohorts; scaling before product/market fit as the fastest way to kill a startup; and one business moving from 7 percent to 40 percent by improving targeting, positioning, and onboarding. This page does not invent other sample sizes, conversion rates, or case numbers.
- CB Insights, “The top 9 reasons startups fail” (report analyzing 431 VC-backed shutdowns since 2023; updated March 5, 2026) — Used as a warning about pattern frequency, not as a personal success forecast. CB Insights reports “ran out of capital” at 70 percent and calls it often the final cause of death. Among the 385 companies with identifiable reasons, more telling patterns include poor product-market fit (43 percent), bad timing (29 percent), and unsustainable unit economics (19 percent). Many companies cited more than one reason. This page does not invent other percentages.
In one paragraph
Summary you can quote
The Sean Ellis product-market-fit survey is a pre-written tripwire: an observable signal, a threshold, a calendar window, and a named action, dated before the first send. The signal is the percent of valid current users who would be very disappointed if they could no longer use the product — exclude N/A; survey only recent real usage (Ellis: last couple of weeks; took a ride, not just signed up). Ellis’s experience: around 40 percent very disappointed is when sustainable growth becomes possible — treat 40 percent as a default bar you may adopt or rewrite before the survey, not a law and not a personal forecast. About 30 responses are directionally useful; 100+ more confident. Open-ended follow-ups from very disappointed users are how you retarget, reposition, and redo onboarding; Ellis describes one business moving from 7 percent to 40 percent that way. CB Insights’ 2026 analysis (431 shutdowns since 2023; 385 with identifiable reasons; 70 / 43 / 29 / 19 percent) is a warning about pattern frequency, not a forecast. Yibud scores, WTP, earlyvangelist filters, interviews, smoke/fake-door/landing tests, and kill/pivot/decision contracts are different instruments. Scores do not prove PMF. Matching the tripwire you wrote first does.
FAQ
Questions founders actually ask
What is the Sean Ellis product-market-fit survey?
One question for people with recent real usage: How would you feel if you could no longer use [ProductName]? Options: Very disappointed / Somewhat disappointed / Not disappointed / N/A I no longer use [ProductName]. The score is the percent very disappointed among valid current users (exclude N/A). This page writes that survey as a tripwire — signal, threshold, window, action — dated before the first send.
Is 40 percent a law of nature?
No. Ellis’s 2019 article calls around 40 percent very disappointed the point, in his experience, when sustainable growth becomes possible. Treat it as a default tripwire threshold you may adopt or rewrite before you survey. It is not a law, not a Yibud rule, and not a personal forecast.
Who should I survey?
People who recently did real usage — Ellis’s example is took a ride, not just signed up. Ideally last couple of weeks. If they never used the product or have not used it in several months, you already know the answer is Not disappointed or N/A. Do not pad the sample with a waitlist.
How many responses do I need?
Ellis only: about 30 responses are directionally useful; at 100+ he was much more confident. This page does not invent other N thresholds or a typical response rate. If you cannot find ~30 people with recent real usage, fix reach or usage first.
Is a Yibud score the same as the Ellis percent? Do I need an account?
No. Yibud scores come from a deterministic rule engine on your idea inputs. The Ellis percent comes from recent real users answering one question. You do not need an account. The method on this page stands alone. Run the free analyzer if you want a flashlight on a weak “users will miss this” assumption before you write the contract. Optional language-model text only polishes prose.
Write the PMF tripwire before the next sprint
Yibud scores weak “users will miss this” assumptions with a deterministic engine. You write signal, threshold, window, and action while you are still honest — then a Sean Ellis survey is evidence, not a vibe.
Generate a Yibud report →