The Sean Ellis Test: Measure Product-Market Fit With One Question
Ask your active users one question — how would they feel if they could no longer use your product? — and score the answers on a four-point scale. The Sean Ellis test is that survey: a single, standardized question sent to people who've actually used your product, with the share who answer "very disappointed" serving as a numeric read on whether you've reached product-market fit. Cross roughly 40% and the evidence says you have it. Land under that line and the same survey tells you what to fix.
Quick Answer: The Sean Ellis test asks active users: "How would you feel if you could no longer use [product]?" with four options — Very disappointed, Somewhat disappointed, Not disappointed, and N/A (I no longer use it). If 40% or more say "very disappointed," that's the widely cited benchmark for product-market fit. Below 40%, don't scale yet — use the follow-up questions and segment analysis to find out what's missing first.
If you're still working out what product-market fit actually means for your situation before you go measure it, What Is Product-Market Fit? How to Know If You've Found It is worth reading first — this survey only makes sense once you know what result you're looking for.
How the Sean Ellis Test Works
The method is deliberately narrow — one question, one scale, sent to a specific group of people — because the value of the test comes from being able to compare your number against a known benchmark. Change the wording, the audience, or the scale, and you've built a different survey that happens to look similar.
The exact question. Ask: "How would you feel if you could no longer use [product]?" Keep the wording close to the original — it's phrased as a loss, not a satisfaction rating, which is what makes it different from an ordinary "how happy are you with X" survey. A satisfaction question measures how users feel about your product in isolation. A loss-aversion question measures how much your product has actually displaced from their workflow or their life, which is a much closer proxy for whether you've become load-bearing rather than merely liked.
The four answer options. Respondents choose exactly one:
- Very disappointed — the response you're counting
- Somewhat disappointed
- Not disappointed
- N/A — I no longer use the product
Resist the temptation to add a five-point scale or a neutral middle option. The forced choice between "very" and "somewhat" is what makes the 40% threshold meaningful — it's calibrated against surveys that used this exact wording and this exact scale, not a generic satisfaction score.
Who to send it to. This is where most first attempts at the test go wrong. Send it to your entire user base — including people who signed up last week and never came back — and you'll dilute the result with people who never got far enough to form an opinion. The test only works on people who've actually experienced your product's core value. A common filter, drawn straight from Sean Ellis's own guidance, is to survey users who have used the core feature at least twice within the past two weeks. Adjust the exact frequency and window to whatever "real usage" looks like for your product, but keep the principle: you're measuring reaction from people who have something to lose, not curiosity from people who haven't tried it yet.
Sample size. A handful of responses can swing a percentage wildly — ten responses with four "very disappointed" answers is 40%, and so is forty responses with sixteen. Only one of those numbers is stable enough to act on. Most practitioners who use this survey suggest waiting until you have at least a few dozen valid responses, commonly cited in the range of 30 to 40, before treating the percentage as a real signal rather than noise.
Timing. Run it after you have a real base of users who've had time to form a habit around the product — not on day one, and not only once. Treat it as a recurring pulse check: an early read tells you whether you're in the right neighborhood, and re-running it as you ship changes tells you whether you're moving toward 40% or away from it.
Running the Survey: A Concrete Example
Say you run a project-management tool for small agencies. You filter your user list to accounts that have created or updated a task at least twice in the last two weeks, and you send the four-option question to that filtered group by email and in-app prompt.
Out of everyone you contacted, a portion responds — surveys like this typically pull response rates well under half, so plan your outreach list large enough that your responses, not your invitations, clear the 30-40 minimum. Once responses are in, the scoring is arithmetic: divide the count of "very disappointed" answers by total valid responses (excluding the N/A group, who told you they've already churned and aren't part of the population you're measuring). That percentage is your score.
The more useful work happens after you have the topline number, in how you segment it. Break responses down by:
- Plan tier or company size — a 55% "very disappointed" rate among agencies with six or more staff and a 15% rate among freelancers isn't a wash to an overall 30%; it's a clear signal about which segment you actually have fit with.
- Acquisition channel — users who arrived through a specific channel (referral, a particular content piece, a specific integration) often score differently than the average, telling you where to spend more.
- Feature usage pattern — cross-reference "very disappointed" respondents against what they actually use in the product. If they cluster around one feature, that feature is probably your real value proposition, whatever your marketing currently says it is.
Every "very disappointed" respondent should also get short qualitative follow-ups: what's the main benefit they get, who else would they recommend the product to, and what would they use instead if it disappeared. Those free-text answers are what turn a percentage into a roadmap — they tell you which words to put on your homepage and which segment to target next, not just whether you passed or failed.
Sean Ellis Test vs. NPS vs. Retention Curves
The Sean Ellis test is one of three metrics founders reach for when they want a number that stands in for product-market fit, and they measure genuinely different things — treating them as interchangeable is where a lot of false confidence comes from.
| Sean Ellis Test | Net Promoter Score (NPS) | Retention Curve | |
|---|---|---|---|
| What it measures | How much users would miss the product if it vanished | How likely users are to recommend the product to others | Whether users who sign up keep coming back over time |
| Question / input | "How would you feel if you could no longer use [product]?" | "How likely are you to recommend [product] to a friend or colleague?" | No survey — behavioral data pulled from product usage logs |
| Output | % answering "very disappointed" | Score from -100 to +100 | A curve plotted over weeks or months, ideally flattening above zero |
| Benchmark commonly cited | ≥40% | Varies by industry; no single universal cutoff | Curve should flatten rather than decay toward zero |
| Main limitation | Self-reported and a single point in time; needs a real sample size to be stable | Measures advocacy, not retention — a product people love can still churn them for unrelated reasons (price, a competitor's feature) | Requires enough usage history to show a real pattern; says nothing about why |
The practical takeaway: use the Sean Ellis test as an early, fast-turnaround signal you can run within weeks of having real users, use NPS as a secondary check on advocacy once you have distribution to worry about, and treat the retention curve as the metric that eventually overrides both, because it's built from what people actually do rather than what they say in a survey. For a deeper look at why retention specifically deserves that final say, see Retention Metrics & PMF Signals: What Numbers Matter.
When the 40% Threshold Actually Matters for Validation
The 40% figure comes from Sean Ellis's own research comparing survey results across early-stage companies — the ones that went on to grow reliably tended to clear that line, and the ones that struggled tended to fall short of it. That history, along with the full survey methodology, is documented in Hacking Growth, the book Ellis co-wrote with Morgan Brown. Treat the number the way it was intended: a decision threshold, not a scoreboard.
Above 40%. This is evidence you can start investing more confidently in growth and distribution, because you have data suggesting the product itself won't be the reason new users churn. It is not permission to stop listening — segment your "very disappointed" respondents and double down on whichever slice of your market they represent, rather than marketing to everyone equally. A product that has fit with one segment and mediocre fit with another should look and sound different depending on who it's talking to.
Below 40%. This is a diagnostic, not a verdict. Go back to the segmentation and qualitative answers described above and look for a subgroup that's already well above the line — a smaller, sharper audience you have real fit with is a far better foundation to build on than a broad audience you have partial fit with. If no segment clears the bar, the qualitative "somewhat disappointed" and "not disappointed" answers usually point at what's missing: a specific feature gap, a workflow the product doesn't fit, or a promise the onboarding makes that the product doesn't keep.
The place this number causes real damage is when it gets treated as more certain than it is. A single survey run once, on a small and unfiltered list, that happens to clear 40% is not the same evidence as a stable 40%+ across repeated surveys and a real, filtered sample — and founders under pressure to show traction are the ones most likely to round the first case up to the second. Why Most Founders Overestimate Product-Market Fit covers the broader pattern of mistaking early enthusiasm for validated fit, and the Sean Ellis test is not immune to it just because it produces a clean-looking percentage.
Key Takeaways
- The question is specific for a reason. "How would you feel if you could no longer use this?" measures loss aversion, not satisfaction — don't reword it into a generic happiness rating.
- Audience filtering matters more than most founders assume. Survey people who've actually used the core feature recently, not your entire signup list, or you'll dilute the result with people who never formed an opinion.
- Don't trust a small sample. A handful of responses can swing the percentage by 10+ points; wait for a real sample — commonly cited as at least 30 to 40 valid responses — before acting on the number.
- The percentage is the start of the analysis, not the end. Segmenting "very disappointed" respondents by plan, channel, or feature usage is where the roadmap-relevant insight actually lives.
- 40% is a threshold, not a competition score. Use it to decide whether to invest in growth or keep iterating — not as a number to optimize for its own sake.
- Pair it with retention data, not just NPS. The survey is fast and self-reported; retention curves are slower but behavioral, and should carry more weight when the two disagree.
- A single passing survey isn't proof. Repeat the test as you ship changes, and treat one good result with real skepticism until it holds up again.
Frequently Asked Questions
What counts as a good Sean Ellis test score?
40% or more answering "very disappointed" is the widely cited benchmark for product-market fit, drawn from Sean Ellis's comparison of survey results across early-stage companies. Treat it as a threshold for confidence, not a target to game — a 40% score from a properly filtered, adequately sized sample is meaningful; the same number from an unfiltered list or a handful of responses is not.
How many responses do I need before I can trust the result?
Enough that a few responses swinging one way or the other wouldn't meaningfully move the percentage — commonly cited as a floor of around 30 to 40 valid responses. Below that, treat any number you get as directional at best, and keep collecting before making a decision based on it.
Is the Sean Ellis test still relevant, or has it been replaced by newer methods?
It's still widely used because it's fast to run and cheap to repeat, which makes it a good early-stage signal before you have enough usage history for a reliable retention curve. It was never meant to stand alone — pair it with behavioral data like retention and, once you have distribution, advocacy metrics like NPS, rather than treating any single survey as the final word.
What should I do if my score comes in below 40%?
Segment your responses before concluding you lack fit entirely — a smaller subgroup scoring well above 40% is often hiding inside a disappointing overall number, and it's a better foundation to build on than a broad, lukewarm audience. Where no segment clears the bar, use the qualitative follow-up answers from "somewhat" and "not disappointed" respondents to find the specific gap, then re-run the survey after you address it rather than assuming one low score is permanent.