Product-Market Fit Survey: One Question, Then 40 Percent

The whole product-market fit survey is one question: how would you feel if you could no longer use this product? Three answers are allowed, very disappointed, somewhat disappointed, not disappointed, and the number that matters is the share who pick the first one. Sean Ellis, the author of Hacking Growth, set the bar at 40 percent in a post on his Startup Marketing blog, now preserved in the Internet Archive copy of “The Startup Pyramid”. In his words, he defined the threshold “after comparing results across nearly 100 startups,” and he added, in the same paragraph, that it is “a bit arbitrary.”

That hedge is the reason the test is worth a small business owner’s time. It does not pretend to be a law. It is a rule of thumb from a man who watched nearly a hundred companies either struggle or take off, and it answers the one question that decides where your next dollar goes: do you have something people would miss, or something they tolerate? If the answer is the second one, spending on ads, a second truck or a bigger space buys you more people who tolerate you. Ellis put it bluntly in the same post: below the line, keep your burn low and put every resource into raising the share of customers who would be very disappointed without you. Marc Andreessen, whose 2007 essay Ellis was building on, defined the underlying idea as “being in a good market with a product that can satisfy that market” (pmarchive.com, June 25, 2007), and that definition fits a landscaping company or a meal-prep kitchen as well as it fits software.

Who you ask decides what the number means

The question only works on people who have actually used what you sell. Ellis asked “existing users of a product,” not prospects and not people who bought once and never came back. A survey sent to your whole email list will mix in people who have no basis for an answer, and their shrug will drag the percentage down for reasons that have nothing to do with your product.

So the list is customers who have bought from you more than once, or who have used your service long enough to know what it is like. For a lawn care company that means clients past their first two visits. For a bakery with a wholesale side it means the cafes who have reordered. For a bookkeeper it means clients past their first month-end. If that list is forty people, you survey forty people. Small numbers make the percentage jumpy, which is why the reading comes later in this piece with the arithmetic shown, but a jumpy number from the right forty people beats a smooth number from four hundred of the wrong ones.

The four questions, in the order that makes them useful

Rahul Vohra, the founder of the email app Superhuman, published the fullest public account of running this survey in First Round Review on November 12, 2018. His company started at 22 percent very disappointed, which he called clear evidence that it had not reached product-market fit, and reached 58 percent within three quarters by acting on the answers (First Round Review). He asked four questions, and the order matters because the later ones explain the first one:

1. How would you feel if you could no longer use this? Very disappointed, somewhat disappointed, not disappointed.
2. What type of people do you think would most benefit from it?
3. What is the main benefit you receive from it?
4. How can we improve it for you?

Swap “Superhuman” for the name of your business or the specific product you are testing, and keep the three answer options exactly as written. Do not add a neutral middle option and do not add a scale from one to ten. The test is calibrated on three choices, and the 40 percent figure means nothing if you change the choices. Keep the survey to those four questions. Every question you add costs you finished responses, and from a list of forty, every finished response is worth 2.5 points of your score.

Send it from your own name, by email or text, to the list described above, with one reminder a week later. Ellis and Vohra both surveyed people inside a product; a small business reaches the same people through the channel it already uses to confirm appointments or send invoices.

Reading the answers: three piles, and the pile most owners skip

Count the very disappointed answers and divide by total responses. That is your score, and if it is above 40 percent, Ellis’s benchmark says you are in the group that gains traction. The rest of Vohra’s method is what turns the score into decisions, and it is where the test earns its keep.

Vohra’s first move was to read the second and third questions only from the very disappointed group. Those people told him who the product was really for and which benefit made it indispensable, and when he narrowed his description of the market to match them and recomputed the score for that group, it read 33 percent (First Round Review). For a small business this is the step that turns a vague “we serve homeowners” into “we serve two-income households with a yard and no Saturday,” which is a sentence you can build an ad around. The six lists behind the Value Proposition Canvas are the natural place to write down what this group tells you.

His second move is the one worth stealing. He split the somewhat disappointed group by their answer to question three. People who named the same main benefit as the very disappointed group were, in his words, the group he paid “very close attention” to, because the product’s core already worked for them and question four told him what was holding them back. People who named some other benefit he chose to “politely disregard,” because the product was never going to be built around what they wanted. That split is the non-obvious part of the whole method: the fastest route to a higher score is not winning over the people who do not care, it is removing one obstacle for the people who almost love you.

Then there is the not disappointed pile, and a small business reads it differently from a software startup. A customer who keeps buying and would not miss you is telling you that you are interchangeable, and interchangeable usually means price. That is not a reason to fire them. It is a reason to stop spending growth money on finding more of them, and to notice that a competitor’s coupon will take them whenever it arrives. Our view is that this pile is the most useful warning the survey gives an owner, because a healthy repeat-purchase report can hide it completely.

A worked example with small numbers

Take a hypothetical meal-prep kitchen with 90 customers who have ordered at least three times. It is a made-up business used to show the arithmetic, not a case study. The owner sends the four questions and gets 54 responses, a 60 percent response rate, which is high and reflects a personal email from someone customers know. The first question comes back 19 very disappointed, 24 somewhat disappointed, 11 not disappointed.

The score is 19 divided by 54, which is 35 percent. Below the line, but with 54 responses a swing of three people moves the score by more than five points, so the honest reading is “near the line, not over it,” not “failed.” Reading question three only from the 19, fourteen of them name the same thing: they stopped cooking on weeknights. Of the 24 somewhat disappointed, 15 also name weeknights. Those 15 are the whole plan for the next month. Their question four answers, in this hypothetical, cluster on two things: delivery arrives after dinner time on Mondays, and the menu repeats every two weeks. Both are fixable without a new kitchen. The nine somewhat disappointed people who named something else, cheaper than restaurants or good for a diet, are not the plan, and the 11 not disappointed are the reminder that a fifth of the respondents would leave for a discount.

Run the same survey again in 90 days to the same kind of list and the owner finds out whether fixing Monday deliveries moved the 15. If it did, the score crosses 40 percent and the next decision, a second delivery van or more ad spend, is being made for people who would miss the business, which is the only kind of growth spending that compounds. The retention campaign built from who stopped buying is the complement to this: that piece works the list of people who left, this one works the list who stayed.

What the survey costs to run

The tooling is the cheapest part, and with a list under a few hundred names the free tiers cover it. Prices below were read on each vendor’s pricing page on November 7, 2026.

Google Forms comes with a Google account at no extra charge, and a Google Workspace Business Starter seat, which adds a business email address, is $7 per user per month on a one-year commitment (workspace.google.com/pricing). Responses land in a Google Sheet, which is all the analysis this survey needs. SurveyMonkey has a free Basic plan limited to ten questions per survey, which is six more than you need; its Advantage Annual plan is $39 per month, billed as $468 a year, and its month-to-month Standard plan is $149 per month (surveymonkey.com/pricing). Typeform has a free plan, and its Basic plan with 100 responses a month is $39 month to month or $28 per month on yearly billing (typeform.com/pricing). For a one-off survey of forty to two hundred people, Google Forms or either free plan does the job; pay for SurveyMonkey or Typeform only if you will run this every quarter and want the branding and logic features.

If you are testing a new product line before it exists, this is the wrong tool, because the question requires use. The 30-day product launch checklist covers the go or no-go decision before launch; this survey is for the ninety days after it, once real customers have real experience to report.

Where AI fits, and where it does not

An AI assistant is good at two jobs here and bad at a third. It can draft the survey email and the reminder in your voice from a two-line brief. And once the responses are in, it can sort fifty open-text answers to questions three and four into piles in a minute, which is the part owners put off because reading fifty answers takes an evening. Paste the responses into ChatGPT, Claude or Gemini with the instruction to group the main-benefit answers into no more than five themes and count each, and you have the Vohra split without the evening.

The job it should not do is decide what the piles mean. The very disappointed verbatims are the most valuable sentences your business will collect this year, and the owner should read every one of them. They are the raw material for the next ad, the next hire’s job description and the sentence on the homepage. An assistant that summarizes them for you is throwing away the part that was worth collecting.

BusinessPrompter, a sister venture of this site, publishes a Product-Market Fit Validator prompt in its Innovation and Growth category, available as a single prompt and a five-step chain to Pro members. Its summary line is “Validate product-market fit before scaling,” which is the exercise above; one caution is that the page’s About panel describes an operational-excellence program for operations leaders in manufacturing and logistics, so read the prompt itself rather than the panel. The BusinessPrompter.com catalog has free prompts alongside the Pro ones.

Ask the question of yourself first

Before sending the survey, answer it about your own suppliers. Your accountant, your payment processor, the distributor who brings the flour. For how many of them would you honestly pick very disappointed? Our guess is that the honest answer is one or two, and a shrug for the rest. That shrug is exactly what a share of your customers feel about you, and the survey only tells you how large that share is. The rest of the method, finding the people one fix away from very disappointed and making the fix, is work you can do with a free form, an evening of reading and ninety days.