How to use PMF Surveys to reach Product/Market Fit Faster (Guide + Case Study)
โA product/market fit (PMF) survey asks users how theyโd feel if they could no longer use your product, and the share answering โvery disappointedโ is your PMF score (Sean Ellisโs 40% test). On its own that score is just a benchmark, and for years it told teams little more than โdo betterโ. The unlock, discovered by Superhuman in 2018, is segmentation: break the results down by user type, job title, or plan, and the average hides a subgroup that loves the product far more than the headline number suggests. Superhumanโs โvery disappointedโ score was 22% overall but 32% once they focused on the right segment, and segmenting the follow-up answers told them exactly what to build next. This guide walks through the Superhuman story and then how to run a segmented PMF survey yourself: pair a comparative-ranking question (pairwise comparison or points allocation) with a segmentation step, so the score becomes a roadmap instead of a grade.โ
โโThe only thing that matters is getting to product/market fitโโ
Product/market fit is the point when a startup begins to see exponential growth in demand: you've built something a lot of people want. According to Marc Andreessen, who first popularised the term in 2007, all startups are either Pre-PMF or Post-PMF. You either have it or you don't, and it should be obvious which side you're on.
The idea caught on quickly, but it stayed a vague target for years. "Getting to" something you can't measure is quite a challenge. Then, three years after Andreessen's famous blog post, Seรกn Ellis changed that with a technique for measuring product/market fit called the 40% rule.
Seรกn Ellis is a bestselling author and former growth leader at companies like Dropbox, Lookout, Eventbrite, and LogMeIn. In May 2010, he wrote a blog post explaining a simple indicator of PMF: ask your users "how would you feel if you could no longer use our product?" and measure the percentage who pick "very disappointed." If over 40% say they'd be "very disappointed", you've got PMF.
Ellis' 40% rule has since become the well-known 'Product/Market Fit Survey', but many don't realise that for over 8 years it was mostly ignored, thanks to one major gap in Ellis' thinking.
Ellis saw the result as a binary outcome: congratulations if you scored above 40%, otherwise keep digging until you work out how to make more of your "somewhat disappointed" participants see the product as a "must-have" instead of a "nice-to-have". But almost every early-stage team hustling toward PMF is already trying to make their product a must-have. They're constantly talking to users, iterating their messaging, releasing features, testing assumptions. Ellis' 40% rule didn't give these teams a better direction than any other research framework; it usually just told them to "do better".
That stayed the case until November 2018, when the founders of Superhuman added one twist to Ellis' 40% rule and finally sent the Product/Market Fit Survey viral: segmentation.
How Superhuman used segmentation to unlock its PMF score
The first time Superhuman's co-founders ran a PMF survey, in early 2017, only 22% of their users were in the "very disappointed" group. That looks like failing the 40% rule, but they understood 22% was just an average, and that some subgroup of users had to have a much higher disappointment score. So they manually tagged users by job title: Engineers, Marketers, Salespeople, and so on. Looking across those groups, the users with the highest overall disappointment score were founders, managers, executives, and people in business development.
When they grouped those participants together and ignored everyone else, their "very disappointed" score jumped from 22% to 32%.
That changed how they read the whole survey. They focused only on the relevant users with those job titles and split them into two groups: users who felt the product was a "must-have" (the "very disappointed" group) and users who saw it as only a "nice-to-have" (the "somewhat disappointed" group). Segmenting this way let them answer some big questions.
What did segmentation let Superhuman ask?
1. Why does the "Very Disappointed" cohort like Superhuman so much? They isolated the "very disappointed" participants and analysed their answers to "What is the main benefit you receive from Superhuman?". The most common responses centred on speed, focus, and keyboard shortcuts.
2. Do "Somewhat Disappointed" users care about these benefits too? They looked at the 45% of "somewhat disappointed" participants and their answers to the same "main benefit" question. Two-thirds mentioned speed, focus, or keyboard shortcuts as their primary benefit, while one-third prioritised other, unrelated things.
3. What is stopping the speed-focused "Somewhat Disappointed" users? By segmenting again to focus only on this subset, they found the highest-priority problems included the lack of a mobile app, missing integrations, attachment handling issues, and some other low-hanging fruit.
How did the survey turn into a roadmap?
The survey gave the Superhuman team two clear priorities:
Double down on the benefits the "very disappointed" users love most (faster speeds, more shortcuts, extra automation).
Solve the highest-priority problems for the "somewhat disappointed" users (mobile app, integrations, attachment and calendar features, better search, read receipts).
Superhuman repeated this at the end of each following quarter and tracked their segmented-PMF score as it climbed:
| Quarter | Segmented “very disappointed” score |
|---|---|
| Q4 2017 | 32% |
| Q1 2018 | 47% |
| Q2 2018 | 56% |
| Q3 2018 | 58% |
In under a year, focused segmentation took them from below the 40% benchmark to comfortably above it.
What can you actually learn from a segmented PMF survey?
The purpose of the PMF survey was misunderstood for almost 9 years. Rather than simply benchmarking your current product/market fit, the "disappointment" data works better as a filtering lens for understanding how your users interpret things like:
Product Benefits
Feature Value
Usability Issues
Customer Problems
Superhuman's story raises two questions for the rest of us: what's the best way to survey users about things like feature value or customer problems, and how do you segment the results the way the Superhuman team did?
Step 1: how do you measure customer preferences and priorities?
As in Superhuman's story, you want to collect both positive and negative information during the PMF survey. That helps you understand the "very disappointed" segment as well as the issues limiting the "somewhat disappointed" users. In a previous post I introduced the Product/Market Fit Matrix, which helps visualise the different lenses we use to frame research like this:
| Research lens | The question it answers |
|---|---|
| Outcomes & benefits | How is the customer's life better as a result of using your product? |
| Customer pain points | What needs, pains or desires is the customer trying to solve? |
| Feature value | Which features drive the most value and deliver the most benefit? |
| Usability issues | What limits the customer's ability to gain that value? |
Whichever of the four you pick, I tend to use a two-part question to gather the right data. The first part is a comparative-ranking exercise, which forces users to compare options and vote for the ones most important or urgent to them. After the ranking, I add a text-response question asking for options they care about that they didn't see during the voting, so I can plug any gaps in my list mid-survey.
My favourite comparative-ranking methods for surveys like this are pairwise comparison (turns your list into a series of head-to-head "pair votes") and points allocation (each participant spends 10 credits on the list according to their preferences). Here are two real examples from a recent survey where we measured "feature importance" according to OpinionX customers:
Step 2: how do you segment by disappointment group?
Comparative-ranking formats produce a score for each option from 0-100 based on how often participants choose it, which makes the data easy to rank and segment. To make segmentation as easy as possible, I use a multiple-choice question to collect the "very/somewhat/not disappointed" data. I also tend to include other multiple-choice questions like pricing plan, organisation type, and job title, all ideal ways to segment the results later.
You can build segments manually in Excel like the Superhuman team did, but a "Segmentation Matrix" gets it done 10x quicker. It's like a combination of a crosstab analysis and a heatmap: each segment is a column and each ranking option a row, with colour coding to help you spot the strong opinions fast.
Here's an example of the Segmentation Matrix in action:
^ You can play around with the Segmentation Matrix from this screenshot here (no login required)
In this example, participants were asked to allocate 10 credits to the features they'd most like built next. The "very disappointed" users clearly want the "Multi-Layouts" feature, whereas the "somewhat / not disappointed" segments spread their votes more inconsistently across many options, one example of how PMF surveys can offer actionable data for roadmap prioritisation.
There's no complex setup involved. On OpinionX, the "Segments Tab" populates itself automatically with this data, and you can configure the segments shown in a couple of clicks. Here's a no-login-required link to the Segmentation Matrix from the screenshot above.
A Segmentation Matrix helps you answer the three questions from Superhuman's case study:
Which segment has the highest percentage of "very disappointed" participants?
What are the top benefits according to those segmented "very disappointed" participants?
What are the barriers stopping suitable "somewhat disappointed" users from getting value?
You don't need to be a data scientist for this, as long as you pick the right tool. OpinionX is a free survey tool with a suite of segmentation features for filtering, comparing, and mapping segmentation data in ways any non-technical person can follow.
PMF surveys are not the only place segmentation helps
So many teams get caught up in benchmarking and lose sight of what surveys could offer if run properly. The same segmentation principle applies to NPS, CES, and CSAT surveys, all of which would be far more useful treated as segmentation datapoints for understanding customer differences rather than as bare benchmarks:
NPS (Net Promoter Score) โ "How likely are you to recommend us to a friend or colleague?" โ a points-rank exercise using expected product-benefit statements to identify why your most engaged customers love your product.
CES (Customer Experience Score) โ "On a scale of very easy to very difficult, how easy was our product to use?" โ an image-based pair-ranking exercise using feature-value statements to see which features are hardest to use.
CSAT (Customer Satisfaction Score) โ "How satisfied are you with our product?" โ a pair-rank exercise with product-usability problem statements to identify barriers to adoption for low-satisfaction customers.
Average results across a wide base of users aren't much use for product strategy. To improve adoption and activation, segmentation tells you what your best customers have discovered and what barriers are blocking your churning users. To improve conversion and monetisation, you need to know the value new customers expect the product to deliver. You can't work any of this out without a way to sort users into segments.
Rather than pulling your hair out doing all this in Excel, use OpinionX instead: over 12,000 teams use it to measure their customers' preferences and priorities, and it comes with purpose-built segmentation features. Create your own product/market fit segmentation surveys with a free OpinionX account today.
PS. My favourite piece of trivia from researching this post: in July 2010, a guy called Max Marmer pointed out to Seรกn Ellis that his 40% rule was missing segmentation.
โIโd consider adding something to the โif you are below 40%โ paragraph about how ideally you want the 40% to be from a sampling of all your users, but segmenting the survey may lead to insights about how different segments value your product, and which segments you might want to focus more of your customer development, and engineering energy on to get to Product/Market fit (ie. 40%).โ
Max understood that the "disappointment" data was better used to segment customers than to benchmark current PMF. He was seven years ahead of the Superhuman team reaching the same realisation in late 2017. It's funny how a good idea can take a while to catch on. :)
Until next time, Daniel
Frequently asked questions
What is a PMF survey? A survey asking active users "How would you feel if you could no longer use our product?" with three answers: very disappointed, somewhat disappointed, not disappointed. The share choosing "very disappointed" is your PMF score.
What counts as good product/market fit? 40% or more of users answering "very disappointed" is the benchmark from Sean Ellis's test. Below that, the product usually needs more iteration before scaling.
Why segment a PMF survey? Because the headline score is an average, and averages hide the subgroup that loves your product most. Superhuman's score was 22% overall but 32% once they focused on the right user segment. Segmenting tells you who your must-have users are and what to build to convert the "somewhat disappointed" group.
How did Superhuman use PMF segmentation? They tagged users by job title, found their most disappointed-to-lose users were founders, managers and executives, then analysed what that group valued (speed, focus, keyboard shortcuts) and what was holding the "somewhat disappointed" group back (no mobile app, missing integrations). That became their roadmap, and their segmented score climbed from 32% to 58% across four quarters.
How do you run a segmented PMF survey? Pair a comparative-ranking question (pairwise comparison or points allocation) about feature value or customer problems with multiple-choice questions for the disappointment group, plan, and job title. Then split the ranking results by those segments to see what each group prioritises.
The PMF score is a starting point, not the answer. What Superhuman proved is that the number you should act on is hiding inside the average, and the teams that reach product/market fit fastest are the ones who go looking for it instead of settling for the headline.
Want more like this? Over 42,000 researchers and product people get The Full-Stack Researcher in their inbox. Subscribe for the next one.
You can build a segmented PMF survey on OpinionX for free, ranking questions, segmentation matrix and all, at $0 for up to 25 participants per survey (full pricing). The 19 PMF survey templates guide has ready-made versions to copy.