How UX Pilot found out it reached product-market fit

Most founders answer "do you have product-market fit?" with a shrug and a vibe. UX Pilot, a bootstrapped AI design tool with over a million users, decided to answer it with a number instead.

I'm Head of Product at UX Pilot, and in the autumn of 2025 we ran the Sean Ellis product-market fit survey on our paid users. This is exactly what happened: the method, the score, and what the answers told us about who our product is really for.

The method: one question, the right people

The Sean Ellis test is deceptively simple. You ask engaged users one question, "How would you feel if you could no longer use the product?", with three options: very disappointed, somewhat disappointed, not disappointed. The share who answer "very disappointed" is your PMF score, and 40% is the benchmark that signals real fit.

Here's the exact setup we used:

1. Audience: paid users only, the people who have actually experienced the core value. Reach of roughly 20,000 across the US, UK, Germany, France, UAE, Netherlands, Canada, Israel, Spain, and Australia.
2. Instrument: an in-product survey via PostHog, shown to active paid users, plus the three open-ended follow-ups that turn a score into a roadmap.
3. Sample: 303 valid responses, about a 1.5% response rate, collected over a few weeks.
4. Scoring: we exported the responses to CSV and imported them into PMFtracker, which calculated the score, segmented respondents, and generated the customer profile automatically.

That last step matters. Running the survey is the easy part; the value is in the segmentation and the qualitative analysis, and doing that by hand across 303 open-ended answers is exactly the tedious work that gets skipped. Importing the raw CSV meant the score and the segments came out in minutes.

The result: 49%, above the line

UX Pilot scored 49%. That clears the 40% Sean Ellis benchmark and lands the product firmly in the strong-fit range. But the single number was the least interesting part. The segmentation is where the survey earned its keep:

The 49.2% "very disappointed" group is the number that matters. Those 149 users are the people who would be genuinely upset to lose UX Pilot, and their answers to the open-ended questions define who the product is for.

Who loves it: the ideal customer profile

We read the "very disappointed" users' answers to "what type of person would most benefit from this product?" The pattern was unmistakable:

This is a real ICP, not a demographic guess. It's a specific person in a specific situation. And when we asked those same users what benefit they get, three themes dominated their own words:

• Speed and time savings (the single most mentioned benefit) - "making screens a breeze," "fast translations of initial concepts into shareable designs."
• Idea generation - users lean on it as a creative partner for exploring directions, not just a production tool.
• Rapid prototyping - quick mockups and wireframes that make stakeholder conversations concrete.

That is the core value proposition, written by the market rather than the marketing team. It's the exact language that should lead the homepage.

Who's on the fence: the barriers

The 85 "somewhat disappointed" users are the growth opportunity, they see value but something holds them back. Their answers to "how can we improve this for you?" pointed at a short, actionable list: tighter design-tool integration, more control over the AI's output, and more editing flexibility. That's a roadmap handed over by the exact people most likely to move into the "very disappointed" group if you serve them well.

This is the whole reason to segment instead of just reading an average. The 49% tells you that you have fit. The "somewhat disappointed" answers tell you how to raise it.

The lesson: PMF is measured, not felt

The point of this case isn't the 49%. It's that UX Pilot stopped guessing. A one-question survey, sent to the right people, scored and segmented, turned "we think users like us" into "49% would be very disappointed to lose us, here's exactly who they are and what's holding the rest back."

You don't need a million users to do this. You need engaged users, the one question, and at least 40 or so valid responses. Everything after that, the score, the segments, the profile, is mechanical, which is precisely what PMFtracker automates.