How to Run a PMF Survey (Step by Step, With the Exact Questions)

It's simple. Before you spend another month building or another dollar on ads, you want to know one thing: would anyone actually miss this if it vanished? A PMF survey answers that in about a week, for free.

But "simple" isn't "easy to get right." The most common way founders waste a PMF survey isn't laziness — it's surveying the wrong people, or wording the question so it flatters them. Let's walk through it so your number is honest enough to bet on.

Step 1: Decide who to survey

This is the step that makes or breaks the whole exercise, so we'll spend real time on it.

You want to survey engaged users — and the working definition is precise: people who have used the core of your product at least twice in the last two weeks. Not everyone who signed up. Not the newsletter list. The people who actually experienced the thing your product is for, recently, more than once.

Why the filter matters this much

Picture two bad surveys. In the first, you blast every email you've ever collected. Half of them barely remember signing up; they answer "not disappointed" out of indifference and crater your score for no real reason. In the second, you survey only your hand-picked beta superfans, who answer "very disappointed" out of loyalty and hand you a flattering 60% that evaporates the moment normal users arrive.

Both numbers are fiction. The engaged-user filter is what keeps you out of both ditches: recent enough that the experience is fresh, repeated enough that they've truly used it, broad enough that they're not just your friends.

Step 2: Write the survey (the exact questions)

Keep it short. Every extra question costs you completions. Here's the full set — one scored question, three open-ended follow-ups. (It's the same Sean Ellis survey template PMFtracker pre-loads, word for word.)

The core question (this is the score)

That fourth option is doing quiet, important work. It pulls churned users out of your denominator so your score reflects the people actually using the product. Don't skip it.

The open-ended follow-ups (this is the gold)

The percentage gets the headlines, but these three questions are where your ICP, your positioning, and your roadmap come from:

• "What type of people do you think would most benefit from [product]?" — Your users describe your ICP for you, often more clearly than you'd describe it yourself.
• "What is the main benefit you receive from [product]?" — The recurring phrase in these answers is your core value and your headline copy.
• "How can we improve [product] for you?" — Ask this especially of the "somewhat disappointed." It's your conversion roadmap.

Four questions total. Two minutes to complete. Don't bloat it — the open-ends only get answered if the survey feels light.

Step 3: Choose how to deliver it

Two main channels, and the right one depends on who you're trying to reach:

If you can, trigger the in-app version after a user completes a core action — they're warm, they've just experienced the value, and their answer is grounded in a real session rather than a vague memory. For broader coverage, follow up by email to engaged users who haven't answered in-app.

Step 4: Collect enough responses

You need enough that one strong opinion can't swing the result — but far less than you'd fear.

The most important thing here is to start now. If you have 40 engaged users, you have enough for a directional read this week. Founders who wait for a "statistically perfect" sample usually wait forever. A rough 33% you measured today beats a flawless score you never run.

Step 5: Calculate your score

Once responses are in, the math is one line:

Remember to drop the "N/A — no longer use it" answers from the denominator. If 31 of 78 valid respondents say "very disappointed," you're at 40%. Here's exactly how to interpret that number and the 40% threshold.

Step 6: Do something with the results (most people stop too early)

A score you file away is wasted. The moment the responses land, run this:

1. Segment by answer. Split users into very / somewhat / not disappointed. You'll treat each group completely differently.
2. Profile the "very disappointed." Who are they? Role, company stage, what they switched from. This group is your ICP.
3. Read the "somewhat disappointed" improvements. Their blockers are your conversion roadmap — the features and fixes that move people into "very."
4. Ignore the "not disappointed." They're usually the wrong audience. Don't let their feedback steer the product.

This is the heart of the Superhuman engine — survey, segment, double down on fans, convert fence-sitters. See the full breakdown of how they ran it.

Step 7: Run it again (this is the real point)

A one-time survey gives you a snapshot. The asset you actually want is a trend line. Schedule the survey to run on a cadence — monthly or quarterly — so you can watch your score move as you improve, catch fit slipping before it's a crisis, and walk into a fundraise with "we went from 26% to 41% over two quarters" instead of a single lonely number.

Re-measuring is also the step that dies first when you run this by hand. Stitching together a Typeform, a CSV export, and a spreadsheet is tolerable once and miserable every month. Which is exactly why most founders measure once and never again — and why putting the loop somewhere automated is what makes it actually stick.

Common mistakes to avoid

• Surveying everyone. Inactive signups crater your score and hide the truth. Engaged users only.
• Leading the question. "You'd be sad to lose us, right?" isn't a survey, it's fishing for a compliment. Use the neutral wording.
• Ignoring the open-ends. The percentage is the least valuable thing the survey gives you. The text is where the strategy is.
• Running it once. Fit moves. A snapshot can't tell you which direction.
• Chasing "not disappointed" feedback. Building for people who'll never love you is how good products lose their edge.

Get those right and you've got a repeatable instrument for the most important question your startup faces — answered with data instead of hope.