The PMF Survey: How to Run the Sean Ellis Test (And Why It's Not Enough)

The Survey: One Question, Three Answers

The survey is deceptively simple:

The rule: If 40%+ of respondents select "Very disappointed," you have a positive PMF indicator. This threshold was derived from Sean Ellis's analysis of hundreds of startups — companies that went on to scale successfully had 40%+ of active users saying "very disappointed," while companies that struggled had less.

That's it. One question. Three options. One threshold. The simplicity is its strength — you can deploy it in minutes with any survey tool and get a result the same day.

For the complete 4-signal PMF framework that puts this test in context, see our full PMF guide.

How to Run the Survey

Step 1: Define Your Population

Who to survey: Active users — people who have used your product in the last 30 days. Specifically, users who have completed at least one core workflow and have had meaningful engagement with the product.

Who NOT to survey:

• Churned users (they've already decided your product isn't for them)
• Trial users who never activated (they never experienced the value)
• Users who signed up more than 90 days ago and haven't used the product in 60 (they've silently churned)
• Internal team members and beta testers (their context is different from real users)

Sample size: Aim for 40 to 100 responses minimum. Below 40, a few outliers can swing your PMF score by 10+ percentage points. If your user base is small, survey everyone who qualifies. For larger bases, a random sample of active users works as long as you hit the minimum.

Step 2: Choose Your Survey Tool

Any survey tool works: Typeform, SurveyMonkey, PostHog Surveys, Intercom, or a simple email with reply options. The tool doesn't matter. The population and timing do. What matters is that the survey reaches active users in-context, not buried in a quarterly NPS survey.

Step 3: Set the Timing

Send the survey at a moment when the user has recently experienced value:

• After they complete a core workflow
• After they generate a report
• After they've used the product for 14+ days

Don't send it: Immediately after signup (they haven't experienced the value). After an error or outage (they're frustrated). During onboarding (they're still learning).

Step 4: Segment the Results

This is the step most teams skip. Break down the results by:

• Plan type: Do paid users say "very disappointed" at a higher rate than free users? (They should.)
• Company size: Do enterprise users say it at a higher rate than SMB users? (This tells you which segment has PMF.)
• Signup source: Do organic signups say it at a higher rate than paid ad signups? (If yes, your ICP is organic, not paid.)
• Tenure: Do 6-month users say it at a higher rate than 1-month users? (If yes, the value compounds over time.)
• Role: Your product might have strong fit with product managers but weak fit with engineers. Segment your PMF score by role to find your true audience.

Segmentation is critical. Do not just calculate one overall PMF score. You may discover that your overall PMF score is 30%, but it's 55% among product managers at mid-market SaaS companies. That segment is your ICP.

Step 5: Compare Over Time

Run the survey quarterly, not once. The trend matters more than the single number:

• 25% → 35% → 42%: You're approaching PMF. Keep building.
• 45% → 40% → 35%: You're losing PMF. Something changed.
• 38% → 42% → 41%: You have PMF. It's stable.

A company at 35% that was at 25% 6 months ago is healthier than a company at 42% that was at 50% 6 months ago. The trend tells you whether you're building something customers increasingly need.

What the 40% Threshold Actually Means

The 40% rule comes from Sean Ellis's analysis of hundreds of startups. He found that companies that went on to scale successfully had 40%+ of active users saying "very disappointed." Companies that struggled had less than 40%.

It's a correlation, not a causation. 40% doesn't cause growth. It correlates with the underlying pattern of users who find your product essential.

The threshold is a heuristic. 39% doesn't mean "no PMF" and 41% doesn't mean "PMF." The 40% threshold works across many products as a rough guide. Your specific segment may have a different threshold.

The Sean Ellis question measures dependency, not satisfaction. That distinction matters. Satisfied users may still leave if a better option appears. Dependent users cannot, because your product has become essential to how they work. This is why the question asks about loss ("if you could no longer use") rather than preference ("do you like"). Loss aversion is a stronger emotional signal than preference, and it produces more honest answers.

Why the Sean Ellis Test Is Not Enough

The Sean Ellis test is Signal 2 of 4 in the complete PMF framework. Here's why it can't stand alone:

It Measures Declared Preference, Not Revealed Behavior

Users say they'd be very disappointed. Then their renewal comes up and they churn. The gap between what users say and what they do is the PMF gap.

The fix: Combine the survey with retention data. If 40% say "very disappointed" AND your retention curve is flat, you have PMF. If 40% say "very disappointed" but your retention curve decays to zero, users are lying to you (or to themselves).

It Averages Across Segments

Your overall rate is 38%. Sounds like you're close to PMF. But Segment A is at 72% and Segment B is at 12%. You have strong PMF for Segment A and no PMF for Segment B. The average tells you nothing.

The fix: Segment everything. Run the survey separately for each meaningful segment. Act on the segment-level data, not the aggregate. For how to distinguish real PMF signs from false signals, see our signs guide.

It's Static

One survey tells you the current state. It doesn't tell you whether you're improving or declining. A single data point is a snapshot; a trend is a story.

The fix: Run the survey quarterly. Track the trend. A product moving from 25% to 35% to 42% is healthier than a product stuck at 40% for four quarters.

The 4-Signal Framework

The Sean Ellis test is one signal of four. Here's the complete framework:

When all 4 signals align, you have PMF. When 2 or fewer align, you don't — regardless of what the Sean Ellis test says. For the complete PMF framework with all 4 signals, see our full guide.

Beyond the Core Question: Follow-Up Questions That Reveal More

The Sean Ellis question tells you whether you have PMF. The questions below tell you why, for whom, and what to do next. Each question reveals a different dimension of your product-market fit.

Value Proposition Questions

"What is the primary benefit you get from [product]?" This is the most important open-ended follow-up. The words your "very disappointed" users choose to describe your benefit should become your marketing copy. If they consistently say "saves me 3 hours a week on reporting," that is your value proposition.

"What would you use instead if [product] did not exist?" The most direct competitive intelligence question. Pay close attention to "Nothing" or "I would go back to doing it manually." These responses indicate you have created a new category.

Target User Questions

"What is your role?" PMF often varies dramatically by role. Your product might have strong fit with product managers but weak fit with engineers. Segment your PMF score by role to find your true audience.

"How long have you been using [product]?" PMF typically strengthens with usage duration. If it doesn't, your product may have an engagement ceiling. Track how the PMF score changes across tenure cohorts.

Improvement Questions

"What is the main thing you would improve about [product]?" The "main thing" constraint forces prioritization. General "what would you improve?" questions produce scattered wish lists. This question produces a ranked priority list when you aggregate responses.

"What does [product] do better than the alternative?" Your competitive advantages as perceived by users, not as assumed by your team. These are the reasons people switched to you and the reasons they stay. Protect these advantages aggressively.

FAQ