Most founders say they are working toward product-market fit. Very few can tell you their current score.
PMF is not a feeling. It is a measurement problem. And the teams that measure it correctly are the ones who know when to accelerate and when to stop pivoting.
In 2010, Sean Ellis surveyed users of dozens of startups and found a clear dividing line. Companies that answered "very disappointed" to the question "How would you feel if you could no longer use this product?" at 40% or above were on a growth path. Below 40%, they were not.
The survey question is simple: "How would you feel if you could no longer use [product]?" with four choices: very disappointed, somewhat disappointed, not disappointed, and N/A (no longer use it).
You track one number: the percentage who answer "very disappointed." The target is 40%.
Benchmark: Below 25% means you have not found fit. 25-40% is a promising signal worth investigating. Above 40% is the green light to scale.
The Ellis survey is the anchor, but it has blind spots. It only captures existing users. It cannot tell you whether the segment who loves you is large enough to build a business on. It does not reflect retention, monetization, or organic growth.
Seven quantitative signals fill those gaps. Together with the Ellis score, they give you a full PMF picture.
Plot your weekly active users by cohort. If retention curves flatten above zero instead of declining to zero, users have found a durable reason to stay. A flat line at any level above zero is a sign of fit in at least a segment.
Track what percentage of new signups come from word-of-mouth or direct referrals with no paid incentive. A rate above 20% without a referral program running means users are evangelizing on their own. That is one of the strongest PMF indicators available.
Measure how long it takes a new user to reach the "aha moment" your power users experience. If new users consistently reach that moment within one session, you have a product that delivers value without friction. If it takes days or requires onboarding calls, retention will suffer regardless of how good the core product is.
A whole-user-base NPS can mask fit in a subset. Run NPS separately for your intended ICP. An NPS above 50 within a defined segment, even if overall NPS is 20, tells you exactly who the product is built for and where to focus growth.
Define how often your product is supposed to be used. Then measure actual usage frequency. If users interact 3 times per week with a product you designed for daily use, there is a habit gap. If users interact more than the intended frequency, the product has found a deeper need than you expected.
In a product with tiers or usage-based pricing, watch how many users hit the ceiling and upgrade without prompting. An organic upgrade rate above 5% per month within a cohort is a strong signal that users are getting compounding value over time.
Exit survey every churned user with one forced-choice question: did they leave because of price, lack of features, changed circumstances, or found a better alternative? If "found a better alternative" is consistently below 10%, you have a defensible product. If it is the top reason, you have a competitive PMF problem, not a retention problem.
Start the Ellis survey when you have at least 50 active users who have used the product more than once. Run it by email or in-app with a single-question survey. Segment results by acquisition channel and user role immediately.
Build a simple PMF scorecard that tracks all 8 signals on one sheet. Review it monthly with your team. Assign a score of 0, 1, or 2 to each signal based on whether it is a red flag, neutral, or green. A total score above 12 out of 16 puts you in the zone where scaling investment is justified.
If your Ellis score is below 25% but 3 or more quantitative signals are strong, look for a subset of users driving those signals. That subset is your real ICP. Rebuild positioning and onboarding around them before measuring again.
Rule of thumb: PMF is not binary. It is a spectrum. Measure it quarterly, segment it ruthlessly, and treat any score below 40% as an invitation to iterate, not evidence of failure.
Teams survey too early, before users have had time to form a real opinion. Responses from users with fewer than 3 sessions are not reliable. Filter them out before calculating your score.
Teams average across the entire user base and miss a strong signal in a specific segment. Always run the Ellis survey with at least two cuts: by user role and by acquisition channel.
Teams use NPS as a proxy for PMF without running the Ellis question. NPS measures satisfaction. The Ellis question measures dependency. They are not the same thing.
Duku helps early-stage and growth-stage teams build the measurement infrastructure to know exactly where they stand on PMF and what to change. If you want a structured PMF audit or a 90-day action plan, reach out.