How to know if you have product-market fit (signals that survive scrutiny).

Most founders claim product-market fit before they have it. The reason is that the early signals (a few enthusiastic users, a wait-list that grows, a press hit that lands) feel like fit, and the difference between feeling and proving is rarely emphasized in founder education. Product-market fit is a specific pattern in the data that survives cohort breakdowns, channel breakdowns, and unfriendly conversations. If your claim breaks under any of those, you do not have it yet.

The single most defensible signal of product-market fit is a retention curve that flattens. According to research published by Lenny Rachitsky surveying 50+ B2B SaaS companies that reached $10M+ ARR, the median company that achieved fit had a 12-month net revenue retention above 110 percent and a logo retention above 85 percent. Companies that later failed to scale had retention curves that looked promising at month one but continued declining through month twelve.

The retention shape matters more than the absolute number. A curve that starts at 80 percent and flattens at 60 percent is fit. A curve that starts at 95 percent and continues falling at month nine is not fit, even if the month-three number looks better. The flattening is the proof. It means a subset of users found durable value and stayed for it.

Sean Ellis popularized a survey question that has become the most-cited PMF benchmark in the category: "How would you feel if you could no longer use this product?" The response options are "very disappointed," "somewhat disappointed," and "not disappointed." If at least 40 percent of users answer "very disappointed," the product has reached product-market fit, according to Ellis's framework.

The threshold is not arbitrary. In Ellis's analysis of companies that did and did not scale, 40 percent was the inflection point above which growth became efficient and below which growth required brute force. The test only works when administered to actively engaged users, not the full user base. Including dormant users inflates "not disappointed" responses and produces a false negative.

The economics of acquisition tell you whether the market is pulling or you are pushing. According to ProfitWell's 2023 benchmark report covering 16,000 SaaS companies, businesses that reached durable fit acquired between 25 and 40 percent of new customers through organic channels (word of mouth, search, direct) by the time they crossed $5M ARR. Businesses that scaled before fit acquired more than 75 percent of customers through paid channels and had CAC payback periods exceeding 24 months.

The organic share is the more honest signal. Paid acquisition can mask the absence of fit by buying growth that does not compound. Organic acquisition only happens when users find enough value to recommend the product without being incentivized to do so.

If you ask five recent customers to describe what your product does and why they bought it, fit looks like five answers that converge on the same problem and the same outcome, in language the customers themselves use. Lack of fit looks like five answers that describe different products, or five answers that match your marketing copy verbatim because the customers cannot articulate the value in their own words.

This test is qualitative and unreliable in isolation, but it is a useful contradictor for quantitative signals that look optimistic. Healthy retention combined with five customers who cannot agree on what the product is signals that you have not yet found the segment.

A waitlist of 10,000 signups is not product-market fit. A TechCrunch feature is not product-market fit. A pilot with a Fortune 500 logo is not product-market fit. None of these are fit because none of them measure whether a subset of users found durable value and chose to keep paying for it. They measure interest, press, and procurement risk, respectively. Interest is cheap. Press is cheaper. Procurement risk only converts to revenue at the end of a long, expensive process.

Founders who declare fit on these signals end up running paid acquisition into a leaky bucket, raising too much money against premature scale claims, and burning eighteen months learning that the early signals did not predict the later ones.

Product-market fit is provable. The proof is a retention curve that flattens, a Sean Ellis score above 40 percent on engaged users, an organic acquisition share that climbs as you scale, and a conversation with recent customers where they describe the same product in their own language. Any subset of these can be faked. The full pattern is hard to fake. If you have all four, you have fit. If you have one or two, you have promising signals that deserve more time. If you have none, you are still searching, and the most important thing you can do is not pretend otherwise. For founders running this evaluation, our how to validate a startup idea framework pairs naturally with the PMF signals above, and the Verdikt methodology shows how we stress-test fit claims inside a memo.

Sean Ellis’s framework asks "how would you feel if you could no longer use this product?" Responses of "very disappointed" above 40 percent of users are the canonical PMF signal. The framework is widely cited because the threshold is concrete and the question is binary enough to survey at scale. Sean Ellis’s original writing on the test explains both the methodology and the limits.

Honest application requires three discipline points. First, survey active users (engaged in the last 30 days), not your whole user list. Inactive users will skew "not disappointed." Second, segment the cohort. Total scores hide segment-level fit. A 35 percent total might be a 65 percent "very disappointed" in one segment and 12 percent in another. The first segment is where the company lives. Third, run the test repeatedly. PMF is not binary; it moves as the product evolves.

Survey-based PMF tests catch up to behavior. The behavioral signals usually show up first. Signal one is retention curves that flatten rather than asymptote toward zero. A cohort retention curve that plateaus at 35 percent at week 12 and stays there is fit; a curve that decays through 5 percent is not. ChartMogul’s SaaS retention benchmarks document the shape of healthy curves across stages.

Signal two is organic word-of-mouth. When users actively refer the product to peers without asking, the value is strong enough that the user takes a social risk. Track this as a percentage of new signups attributed to direct referral. Pre-PMF teams see 5 to 10 percent; PMF teams see 30 to 50 percent.

Signal three is willingness to pay above the price ceiling. If you can raise prices 30 percent and see no change in conversion, you are priced below the value the customer perceives. The price ceiling is the PMF signal that founders are most afraid to test, but it is the cheapest one to run.

Signal four is unprompted feature requests that share a coherent narrative. Pre-PMF teams get requests across a wide and incoherent surface; PMF teams get requests that cluster into 2 to 3 themes, which means users have an internal model of what the product is and are extending it.

Three patterns look like PMF and are not. First, founder-network usage: friendly users sign up, give feedback, and stay engaged because they want the founder to succeed. Strip these from the cohort before measuring anything. Second, free-tier engagement that does not convert to paid: free users behave as if they have product-market fit because the cost-benefit is asymmetric. Measure PMF only on paid cohorts. Third, growth driven by a single paid channel that is structurally unscalable. If the channel breaks, the apparent fit evaporates because it was channel fit, not product fit.

Marc Andreessen’s essay on product-market fit is the original modern statement of the concept and remains the cleanest reference. The key sentence: "The customers are buying the product just as fast as you can make it." The behavioral test is whether you are demand-constrained or supply-constrained. PMF is the demand-constrained state.