Product-market fit is the most cited concept in early-stage startups and the least well-measured. Here is how founders actually distinguish real traction from polite early adopters.
Every investor asks about it. Every founder claims to be chasing it. What is product market fit, exactly? Marc Andreessen popularized the term in a 2007 blog post, defining it simply as being in a good market with a product that can satisfy that market. That definition is accurate and almost useless in practice. It tells you what PMF is in the same way that "being healthy" tells you how to train for a marathon.
The problem is that PMF feels different depending on what stage you're at. In the earliest days, almost any interest from users can feel like validation. Someone pays you $50. A friend's company agrees to pilot your software. A few hundred people sign up after a TechCrunch post. None of that is product-market fit. What you're measuring is curiosity, not need.
The distinction matters because scaling before you have PMF is one of the most efficient ways to destroy a startup. You hire salespeople to sell something people don't urgently want. You spend on ads to drive users who churn in week two. You build a support team for a product that's solving the wrong problem slightly faster than it did last quarter. The costs compound. The product doesn't.
Sean Ellis, who led growth at Dropbox before founding GrowthHackers, developed the cleanest diagnostic tool available: ask your users, "How would you feel if you could no longer use this product?" Give them four options: very disappointed, somewhat disappointed, not disappointed, or I've already stopped using it. If at least 40% say very disappointed, you have PMF. Below that threshold, you don't, and you should keep iterating before pushing on growth.
Rahul Vohra, co-founder and CEO of Superhuman, took this framework and ran it in careful detail in 2018, writing about it extensively for First Round Review. When he first ran the survey, only 22% of Superhuman's users said they'd be very disappointed without the product. That's a real number from a company that had already attracted serious press attention and a waitlist. Vohra didn't panic and didn't scale. Instead, he segmented the responses: who were the people in the "very disappointed" bucket, and what specifically did they value? He rebuilt Superhuman's positioning and product roadmap around those users exclusively. A year later the number crossed 58%. That's when they started investing seriously in growth.
What makes the Ellis test useful is that it forces a distinction a lot of founders avoid: between people who like your product and people who need it. Polite early adopters will tell you your product is great. They'll give you a five-star review and then quietly go back to their spreadsheet. The 40% threshold is trying to find the users for whom your product is irreplaceable, not merely appreciated.
Retention Curves Tell You More Than Any Survey
Surveys are self-reported. Retention curves don't lie. If you look at your monthly or weekly cohort retention chart and the line keeps dropping toward zero, you don't have PMF, regardless of what users tell you in a survey. What you want to see is a curve that flattens: users who are still around at month three at roughly the same rate as month two, which means a core group has made your product a habit.
Slack is the most cited example for a reason. When Stewart Butterfield and his team pivoted from Glitch, a failed online game, to what became Slack in 2013, the retention signal was immediate and unmistakable. Teams that started using Slack just kept using it. Butterfield has described the early growth as something that felt qualitatively different from anything else he'd built: people were asking for it, not just agreeing to try it. The product entered public beta in August 2013 and drew thousands of new teams in its first 24 hours. That kind of pull is what a flat retention curve looks like in practice.
For consumer products the benchmark is often a Day 1, Day 7, Day 30 retention framework. For SaaS businesses, monthly cohort retention below 80% after the first three months is usually a red flag. There's no universal number that works across categories, but the shape of the curve matters more than any single data point: you want a line that drops steeply in the first few weeks as casual users fall off, and then levels out on a stable base of people who've genuinely built your product into how they work.
What Founders Get Wrong
Revenue is not a proxy for PMF. You can have paying customers and still be nowhere near it, especially in B2B. Enterprise deals close for reasons that have nothing to do with urgent product need: a champion at the company has political capital to spend, a vendor is being replaced, a budget needs to be used before year-end. The customer is not necessarily the user, and the contract tells you something about the sales process, not about whether anyone is actually relying on what you built.
NPS scores are even less reliable. Net Promoter Score is designed to measure loyalty at scale, not product-market fit in its early stages. A score of 50 at 200 users tells you almost nothing about whether you'll hold on to those users at 2,000. What matters is whether the people who've stopped using your product can tell you exactly why, and whether that reason is fixable or fundamental.
Word-of-mouth growth is the leading indicator most founders underweight. If your growth isn't coming partly from users telling other users, that's a signal worth taking seriously. It doesn't mean PMF is impossible without organic referrals, but referral behavior is a behavioral expression of the same feeling the Ellis survey is trying to measure: I'd be very disappointed without this product, so let me make sure someone I know has it. When Dropbox launched its referral program in 2009, signups grew 3,900% in 15 months, according to figures Drew Houston shared publicly. That result was only possible because the underlying product had already earned that kind of enthusiasm from people who were already using it.
There are also false signals that look like PMF from the inside. A product that's free, novel, or heavily promoted will attract usage that evaporates once the novelty fades or the free tier tightens. The honest test is churn under normal conditions: when you stop actively pushing users toward the product, does usage hold or does it collapse? If you have to keep pouring energy into retention just to hold your numbers flat, the product isn't doing the work on its own.
Frankly, most founders who say they're "working toward PMF" are actually working toward the feeling of PMF, which is a different thing. The feeling is confidence. The real thing is a number that doesn't depend on your mood: 40% on the Ellis survey, a retention curve that flattens, organic referrals arriving without a push. Those don't require interpretation. They don't need a VC to confirm them. You run the measurement, you read the result, and you know where you are. Everything else is story.
Also read: Revenue Based Financing for Startups Beats Equity but Only If You Do the Math • Liquidation Preferences in Startup Deals Determine Who Gets Paid at Exit • The liquidation cascade crypto traders fear most is entirely mechanical
