Most founders mistake early traction for product-market fit. The ones who survive long enough to find it learn to read the right signals fast, not optimistically.
The uncomfortable truth about how to find product-market fit is that most founders are looking for the wrong thing. They want a moment: a chart that bends upward, a round that closes, a press mention that validates the idea. What they get instead is a slow accumulation of signals that most people misread, ignore, or rationalize away until the bank account makes the argument for them. Runway isn't a variable you control. The market's response to your product is. Those two facts together are why the search for fit has to be methodical, not optimistic.
Growth is a lousy early signal. It tells you your marketing worked, or your pricing was right, or you caught a lucky moment in a news cycle. Retention tells you whether any of that matters. If people come back without being prompted, that's the first honest signal you have. If they don't, no acquisition strategy in the world fixes it.
The specific benchmark varies by category. For a daily-use consumer app, weekly retention below 20% after the first month is a clear problem. For a B2B tool used for quarterly reporting, that standard doesn't apply. The question to ask isn't "what's our retention rate" but "given what this product is supposed to replace in someone's work or life, are they actually letting it?"
This is where the Sean Ellis test earns its reputation. Ellis, who helped scale Dropbox and LogMeIn before founding GrowthHackers, proposed a single survey question in 2010: "How would you feel if you could no longer use this product?" If 40% or more of respondents say "very disappointed," you're at or near fit. Below 40%, you're not. Rahul Vohra used this framework at Superhuman after the email client had already been live for a year and was growing. The survey came back at 22%, not enough. More importantly, it revealed exactly who the "very disappointed" users were: founders and executives who lived in Gmail and needed speed above everything else. Superhuman narrowed its ideal customer profile, retooled its onboarding, and crossed the 40% threshold within a few months. First Round Review documented the whole process in detail.
What product-market fit signals actually look like on the ground
Marc Andreessen's original formulation from 2007 is still the most honest description anyone has offered. "You can always feel when product-market fit isn't happening," he wrote. "The customers aren't quite getting value out of the product, word of mouth isn't spreading, usage isn't growing that fast. And you can always feel product-market fit when it is happening. The customers are buying the product just as fast as you can make it. You're hiring sales and customer support staff as fast as you can."
Slack's first day is the canonical example of the felt version. When Stewart Butterfield's team launched to a small group of companies in August 2013, 8,000 people signed up for the waitlist overnight, and daily active users hit 8,000 within the first 24 hours. The product spread because people wanted their colleagues on it. That's pull, not push. Butterfield's team had spent years failing to build a video game called Glitch. The internal communication tool they built for themselves turned out to be what the market wanted, and they followed the demand instead of defending the original vision.
The problem is most founders aren't Slack, and waiting to feel that kind of pull is expensive when runway is shrinking. The more practical version of fit detection is watching leading indicators week over week, before the feeling ever arrives.
How to measure product-market fit with leading indicators
Three metrics, watched together, tell you more than any single dashboard. First, cohort retention curves: plot each week's new users and track whether their engagement flattens rather than continues to decline. A flattening curve, even at a low absolute number, means some segment has found a reason to stay. Second, organic referral rate: what percentage of new signups name another user as the reason they tried the product? Above 20%, something real is happening. Third, for B2B, watch contraction revenue versus expansion revenue. If existing customers are upgrading and expanding more than they're downgrading or churning, that's retention expressed in dollars, and dollars don't lie the way NPS scores sometimes do.
None of these need to be large to be meaningful early on. A cohort of 50 users where 30 are still active at week eight is a better signal than a cohort of 5,000 where 100 remain. The size of the pool matters far less than the shape of the behavior inside it.
The mistake that kills companies before they find fit
The fastest way to burn through your runway without finding product-market fit is to iterate on features when you should be iterating on the customer. Founders who are talking to a heterogeneous mix, some enterprise, some SMB, some consumer, all with slightly different complaints, tend to build a product that half-solves twelve problems rather than fully solving one. The answer to flat retention is almost never a new feature. It's usually that you're serving too many different people at once.
Brian Chesky and Joe Gebbia flew to New York in 2009 to photograph the apartments listed on Airbnb themselves, because listings with professional photos converted far better. That's not a product decision. It's a customer discovery decision made visible. They found the actual barrier to conversion by being in the room with it and removed it directly. The company's early growth didn't come from a brilliant product sitting on a server. It came from founders who understood their users' real problem at a granular level.
The equivalent for most early-stage companies is getting off the roadmap entirely. Stop asking what to build next and start asking who is actually using this, why, and what they'd do if it disappeared tomorrow. Find 40 people who'd say "very disappointed" and understand precisely what they have in common. That's your market. If you can't find 40, you haven't found the thing yet, and that's more useful information than another month of building in the wrong direction.
The harder situation is when you're close but not there. The Ellis score is 28% instead of 40%. Revenue is growing but support tickets are growing faster. That zone is where a lot of companies die, not because the idea is wrong but because they can't tell whether to push forward or pivot. Go back to the "very disappointed" cohort. If those 28% share a specific job title, company size, or workflow, the problem isn't the product, it's the top of the funnel. You're acquiring users for whom the product was never right. Narrow the audience and re-run the survey in 90 days. If the number climbs, you found something real. If it stays flat across several focused cohorts, the product needs a different direction, and knowing that with money still in the bank is worth considerably more than knowing it when you don't have a choice.
Slack found fit with team messaging, then had to find it again when it moved upmarket into enterprise, and again when Microsoft Teams arrived and forced a repositioning. The original signal just means you've earned the right to keep going. What you do with that right is the actual work.
Also read: How to Get Your First 1,000 Customers as a Startup Without Spending on Ads • How to Price Your SaaS Product Without Guessing or Copying Competitors • How to Build a Personal Brand as a Founder Before You Have Anything to Sell
