In 2019 I backed a Finnish enterprise software company with every visible signal in place: real traction, a strong team, good timing, revenue about to land. The story was easy to believe. Three customers had signed. None of them had changed how they worked, who owned the risk, or what they were willing to pay to avoid it. The company had wins. It did not have a system underneath them.
That gap is not rare, and it is not new. Interviews had suggested interest. A couple of early deals had looked like demand. Discovery had done exactly what discovery is supposed to do: it pointed at something worth testing. The mistake came one step later, when interest quietly became a growth narrative before anyone checked whether a real buyer had moved a real budget for a real reason. Discovery suggests. Proof validates. Across more than 10,000 startups I've reviewed as an investor, confusing the two is the single most expensive mistake founders make, more expensive than any one bad hire or wrong market, because it doesn't announce itself until the bill arrives.
Founders call the resulting confusion product-market fit, and the term deserves its reputation as the most overloaded phrase in the industry. Everyone defines it differently, so advice about it has had to survive translation by staying general, and general advice about a stage with this many moving parts is barely advice at all. Fit is also diagnosed backward: you find out you had it once growth confirms it, by which point a founder has usually already hired against it, raised against it, and built a roadmap on the assumption it was already there.
That backward-looking version of fit used to be survivable, because the product itself functioned as the moat. If a founder misjudged fit, the product (code, features, interface polish) was hard enough to copy that a competitor needed real months to close the gap, long enough for the mistake to surface and get corrected quietly. I now watch pitch decks and working demos get assembled in weeks that would have taken a small team most of a year to build a decade ago. That shift removes the cushion. When the product itself stops being defensible, buyer-by-buyer understanding of who actually pays, why, and what survives their real cost of change becomes close to the only moat left. Reading fit slowly, after the fact, used to cost a company a difficult year. It can now cost the company outright, while the founder is still feeling toward the right answer.
The fix isn't a sharper feeling. It's a different question, asked one layer at a time instead of all at once. Buyer Proof asks who pays: a specific person with budget authority, a visible cost to whatever workaround they're using today, and a reason to act now instead of next quarter. Value Proof asks what survives: whether customers who resemble your best customer get measurable value once the full cost of adopting you (implementation, training, the internal risk someone took by championing you) is subtracted from what you sold them, and whether enough of that value returns to your business rather than staying with the customer while your team quietly absorbs the delivery work. Read this way, product-market fit isn't a single moment a founder arrives at. It's several rounds of Buyer Proof and Value Proof holding at once, across more than one customer, without the founder personally in the room closing the gap each time.
Product admiration is the most convincing false signal here, because it feels like validation. A user who says they love the product, opens it daily, and would be sad to lose it hasn't told you whether they've changed the behavior your company's survival actually depends on. The cleanest test cuts through this: if the product vanished tomorrow, what would literally stop working for that customer, not what they'd miss, what would break. Admiration that doesn't survive that question is real. It's also still discovery, dressed as an answer.
Run the test directly against your own numbers. Take out the friendly buyers, the discounted deals, and the accounts your best salesperson personally rescued. Look at what remains. Do the wins that are left share a buyer, a trigger, and a reason for acting now, or did each one close through its own separate exception? Would one of those customers reach the same value again, under the same conditions, without you or your top closer carrying the deal? If you quietly removed the extra support your team provides to keep an account happy, would the relationship survive? An honest "we haven't checked" to any of these means you don't have fit yet. You have discovery that was never tested, wearing a revenue number as a disguise.
None of this argues for distrusting every early signal or slowing down to prove everything twice. Interviews, usage, and early revenue are supposed to point somewhere; gathering that evidence was never the mistake. The mistake is promoting it into a verdict before a specific buyer, a specific behavior, and a specific number have actually said so. That discipline costs more time up front than declaring the round won on a good quarter. It costs far less than discovering the same gap eighteen months and one funding round later, in front of a board instead of a bank statement.
The Finnish company eventually earned real Buyer Proof, in a segment considerably narrower than the one its early revenue had suggested. It took longer than it should have, because the first wins got treated as the destination instead of a question still open. The company that gets there first from here won't be the one with the best product. It will be the one that can say, specifically, which buyer, which trigger, and which number already proved it, because the product stopped being the moat, and the understanding underneath it is now the only one left.
More scars than trophies.