Your Survey Said 78% Would Pay for This. Your Launch Said Otherwise. Here's Why.

TL;DR

• Surveys measure intent, not behavior, and the two are not the same thing.
• People don't lie in surveys to deceive you, they just have no skin in the game, so a "yes" costs them nothing.
• A stated preference ("I'd pay for that") predicts almost nothing about a real purchase decision.
• If there's no cost to saying yes, the yes means nothing. Design tests that require an actual decision instead.
• A fake checkout or a real deposit will tell you more than a thousand survey responses.
• Companies like Zappos, Airbnb, Stripe, and Buffer validated demand by manually testing real decisions, not by asking opinions.
• Replace "would you pay for this?" with something that finds out whether they actually will.

Your survey said 78% would pay for this. You launched. Four percent did.

Nothing about your survey was dishonest. Nobody was lying to you. The number was real, the responses were sincere, and it still told you almost nothing about what would actually happen when money was on the line.

People Lie in Surveys, Just Not on Purpose

Nobody fills out a survey intending to mislead you. But answering "would you pay for this?" costs a respondent nothing. There's no wallet involved, no calendar commitment, no real trade-off against the other things they could be doing with their money. Saying yes is easy, free, and mildly pleasant, because it feels like being helpful.

That's the whole problem. A "yes" that costs nothing tells you nothing. It measures how a person feels about your idea in the abstract, in a moment with zero consequences, which is a completely different question from whether they'll open their wallet when the abstract becomes real.

Intent Is Not Behavior

"Would you pay for this?" is a question about intent. "Did you pay for this?" is a question about behavior. Founders routinely ask the first question and act as if they've answered the second.

Intent is cheap to state and expensive to predict from. Someone can genuinely believe they'd pay for your product and then, faced with an actual checkout screen, get distracted, deprioritize it, decide the price is a little too high this month, or simply forget it exists by the time they'd need to act. None of that makes their survey answer a lie. It just means intent was never a reliable stand-in for what they'd do.

If you want to know what someone will do, you have to put them in a position where they actually do something, not just where they say something.

The Only Validation Question That Actually Predicts Purchase Behavior

There isn't a clever survey phrasing that fixes this. The fix isn't a better question, it's a different kind of test entirely.

Stop asking people to imagine a hypothetical. Give them a real decision, even a small one, and see what they do with it. Ask for an email in exchange for early access. Ask for a deposit to hold a spot. Put up a checkout page and see how many people actually click "buy" for a product that doesn't exist yet. Every one of those requires something a survey never does: a real cost, paid in time, money, or reputation.

The question isn't "would you pay for this?" The question is "will you, right now, do the smallest real version of paying for this?"

The Smoke Test That's More Predictive Than 1,000 Survey Responses

A smoke test is exactly this: a landing page, a fake checkout, a waitlist that requires a card on file even if it's never charged. It's not deceptive if you're upfront about the product not existing yet. What it gives you is a real signal instead of a polite one.

Run the survey if you want, it's not useless as a first pass at understanding language and framing. But treat it as generating hypotheses, not confirming demand. Then run the smoke test and watch what people actually do when the yes has a cost attached. The conversion rate on that page, not the percentage who said yes to a hypothetical, is the number that predicts your launch.

How Real Companies Tested This Before They Scaled

Zappos started because Nick Swinmurn wanted to know if people would actually buy shoes online, not whether they said they would. Instead of building inventory or a warehouse, he photographed shoes at local stores, listed them on a simple website, and when an order came in, he'd walk to the store, buy the pair at full price, and ship it himself. No survey could have told him what mattered: whether real people would enter a card number and buy shoes sight unseen. The orders did.

Airbnb's founders didn't run a market research study to see if people would rent out their spare rooms to strangers. They found their first hosts, photographed the listings themselves, and personally managed the early bookings by hand. Watching real bookings happen, not survey answers about willingness to host, is what told them the model worked.

Stripe's founders skipped the pitch entirely. Instead of asking developers whether they'd want a simpler way to accept payments, they'd sit down next to them and install the actual integration on the spot, a practice that became known as the "Collison install." What they were measuring wasn't opinion, it was whether a working integration got used once it existed.

Buffer tested willingness to pay before writing a line of the actual product. Joel Gascoigne built a two-page site: the first page described the idea, and clicking "plans and pricing" led to a second page with real prices and a signup form. The click-through rate on that second page, a real decision path rather than a survey question, told him whether people wanted it enough to look at a price tag.

Each of these skipped the step where you ask people what they think. They went straight to giving people a real decision to make, then watched what they did with it.

Behavioral Evidence Beats Stated Preference, Every Time

None of this means people are bad at predicting their own behavior on purpose. It means stated preference and revealed preference are measuring different things, and only one of them is measuring what you actually need to know before you build.

The next time a survey hands you an encouraging number, don't build on it. Design a test that costs your respondents something real, and let their actual behavior, not their agreeable answer, tell you whether you have something worth building.

A survey tells you what people are willing to say. A checkout page tells you what they're willing to do. Only one of those is the number that matters.