To validate a startup idea, talk to ten prospective customers, build a smoke test that asks for a real commitment, run a small paid pilot instead of a free beta, and track what people do instead of what they say. Set a kill threshold before you start, so the data can change your mind.
In March 2023 I watched a founder in Austin spend four months building a scheduling tool for dog groomers before she talked to a single groomer. She had 40,000 dollars of savings in it. When she finally called ten shops, six of them already used Square for booking and saw no reason to switch. That is not a hypothetical. That is how most failed products start: with a build, then a search for someone who wants it.
Validation flips the order. You find out if anyone wants the thing, then you build it. It sounds obvious written down. It is rarely how founders actually behave, because building feels like progress and talking to strangers feels like delay.
Talk to ten prospective customers before you write code
Ten conversations is a small number on purpose. You are not running a market research study. You are checking whether the problem you think exists actually shows up in someone's week, and how they currently deal with it.
Ask about the last time the problem happened, not whether they would use your solution. "Would you use an app that does X" gets a polite yes almost every time. "Walk me through the last time this went wrong for you" gets a real answer. A SaaS founder I worked with in Denver ran this exact swap in her interview script and went from nine out of ten polite yeses to four out of ten people who could describe a specific recent incident. Those four became her first customers.
Build a smoke test that asks for money
A smoke test is a landing page, a waitlist, or a pre-order button for a product that does not exist yet. The part founders skip is the ask. A page that collects an email address tells you almost nothing, because an email costs nothing to give. A page that asks for a card, a deposit, or a signed letter of intent tells you something real.
An email address is not a customer. A deposit is a customer.
One team I advised built a 40 dollar landing page for a niche invoicing tool, ran 200 dollars of ads to three audiences, and required a 25 dollar deposit to join the beta. Eleven people paid it in nine days. That was the signal that got them to build the first version, not the 400 email signups from the version without a deposit.
Run a paid pilot instead of a free beta
Free users are not a validation signal. They are a courtesy signal. If your pilot is free, you cannot tell the difference between someone who genuinely needs your product and someone who is being polite because it costs them nothing.
Charge something for the pilot, even a fraction of what you plan to charge later. A three-person agency I spoke with in Manchester ran a paid pilot at 150 pounds a month against a planned 400 pound price point. Two of five pilot customers churned within the first month at 150 pounds. That told the founders their retention problem existed before they ever tried to charge full price, and saved them from scaling a leaky product.
- Price the pilot low enough to be easy to say yes to
- Price it high enough that saying no is a real decision
- Track churn inside the pilot itself, not just signups
Watch behavior, not opinions
Surveys measure what people think they would do. Usage data measures what they actually did. According to Nielsen Norman Group's research on user behavior, stated preferences and observed behavior frequently diverge, which is why watching a pilot user's actual weekly activity tells you more than a satisfaction survey ever will.
Set up three numbers before launch: activation rate, week-two return rate, and willingness to pay at the real price. A product with strong opinions in interviews and weak week-two return rate is not validated. It is well-liked in theory and abandoned in practice, and those are different outcomes.
Set a kill threshold before you start
Decide the number that would make you stop before you are emotionally invested in the answer. Write it down. "If fewer than four out of ten paid pilot customers renew, we stop and rework the offer" is a kill threshold. "We will know it when we see it" is not.
The founder in Austin eventually pivoted her scheduling tool toward independent mobile groomers instead of shops, after finally running the interviews she had skipped. It took her six more weeks and a much smaller rebuild than starting over. The interviews did not save her four months. They saved the next four.
Bottom Line
Learning how to validate a startup idea comes down to one habit: get real commitment before you get real conviction. Ten honest interviews, one smoke test with a price attached, and a paid pilot with a kill threshold will tell you more in three weeks than three months of solo building ever could. Pick one idea you are sitting on right now and run the ten interviews this week, before you write another line of code.