Every founder has had the conversation. You describe your idea, the other person nods, says it sounds useful, maybe offers a suggestion or two. You leave energised, note it as validation, and go build. Months later the product ships and that same enthusiastic person does not sign up.
They were not lying. They were being polite, which is what people do when a stranger describes something they are clearly invested in. The failure was not in their answer — it was in the question.
Customer interviews are the cheapest way to avoid building the wrong thing, and most founders extract almost nothing from them because they run the conversation as a pitch with pauses. The fix is a set of habits that are simple to describe and genuinely difficult to hold to.
Never Describe the Idea First
The moment you explain what you are building, the conversation is compromised. You have told the other person what answer would make you happy, and most people will oblige. Everything after that point is contaminated.
Structure interviews so that your idea never comes up until the end, if at all. The entire useful portion happens before the reveal. This feels deeply unnatural — you are talking to them because of the idea — but the value is in what they say when they do not know what you want to hear.
If they ask what you are working on, defer gently: "I will tell you at the end, I do not want to bias what you say." Almost everyone finds this reasonable.
Ask About the Past, Not the Future
The single most reliable rule: people are unreliable narrators of their future behaviour and reasonably reliable reporters of their past.
"Would you use a tool that automated your invoicing?" invites imagination, and imagined future selves are always more organised and more willing to pay than real ones. The honest answer is nearly always yes, and it means nothing.
"Walk me through the last time you sent invoices" produces evidence. How long it took. What broke. What they tried. Whether they went looking for a solution or simply absorbed the pain. That last detail is the one that matters most, because a problem people have not tried to solve is usually a problem they are content to live with.
Every question worth asking can be rephrased in the past tense. Do that reflexively.
Chase Specifics Relentlessly
Generalisations are where useful information goes to die. "It's pretty time consuming" tells you nothing. "It took me about three hours last Thursday because the export from the old system kept timing out" tells you everything.
When someone generalises, ask for the instance. When did that last happen? What did you do? How long did it take? What happened next? Keep pulling until you have a concrete scene with details in it.
A useful signal: if you cannot picture the situation clearly enough to describe it to someone else, you have not asked enough follow-up questions yet.
Follow the Money and the Effort
Talk is free, so opinions about a problem's severity are worth little. What people have already spent tells you far more.
Ask what they currently pay to address this — in software, in contractor time, in someone's salary. Ask what they have tried and abandoned. Ask whether anyone has built an internal workaround, because a homemade spreadsheet held together by macros is one of the strongest buying signals that exists. Somebody cared enough to build something.
Conversely, a problem that nobody has spent any money or effort on is probably an irritation rather than a need, regardless of how emphatically they complain about it.
Learn to Recognise Compliments
Compliments feel like progress and contain no information. Train yourself to notice them and treat them as a prompt to dig rather than a result to record.
"That's a great idea" — what makes you say that? What would you use it for? "I would definitely use that" — what are you using now? What would you stop doing? "Everyone I know has this problem" — who specifically? Have you talked to them about it?
Sometimes the follow-up reveals something real. More often it reveals that the enthusiasm was social rather than substantive, which is itself valuable to know before you spend a quarter building.
Sit With the Silence
The most common interviewing mistake after leading is filling pauses. A question lands, the other person pauses to think, and the founder — uncomfortable — rushes in with a clarification or a suggested answer.
That pause is where the considered response lives. The first answer people give is usually the socially convenient one; the second, arriving after a few seconds of actual thought, is the honest one.
Ask the question and stop talking. Count to five if you have to. It will feel far longer to you than to them.
Aim for Disconfirmation
Reframe the goal of the interview. You are not there to find out whether people like your idea. You are there to find out whether you are wrong about something.
Write down your riskiest assumption before the conversation — the belief that, if false, means the product should not exist. Then design questions specifically capable of disproving it. If your assumption is that finance teams spend hours reconciling invoices manually, your questions should be able to reveal that they do not.
An interview that could not possibly have changed your mind was not an interview. It was a performance.
Do Not Ask for Solutions
Users are excellent at describing problems and unreliable at designing solutions. When someone says "you should add a bulk export button," the request is data about a problem, not a specification.
Ask what they would do with it. Often you discover the real need is something else entirely — they want to share a subset with a client, and export is just the only mechanism they can imagine. The feature they asked for would have solved the problem badly.
Take requests seriously as symptoms. Take your own judgment seriously as the diagnosis.
Talk to Enough People, Then Stop
Five to eight conversations with a coherent segment will surface most of the major patterns. If you have done ten interviews and the ninth and tenth told you nothing new, you have what you need for now.
The important qualifier is coherent segment. Five interviews across five unrelated industries produce five unrelated stories and no pattern. Five interviews with people in genuinely similar situations produce a signal you can act on.
Write It Down Immediately
Memory reshapes conversations toward what you hoped to hear, and it does so within hours. Take notes during the conversation — verbatim quotes wherever possible — and write up your interpretation the same day, kept clearly separate from what was actually said.
Those verbatim quotes turn out to be enormously valuable later, and not only for product decisions. The language customers use to describe their problem is the language that should appear on your landing page, in your onboarding, and in your outbound email. Founders who interview well end up writing better copy almost as a side effect, because they stopped inventing phrasing and started borrowing it.
The Discipline Is the Hard Part
None of this is complicated. Ask about the past. Do not describe your idea. Chase specifics. Sit with silence. Look for evidence of spending. Try to be proven wrong.
It is difficult only because you care about the answer, and caring is precisely what makes leading questions so tempting. The founders who get real value from customer conversations are not better interviewers by nature. They have simply decided that being right matters more than feeling right, and built the habits that follow from it.