Customer Discovery Interviews: How to Talk to Users Before You Build

The most common cause of startup death is not running out of money or getting outcompeted. It is building something nobody needs, which is behind roughly 42% of failures. The cheapest insurance against it costs nothing but a few uncomfortable hours: customer discovery interviews, real conversations with the people you think have the problem, before you build anything.

Most founders skip them, and most of the founders who do them get answers that lie. Both failures have the same fix: knowing what a discovery interview is actually for, and which questions produce truth instead of politeness. Here is the working version.

What a discovery interview is, and is not

A customer discovery interview is a conversation about someone's life, not about your idea. You are trying to learn how they experience the problem today: what they do, what they have tried, what it costs them, what they gave up on. You are not pitching, previewing, or floating features.

That distinction sounds small. It is everything. The moment you reveal your idea, the conversation stops producing information and starts producing encouragement. People are kind. They will tell you your idea sounds great, that they would definitely use it, and none of it predicts anything. Enthusiasm in a conversation costs the speaker nothing, so it carries no signal.

So the whole discipline compresses into one rule, popularized by Rob Fitzpatrick's book The Mom Test: ask about their past behavior, never about your future product. Past behavior is fact. Hypothetical interest is fiction, delivered warmly.

Who to talk to, and how many

Talk to the people you believe have the problem, in the specific niche you plan to serve. If your app helps new parents fix infant sleep, you need new parents currently losing sleep, not your college friends who think babies are cute.

Two sourcing rules:

• Not friends and family. They want you to succeed, which makes them the least reliable witnesses on earth. If you must start there for practice, discard the data.
• Go where they already talk about the problem. The subreddits, Facebook groups, Discords, and forums where your niche gathers are full of people mid-complaint. A short, honest DM ("I'm researching how people handle X, could I ask you about your experience for 15 minutes? Not selling anything") gets more yeses than you expect, because people enjoy being asked about their problems.

On volume: five interviews is the minimum before any conclusion, and patterns usually stabilize somewhere between ten and fifteen. For a consumer app, 15 to 20 minutes per conversation is plenty. You are looking for repetition: when the third and fourth person independently describe the same frustration in the same words, you have found something real. Those repeated words, verbatim, are also your future landing page copy.

The questions that work

Open with a screener to confirm they actually live the problem ("How often do you..."), then work through some version of this sequence:

1. "Walk me through the last time you dealt with X." The single best question in discovery. It forces a concrete story, with real details, real tools, real friction, instead of an opinion. 2. "What have you tried to fix it?" This is the load-bearing question. Someone who has downloaded three apps, bought a book, built a spreadsheet, or paid for a course has proven the problem matters with their own time and money. Someone who has tried nothing has told you it is an annoyance, not a problem. 3. "What happened with each of those? Why did you stop?" The failures of existing solutions are your product spec. The reasons people quit them are your differentiator. 4. "What does this cost you?" Time, money, sleep, stress. Get them to say it out loud. A problem someone can quantify is a problem someone will pay to remove. 5. "When this happens, what do you do instead?" Workarounds are gold. A duct-tape workaround means the pain is real and unserved. 6. "Who else do you know who deals with this?" Ends the interview, sizes the niche, and recruits your next interview in one move.

Then mostly stay quiet. Follow up with "why?" and "tell me more about that." Silence after their first answer routinely produces the second, truer answer. Take notes on their exact phrasing, not your paraphrase.

The questions that lie to you

Strike these from your vocabulary, and treat any answer to them as noise:

• "Would you use an app that...?" Everyone says yes. Saying yes is free and ends the conversation pleasantly.
• "Would you pay for...?" Hypothetical wallets open easily. Real ones do not. The gap between "I'd pay for that" and a completed checkout is the graveyard where the 42% are buried.
• "Do you think this is a good idea?" You are asking to be complimented, and you will be.
• "Don't you hate it when...?" A leading question that supplies its own answer. You are dictating the pain, not discovering it.

If you catch yourself explaining the idea to warm someone up to a question, stop. The interview has become a pitch, and everything after that moment is contaminated.

End with a commitment, not a compliment

Words are free, so the close of a good interview asks for something that is not free. After the conversation, if they seemed genuinely engaged, make a small ask:

• "I'm putting up a page where I'll share what I build for this. Can I add you to the waitlist?" Then send the link and see if they sign themselves up.
• "Could you intro me to the two people you mentioned?"
• For bolder founders: a founding-member preorder or deposit.

Each accepted ask is real evidence, because it cost them something: their email, their social capital, their money. A pattern of warm interviews that all decline the ask is itself a finding, and a cheap one compared to six months of building.

From conversations to a decision

After ten or so interviews, sit down with your notes and answer four questions honestly:

1. Did the same problem come up unprompted, in similar words, across most conversations? 2. Have several of these people already spent time or money trying to fix it? 3. Do the existing solutions fail for a reason you can actually address? 4. Did real people accept a commitment ask?

Four yeses: build the smallest version aimed at the exact story you heard most. Mixed answers: your idea probably needs re-aiming at the sharper problem you heard, which is discovery doing its job. Mostly noes: you just saved yourself the six months and the heartbreak, for the price of a dozen conversations.

Either way, you walk away with something no amount of building produces: the problem in your customers' own words, a list of failed alternatives to beat, and a first waitlist of people who asked to be told when you launch. That is not a delay to building. That is the head start.