Customer Discovery: How to Run Interviews That Actually Tell You Something
Here's how most customer discovery goes:
You schedule a 30-minute call. You walk through your idea. You ask "would you use something like this?" The person says "oh yeah, definitely — this sounds really useful." You hang up feeling validated. You do ten more calls. They all go roughly the same way.
Then you build the thing. And nobody buys it.
The problem wasn't the conversations. The problem was what you were asking. You asked about the future, not the past. You asked about opinions, not behavior. You were running a validation exercise dressed up as discovery — and because humans are fundamentally nice and don't want to disappoint you, you got exactly the answer you were looking for.
What Customer Discovery Actually Is
Not a sales call. Not a pitch. Not a product demo with a feedback survey.
Customer discovery is a structured conversation to understand your potential customer's world — not to explain yours. You're there to learn: what problems they actually have, how they're currently solving them, what they've tried before, how they make decisions, and what it costs them when nothing changes.
If you're spending most of the call talking, something has gone wrong.
The Mom Test: Why "Would You Use This?" Is a Useless Question
In The Mom Test, Rob Fitzpatrick makes the observation that should be required reading for anyone building a product: people will lie to you, but not on purpose.
Most people are kind. They don't want to crush your enthusiasm. So when you pitch your idea and ask "would you use this?", they say yes — even when they wouldn't. They're not deceiving you; they're being polite.
Fitzpatrick's insight: you have to make it impossible for people to lie to you — not by confronting them, but by asking questions they can't answer politely. Specifically, questions about the past and present, not the future.
"Would you use this?" is a hypothetical about a future that doesn't exist. Compare it to: "Tell me about the last time you dealt with this problem. What did you do?"
That question has a real answer. If they can't recall a specific recent instance, that's extremely important information about how acute the problem actually is. Past behavior doesn't lie the way future intent does.
Customer Interview Questions That Actually Work
To understand the problem:
- "Tell me about the last time you dealt with [the problem]. What happened?"
- "How often does this come up? When did you last run into it?"
- "Walk me through what you did when it happened."
To understand current solutions:
- "How are you currently handling this? Walk me through your process."
- "What tools or workarounds do you use today?"
- "What do you hate about the way you're doing it now?"
To understand what's been tried:
- "What have you already tried to solve this? What worked, what didn't?"
- "Have you looked for a better solution? What stopped you from switching?"
To understand cost and priority:
- "What does it cost you when this goes wrong — in time, money, frustration?"
- "Where does this fall on your priority list?"
- "What would need to be true for you to actually pay for a solution?"
To understand the decision:
- "If you found something that solved this, who else would need to sign off?"
- "Who else in your organization deals with this problem?"
Questions to avoid: "Would you use X?" / "Do you think X would be valuable?" / "Would you pay for X?" / "What features would you want?" — all hypothetical, all produce invented answers.
What Good Customer Discovery Actually Reveals
The most valuable thing you learn is often not what you expected.
Workarounds are the strongest signal a problem is real. If someone has built an elaborate spreadsheet, a multi-step manual process, or a hack involving three tools duct-taped together — that's not just a problem. That's a problem painful enough that they built their own solution. Workarounds are a far stronger signal than "yes, that's annoying." They tell you the person cared enough to do something about it.
The buyer isn't always who you assume. Many founders discover mid-interview that the person they're pitching isn't actually the decision-maker — and the person who controls the budget has completely different priorities. "Who else would need to be involved in a decision like this?" asked every time prevents you from building a product for the wrong person.
Price reveals itself through context, not direct questions. "Would you pay $X/month for this?" is a hypothetical that produces an invented answer. But "what are you currently spending to handle this?" and "what did it cost the last time this went wrong?" give you real data about willingness to pay without asking anyone to commit to a number.
How Many Interviews Do You Need?
It depends on the consistency of what you're hearing.
In qualitative research, "saturation" describes the point at which additional interviews stop revealing new information. In practice, founders often start seeing clear patterns after 5–10 interviews with the same customer type.
The goal isn't statistical significance. You're looking for signal: consistent, repeating themes that suggest a real, common problem. If 7 of 10 people describe the same pain in nearly identical terms — unprompted — that's meaningful. If answers are all over the map after 10 conversations, either the problem isn't universal or you're talking to the wrong people.
Interview within a customer type. Five conversations with HR managers at 50-person startups teach you different things than five conversations with HR directors at Fortune 500 companies. Segment deliberately. Don't mix signals.
How Discovery Connects to ICP, Value Proposition, and PMF
Customer discovery is the foundation everything else is built on.
When you do discovery well, you learn who has the problem most acutely — directly shaping your ideal customer profile. Not who you wish would buy your product, but who actually has the problem, has budget to solve it, and is motivated to change.
The language your customers use to describe their problems is the raw material for your value proposition. When you describe back the problem in their own words — their language, not your jargon — it lands differently. It sounds like you understand them, because you do.
And product-market fit is validated when customers describe the value you deliver in the same terms you used when you understood the problem in discovery. The loop closes. Discovery informs ICP. ICP focuses discovery. Value proposition comes from what you hear. PMF is evidence you heard it right.
What Customer Discovery Can't Tell You
Discovery is the qualitative layer — the texture of the pain, the workarounds people invented, the language they use when nobody's pitching them.
But it won't tell you how many people have this problem. It won't tell you what the market is already paying to solve it. It won't tell you whether five well-funded competitors are already competing for the same customer.
That's the quantitative layer. DimeADozen.AI handles it — market size, competitive landscape, demand signals, pricing benchmarks. Built from real market data. The conversations tell you what to build. The market data tells you whether it's worth it.
Summary: The Rules of Good Customer Discovery
- Ask about the past and present. Never "would you" — always "what did you" and "how do you."
- Your job is to listen. If you're talking more than 30% of the time, recalibrate.
- Workarounds are the strongest signal a problem is real and painful.
- Patterns emerge after 5–10 interviews with the same customer type. Trust the repetition.
- The buyer isn't always who you think. Ask "who else is involved?" every time.
- Qualitative and quantitative research are complementary, not interchangeable.
Start with the conversations. Then run the numbers →