Customer Discovery

How to talk to real people before you build anything.

Start · 6 min

What is this?

Customer discovery is the practice of talking directly to the people you think have the problem you want to solve, before you spend time or money building a solution.

It's not a sales pitch and it's not a survey about your idea. It's a conversation about their life, their workflow, or their frustrations — with your idea barely mentioned.

Why it matters

Most failed startups don't fail because the product was built badly. They fail because nobody actually wanted it. A handful of honest conversations early on is the cheapest way to find that out before you've invested months building something.

What you're trying to learn

Learn whether the problem you think people have is one they actually, currently experience — in their own words, not yours.

Before you start

Pick a specific type of person, not "anyone who might use this" — e.g. "ops managers at 20-50 person logistics companies," not "business owners."
Find 10-20 of them through your own network's edges, relevant online communities, LinkedIn search, or a niche forum — not just people who already like you.
Reach out with a short, honest ask: you're researching the space, not selling anything, and you want 15-20 minutes to learn how they currently handle something.
Block real time for each conversation and for writing notes right after — a rushed interview or a forgotten write-up loses most of the value.

Step by step

1List 10-20 people who plausibly have the problem you're interested in — not friends and family who will just be nice to you.
2Ask for 15-20 minutes to learn about how they currently handle the situation, not to pitch them anything.
3Ask about their past behavior and specific recent examples, not hypothetical future behavior ("Tell me about the last time this happened" beats "Would you use an app for this?").
4Let them talk more than you do. Silence is fine — resist the urge to fill it by explaining your idea.
5Write down what you heard right after each conversation, while it's fresh.
6Look for patterns across conversations, not conclusions from any single one.

What to ask / do

"Walk me through the last time [the problem] came up. What were you doing right before?"
"What did you do about it? Who else was involved?"
"How did you end up handling it in the end?"
"What did that cost you — in time, money, or something else?"
"How often does this come up?"
"What have you already tried to fix this? What happened?"
"If you had a magic wand, what would you change about how this works today?" (only near the end, if at all)

Example conversation

YouThanks for making time. I'm not here to sell you anything — I'm trying to understand how teams like yours currently handle scheduling changes. Can you walk me through the last time a shift change came up?

ThemSure — last Tuesday, actually. One of our drivers called in sick at 6am.

YouWhat happened right after that call?

ThemI had to call around to three other drivers to see who could cover, then update the route sheet by hand.

YouHow long did that take, start to finish?

ThemMaybe 40 minutes, and I was late updating dispatch.

YouHas that kind of thing happened before? How often?

ThemAt least once a week, honestly.

YouHave you tried anything to make that easier?

ThemWe looked at a scheduling app once but it didn't handle our routes well, so we gave up on it.

What the signal looks like

Good signal

They describe a specific, recent, real event with concrete details (who, when, what happened) without you prompting for specifics.
They've already tried something — a tool, a workaround, a spreadsheet, hiring help — to deal with the problem themselves.
They get visibly more animated or frustrated describing the problem, unprompted.
The same underlying story shows up, in different words, across multiple unrelated interviews.

Weak or negative signal

"That's a great idea" or "I'd probably use that" — a hypothetical reaction to your idea, not a report of their actual experience.
Vague, generic answers with no specific example ("yeah, that happens sometimes") even after you ask for a recent instance.
Enthusiasm only appears after you describe your solution, not while they were describing their situation.
They can't recall the last time it actually happened when asked directly.

Common mistakes

Describing your solution and asking "would you use this?" — almost everyone says yes to be polite.
Only talking to people who already agree with you.
Treating one enthusiastic conversation as proof, or one lukewarm one as failure.
Asking about the future ("would you pay for X") instead of the past ("what did you do last time").

Checklist

You have talked to at least 10-15 people outside your existing network.
You asked about their past behavior, not their predicted future behavior.
You wrote down notes after each conversation.
You can describe a real, specific example a real person gave you — not a paraphrase of your own idea.
You've noticed at least one pattern across multiple conversations.

You're done when: You've completed 10-15 real conversations outside your existing network, written notes after each one, and can point to a specific pattern that showed up more than once — in the interviewees' own words, not your paraphrase of them.

What to do next

If a real, repeated pattern showed up: turn it into a Mission for the next step (often Problem Validation, to check how painful and frequent it really is).
If interviews mostly produced polite, hypothetical answers: that's useful information too — go back to who you're talking to, or reconsider the problem itself, before building anything.
Either way, use Record What I Learned on the related Mission to capture what you found — that's the only step that can update your model.