How to Talk to Customers: A First-Time Founder's Guide

The single highest-leverage skill for a first-time founder isn't coding, design, or pitching. It's the ability to sit across from a stranger and hear what's actually true. Almost every failed early-stage venture failed because the founder couldn't (or wouldn't) do that.

The good news: it's a skill, not a personality trait. You can get noticeably better at it in a week. This guide is what we wish someone had told us before our first 50 interviews.

The one rule that fixes 80% of bad interviews

Ask about the past, never about the future.

"Would you use this?" "Would you pay for it?" "Do you think this is a good idea?" — all garbage. People are extremely bad at predicting their own future behaviour, and even worse when they're trying to be nice to a founder who clearly wants a yes.

"Tell me about the last time you ran into this problem" — gold. The past is real. The future is a wish.

Find the right people

Five strangers in the problem beat fifty friends with opinions. Strangers don't love you, don't owe you, and don't soften the truth. Find them in:

• Reddit threads where people complain about the problem
• LinkedIn comment sections under industry posts
• Slack / Discord communities of practitioners
• Reviews of competitors (one-star and three-star, not five-star)
• The person who introduces you to two other people with the problem

How to ask for the interview

Short, specific, no pitch. Something like:

"I'm researching how [specific people] handle [specific problem] today. Not selling anything. Could I ask you about your experience for 20 minutes this week?"

Don't say "founder." Don't say "validate." Don't attach a deck. The fewer words the better.

The interview structure that works

1. 2 min — set the frame. "I'm trying to understand X, not pitch anything. There are no wrong answers."
2. 3 min — context. Their role, their day, the shape of their work. Build the rapport, learn the vocabulary.
3. 10 min — the past. Walk through the last specific time the problem happened. What did they do? What did it cost? What did they try? Why did they stop?
4. 3 min — the workaround. What are they using today, however ugly? Spreadsheets, tools, people, manual processes.
5. 2 min — the ask. "Who else like you should I talk to?" Always.

That's it. Twenty minutes. No demo. No pitch. No "so what we're building is…" — that ruins the next 30 interviews because they'll all start telling you what your thing should be.

Questions that pull truth out

• "Walk me through the last time this happened."
• "What did you do right before that?"
• "What did you try that didn't work?"
• "How much time / money did it actually cost you?"
• "What would you have to stop doing to use a new solution?"
• "What did you Google?"
• "What's the workaround you're a bit embarrassed about?"
• "Who else has this problem worse than you?"

Questions that poison the data

• "Would you use a product that…?"
• "How much would you pay for…?"
• "Do you think this is a good idea?"
• "What features would you want?"
• Anything starting with "Imagine if…"

Signals to listen for

• Emotion. Frustration, sighing, swearing — real problems have emotional residue.
• Workarounds. The uglier and more elaborate, the bigger the pain.
• Money already spent. Existing budget is a far stronger signal than enthusiasm.
• Specifics. "Last Tuesday" beats "we usually." Vague answers mean the problem isn't lived.
• Unprompted introductions. "You should talk to my friend X" — biggest tell of all.

Signals to ignore (even when they feel great)

• "Cool idea!"
• "I'd definitely try it."
• "You should talk to my investor friend."
• Polite enthusiasm with no specifics.

If you hang up feeling great but can't quote one specific thing they said, the interview was a hug, not data.

What to do right after the call

Spend 5 minutes writing down:

1. The exact words they used for the problem (you'll want them for your landing page).
2. The workaround they currently use.
3. What surprised you.
4. What you now want to ask the next person.

The fourth one is how interviews compound. By interview 10, you're not asking the same questions you started with — you're testing hypotheses that came out of interviews 1–9.

How many interviews?

Until you can predict what the next person will say. Usually that's 10–20 in any given segment. If you can't predict yet, keep going. If you can, change segment or change question.

How to practice this on Playground

Talking to real customers is uncomfortable, especially at the start — so most founders avoid it just long enough for their venture to die. The Chat module includes a fictitious customer persona you can practice with: ask the questions out loud, hear how they land, get coached on what you missed. It's the cheapest rep you'll ever do. And if selling specifically is the scary part, see founder-led sales.