Talking to Users

Everyone tells you to talk to users. Almost nobody tells you that most founder interviews are worthless, because they're designed, unconsciously, to collect compliments.

The concept

You cannot learn whether people want your product by asking them whether they want your product. People are polite, they're bad at predicting their own future behaviour, and they will happily tell you an idea sounds great because the alternative is an awkward conversation.

The purpose of a user interview is not to validate your idea. It is to learn about the customer's life, what they currently do, what it costs them, what they've already tried, and what they've already paid for. Their past behaviour is data. Their predicted future behaviour is fiction.

This principle is usually credited to Rob Fitzpatrick's The Mom Test: ask questions so grounded in fact that even your mother couldn't lie to you about the answer.

The rule, in one table

The reframe: stop asking about your idea. Ask about their life. If their life is full of the problem, you have something. If you have to explain why they should care, you don't.

The three signals that actually mean something

1. They already pay for a solution, money is the only unambiguous vote. Even a bad, expensive incumbent proves a budget exists.
2. They've built a workaround, a spreadsheet, a Zapier chain, a WhatsApp group, a hired VA. A workaround is a problem someone cared enough to solve badly.
3. They can tell you when it last happened, specifically, vivid, dated, detailed recall means it's real. Vagueness means it isn't a priority.

Everything else, enthusiasm, "that's cool," "I'd definitely use that", is noise.

What "enough interviews" means

• 5 interviews, you'll hear the same thing twice; enough to correct a badly wrong assumption
• 10-15, patterns stabilise; this is the normal target for a first product
• 20+, diminishing returns unless you're segmenting

Stop when new interviews stop surprising you. That's saturation, and it usually arrives around 12.

B2B vs consumer

These are genuinely different exercises:

In B2B, always find the buyer, not just the user. The person who suffers the problem often cannot authorise spending. Interviewing five enthusiastic users and zero budget-holders is how B2B founders spend a year building something nobody can purchase.

📐 Best practice

Ask about the past, never the future. "When did this last happen?" not "would you?"

Shut up. The interview is going well when you're talking less than 20% of the time. Silence after an answer is the highest-yield technique there is, people fill it with the real answer.

Ask "why" three times. The first answer is a rationalisation. The third is usually the truth.

Don't pitch until the end, if at all. The moment you describe your solution, the interview becomes a politeness exercise. If you must demo, do it in the last five minutes and treat everything after as contaminated.

Ask for specifics, then follow the specifics. "What did you do next?" "How long did that take?" "What did that cost you?"

Take verbatim notes. Their exact words become your copy (Chapter 43). Paraphrasing destroys the most valuable output.

Record with permission, transcribe, and tag. Otherwise you'll remember the interviews that agreed with you.

End with: "Who else should I talk to?" Referral is the cheapest recruiting channel and it compounds.

Look for the commitment, not the compliment. The real close is: "Can I follow up when I have something?" or "Would you be willing to pay a deposit?" Enthusiasm is free; commitment costs them something.

💀 Common mistakes

Pitching instead of listening. The default failure. You leave with compliments and no information.

Asking leading questions. "Don't you find it frustrating when…?" tells them the answer you want.

Only interviewing friends. They will lie to protect the relationship. Useful for practising the script; useless for evidence.

Only interviewing enthusiasts. People already deep in a category are unrepresentative. Talk to people who have the problem but aren't obsessives.

Treating "I'd use that" as validation. It costs nothing to say. Ask what they use now and what they pay.

Not writing down verbatim quotes. You lose the exact language, which is the most commercially useful artifact.

Stopping after three interviews because they went well. Three polite conversations is not a signal.

Skipping interviews entirely because you "are the user." Being your own customer is a genuine advantage, it gives you speed and taste, but you are one data point, and probably an unusual one. It shortens the research; it doesn't replace it.

Interviewing users but not buyers (B2B). Fatal and common.

The professional workflow

 1. WRITE THE HYPOTHESIS you're testing, in one sentence.
    "I believe [customer] struggles with [problem] and currently
     handles it by [workaround]."

 2. RECRUIT 10-15
    Consumer: communities, subreddits, Discords, IG/TikTok comments,
              friends-of-friends (2 degrees out), $20-50 incentive
    B2B: LinkedIn search by title, industry Slack, warm intros,
         cold email offering to share findings

 3. SCRIPT — 6-8 questions, all past-tense. (Template below.)

 4. INTERVIEW — 20-30 min. Record. Talk <20% of the time.
    Follow specifics. Ask why three times. Don't pitch.

 5. NOTE VERBATIM QUOTES immediately after, while fresh.

 6. TAG AND CLUSTER
    Per interview: current solution · workaround · trigger moment ·
    what they've paid for · emotional driver · exact phrasing

 7. LOOK FOR: repeated problems · repeated language ·
    existing spend · existing workarounds

 8. DECIDE: is the problem real, frequent, and expensive enough
    that someone would change their behaviour?

A script that works in any category

1. "Tell me about the last time you [did the thing]."
       → then follow every specific. This is 40% of the interview.

2. "How do you handle that today?"
       → the incumbent, including 'nothing' and 'a spreadsheet'

3. "What's the most annoying part of that?"
       → the wedge

4. "Have you tried anything else? What happened?"
       → the graveyard. Why did previous solutions fail?

5. "What does it cost you when it goes wrong?"
       → time, money, stress — this is your value proposition

6. "Do you pay for anything in this area?"
       → budget exists / doesn't

7. (B2B) "If you wanted to buy something like this, who'd have to say yes?"
       → the real buyer

8. "Who else should I talk to?"
       → compounding recruitment

Tools, websites & costs

Realistic cost: $0 if you recruit from communities, $300-750 if you pay 15 consumer participants $20-50. B2B usually costs nothing but time.

Time cost: ~10 hours for 15 interviews including scheduling and synthesis. It is the cheapest de-risking available.

Alternatives & trade-offs

Surveys scale and are shallow. Good for quantifying something qualitative interviews surfaced ("how many of you also use a spreadsheet?"). Bad for discovery, you can only ask what you already thought of.

Analytics on an existing product beats interviews for what people do, and tells you nothing about why. Use both: analytics finds the cliff, interviews explain it.

Support tickets and reviews are free, honest, and biased toward the angry. Excellent raw material; not a substitute for conversation.

Public review mining (Chapter 03) is the cheapest substitute when you genuinely cannot reach users. It gives you real language and real complaints from real people, but you can't ask a follow-up, and that's a large loss.

Being your own customer is a real, legitimate advantage, Superhuman, Linear and Basecamp all started this way. It gives you taste and speed. Its failure mode is building for an unusual version of yourself. Mitigate by interviewing 10 people who are like you but not you.

"Just ship it and watch" works when building is genuinely cheap and the market is unknowable. It is a very expensive way to learn something an afternoon of conversation would have told you.

Checklist

• I wrote the hypothesis I'm testing before scheduling anything
• I have talked to 10+ people who are not friends
• Every question is about the past, not the future
• I talked less than 20% of the time
• I have verbatim quotes, not paraphrases
• I know what each person currently uses and currently pays
• I found at least one workaround someone built
• (B2B) I have spoken to someone who can authorise spending
• I asked "who else should I talk to?" every time
• New interviews have stopped surprising me
• I can state the problem in my customers' words, not mine

📓 Case Study: the shortcut, and what it costs

Project: SOLIS. This case study is included because of what didn't happen.

No user interviews were conducted. Not one. The founder went from market research directly to product design.

That is a deviation from best practice, and it's worth being precise about what substituted for it, because the substitutes were not nothing:

• Public review mining at scale. The universal complaint, "Most Stoic apps show you a quote, ask you to journal, and charge $50/year… they skip the hard part… more flavor than substance", came from real users in public, in their own words. That's genuine voice-of-customer.
• Founder-as-customer. The founder was adjacent to the target audience and had the problem.
• A competitor's funnel, screen-recorded end to end. Watching a $20k MRR product's onboarding is a form of behavioural research, you're seeing what a company with real conversion data decided to ask users.
• Demographic evidence. Stoic Week surveys showing 60% male participation (2019), 57% (2020), versus meditation apps at ~65% women, established the audience existed at the scale assumed.

What that stack genuinely provided: the language, the complaint, the price anchor, and confidence that the audience was real.

What it could not provide, and this is the point:

1. No follow-up questions. A review says "more flavor than substance." You cannot ask what would substance look like to you? That answer would have shaped the curriculum.
2. No falsification. Every substitute was a source the founder chose to consult. There was no mechanism for a real person to say "I wouldn't use this, and here's why" before 24 days of building.
3. No pricing conversation. Nobody was ever asked what they pay for self-improvement, or what would make $49.99/year feel reasonable versus expensive.
4. No workaround discovery. The single highest-signal finding in interviews, what people have jury-rigged, was never collected. If the target user is already keeping a physical journal, or following a specific YouTube channel, or has a Notion template, that's the real incumbent. It was never checked.

The traceable consequence. The onboarding funnel was designed, built, and then substantially rebuilt after a competitor teardown on day 22 (Chapter 03). Ten user interviews in week one would plausibly have surfaced the same psychological requirements, investment, commitment, personalised loss framing, before the first version was built rather than after.

🚩 The unresolved risk. SOLIS launched to nobody, so the core assumption, that this audience will pay $49.99/year for a structured course rather than a journal, remains completely untested. Interviews wouldn't have proven it either (only money does that, see Chapter 05), but they would have produced early falsification. Fifteen conversations, ten hours, zero dollars.

In fairness: for a solo founder building fast in a category they personally inhabit, review mining plus a competitor teardown is a defensible compressed substitute. It is not equivalent, and the honest framing is that it trades certainty for speed. If your build is longer than a few weeks, or you are not your own customer, the trade stops being defensible.

Lessons

1. Ask about the past. The future is fiction. People cannot predict their own behaviour.
2. The interview is going well when you're barely talking.
3. Money spent and workarounds built are the only strong signals. Enthusiasm is free.
4. Capture verbatim language, it becomes your copy.
5. In B2B, find the buyer, not just the sufferer.
6. Being your own customer is an advantage, not a substitute. You are one unusual data point.
7. Review mining is the best cheap substitute, real words, real complaints, but you can't ask a follow-up, and that's most of the value.
8. Interviews are for falsification, not validation. If you can't imagine an interview changing your mind, you're not running one.
9. Ten hours and zero dollars is the entire cost. Skipping it is almost never a budget decision.

Next: Chapter 05: Validating Demand Before You Build →