How to Run User Interviews: A Founder’s Playbook
May 2, 2026
Most founders walk out of customer conversations convinced they’ve found product-market fit — when all they’ve found is a polite person who didn’t want to hurt their feelings.
User interviews are one of the highest-leverage activities in early-stage product development. Done well, 15–20 conversations can tell you more than six months of analytics. Done poorly, they generate false confidence that sends you deeper down the wrong path. The difference is not how many interviews you run — it’s how you ask.
This post covers:
- Why standard interview questions mislead founders
- The three Mom Test rules that produce honest signal
- A repeatable 45-minute interview structure
- A side-by-side table of good vs. bad questions
- A real example of synthesis done right
- What to run this week
The Problem With How Founders Ask Questions
Most founder-run interviews fail for the same reason: founders pitch with their questions. “How valuable would it be if our tool automated X?” is not research — it is a leading question dressed up as curiosity. The interviewee hears your excitement, wants to be supportive, and says “that would be really useful.” You write it down as validation.
This dynamic is so common that Rob Fitzpatrick named it in The Mom Test: if you ask your mom whether your idea is good, she’ll say yes. Not because it is, but because she loves you. Most users behave the same way — they want to be helpful, not honest. The fix is not to ask better questions about your idea. It’s to ask questions that don’t involve your idea at all.
At Decagrowth, we’ve seen founders spend months building features “validated” by interviews that were essentially pitching sessions. The pattern is always the same: enthusiastic calls, low activation once the product ships. The interviews felt like signal; the data said otherwise.
The Mom Test: Three Rules That Change How You Listen
Rule 1: Talk About Their Life, Not Your Idea
The most reliable interviews never mention the product until the final five minutes. You are trying to understand a person’s actual workflow, frustrations, and current behaviors — not validate a feature you already built.
Ask: “Walk me through the last time you had to [do the thing your product addresses].” Ask: “What’s the hardest part of that process today?” Ask: “How are you solving it right now?” Their answers reveal the real problem, its real frequency, and whether it is painful enough to change behavior. If they cannot name a recent example, the problem probably isn’t acute enough to build a business around.
Rule 2: Ask About Past Specifics, Not Future Hypotheticals
“Would you use a product that…” is the single most misleading question in a founder’s interview playbook. Future intentions are unreliable — people consistently overestimate what they will do, how much they will pay, and how readily they will switch tools.
Replace every future question with a past one. “How much did you spend last year solving this problem?” beats “How much would you pay for a solution?” “Have you ever paid for a tool that does part of this?” beats “Would you pay for a tool that does this?” Specific past behavior is one of the best proxies for future action you have access to.
Rule 3: Listen More Than You Talk
A useful target: speak less than 20% of the time. If you are filling silence with your own thoughts about the product, you are pitching, not researching. When a user says something surprising, follow it. “Tell me more about that” and “what did you do next?” are more valuable than any prepared question. The unexpected tangent is usually where the real insight lives.
Send two people if you can — one to lead, one to take notes. The lead stays in the conversation; the notetaker catches the details that slip past in the flow. Trying to do both roles at once means doing neither well.
How to Structure a 45-Minute Interview
A 45-minute window lets you go deep without burning the interviewee’s goodwill. Here is a repeatable frame:
- 0–5 min: Small talk and brief context-setting. Ask permission to record. Keep it casual — the first five minutes set the emotional register for everything that follows.
- 5–20 min: Open-ended exploration of their workflow and pain. No product mention. Ask them to walk you through a recent specific example.
- 20–35 min: Drill into specifics. What did they try? What failed? How much did the problem cost them in time or money? Push for numbers and concrete incidents.
- 35–40 min: Share the concept or show a minimal mockup, then go quiet. Watch what confuses them before you explain anything. Their first-contact reaction is data you cannot get a second time.
- 40–45 min: Ask for a referral. “Who else do you know who has this problem?” A warm introduction to the next interviewee is often the most valuable output of the entire call.
Good vs. Bad Questions
| Bad question | Why it fails | Better version |
|---|---|---|
| “Would you use a product that does X?” | Future hypothetical; invites politeness | “The last time you faced X, what did you do?” |
| “How much would you pay for this?” | Imaginary budget with no anchor | “How much did you spend solving this problem last year?” |
| “Don’t you think Y is a big problem?” | Leading; telegraphs your thesis | “What are the biggest friction points in your workflow?” |
| “What features would you want?” | Puts users in product manager mode | “What does your current tool get badly wrong?” |
| “Would this solve your problem?” | Asks them to evaluate your solution | “How are you solving this today, and what breaks?” |
A Real Example: How Superhuman Found Their Cohort
When Rahul Vohra was preparing Superhuman for growth, his team ran a single synthesis question after interviews: “How would you feel if you could no longer use Superhuman?” with three options: very disappointed, somewhat disappointed, not disappointed. The goal was to find the users who would genuinely miss the product — not the ones who said kind things on a call.
He found that users who answered “very disappointed” shared a tight profile: busy professionals who sent more than 50 emails a day and valued speed above all else. He then filtered every product and growth decision through that cohort. Users in the “somewhat disappointed” bucket were deprioritized — not because they didn’t matter, but because building for them would dilute the core experience for the right segment.
Superhuman reached 58% “very disappointed” before they opened up growth. Vohra’s stated threshold: 40% as the floor where a product is ready to grow. Below 40%, you are in discovery mode. Above it, you can start investing in distribution.
The lesson is not to copy the exact question. It’s that a single well-designed synthesis prompt, run consistently across 20 interviews, can surface the exact segment worth building for. Your interviews need a signal-extraction layer on top of them, not just raw notes filed somewhere. That extraction layer is what connects discovery work to the activation metric you actually ship against.
Synthesis: Turning 15 Conversations into a Decision
Raw notes are noise. Signal comes from pattern-matching across conversations. After each interview, write three things down immediately while the call is still fresh:
- The specific problem they named and how recently it occurred
- Their worst complaint about whatever they use today
- One exact phrase they used that you did not expect
After 15–20 interviews, cluster these notes by theme. Problems that surface in eight or more conversations are real. Language that appears in multiple transcripts verbatim belongs in your positioning and copy — because your users already speak it. Segments where urgency is highest are where you build first.
If you cannot find one dominant pain running through at least two-thirds of your interviews, you either have not found the right segment yet, or the problem is not acute enough to support a business. Both are valid findings. The founders who compound fastest treat a null result the same way they treat a positive one: they shift the segment, not the product, and go find the right conversations to have.
What to Do This Week
- Schedule five interviews — reach out directly to people you believe have the problem. Don’t recruit through a form. A personal ask converts far better than a survey link.
- Write a one-page interview guide with five or six open questions about their current workflow. No product mention until the last ten minutes.
- Record every session with permission. Assign one person to lead and one to take notes. Don’t try to do both.
- After each call, write three bullets before you do anything else: biggest pain named, current workaround, one phrase that surprised you.
- After five interviews, compare notes side by side. Repeated language and repeated problems are your first real signal. Isolated observations are not.
If you’re in the middle of customer discovery and want a peer to pressure-test what your interviews are showing, reach out to us at Decagrowth. We run this process inside our own products and with the founders we work alongside. Read more about how we work before deciding if the conversation makes sense.