Rob Fitzpatrick's The Mom Test is the most useful hundred pages a first-time founder can read. The book's central insight: most customer interviews lie, and they lie because of how the founder asks, not because the customer is dishonest. People are bad predictors of their own future behavior and excellent at being polite.
What the book leaves underspecified — by design — is the exact script. Founders read it, get the principle, sit down for the first interview, and freeze. The translation below is the script we use to run validation interviews inside MoatKit, distilled into five questions, five anti-patterns, and an interview structure.
The principle, restated
The Mom Test says: don't ask people about their opinions of your idea. Ask them about their past behavior with the problem.
"Would you use a tool that does X?" is a question your mom would say yes to. "Tell me about the last time you faced X — what did you actually do?" is a question that pulls real information, even from your mom, because it's about a past event, not a hypothetical product.
The five questions
1. Tell me about the last time you faced this problem.
Past-tense, specific. If they can't recall a specific time, the problem isn't acute enough — and that's information. If they can, you'll get the context: how often it happens, what triggered it, who else was involved, how it ended.
Listen for: time of day, named tools they reached for, named people they asked, specific outcomes. Vagueness here means low salience.
2. What did you do about it?
The action they took is your real competitor. The buyer's current workaround — even if it's "I just live with it" — is what your product has to beat. If they spent two hours on a spreadsheet, your product is competing with two hours and a spreadsheet, not with their idea of "a better tool."
Listen for: a process, not a product. Most founders are surprised to discover the workaround is rarely a competing tool — it's a manual sequence the buyer already accepts.
3. What would have made that better?
Pull on the workaround. Where did it break down? What did they wish they could skip? What did they spend the most time on? This is where the buyer self-describes your actual product spec, not the imagined one you've been holding.
Listen for: time savings (most common), confidence (second most common), and audit-ability (B2B). Not "ease of use" — that's a non-answer.
4. How much would not solving this cost you next year?
Cost framing in actual currency. Hours per month × hourly rate is fine. Customers lost × average customer value is fine. The point isn't accounting precision — it's forcing them to put a number on it, because the act of doing so reveals whether the pain is real.
If they can't put a number on it within thirty seconds, the problem isn't urgent enough to drive a purchase. If they can, you have the upper bound on what they'd pay to solve it (rule of thumb: they'll pay up to 10–20% of the annual cost they just named).
5. Who else in your company / household cares about this?
Especially for B2B. Most purchases involve more than one person. Identifying the second stakeholder early tells you who you'll need to convince, what their pain is, and whether the buyer in front of you has actual authority to spend.
Listen for: budget holder (often different from problem-feeler), end-user (often different from buyer), and skeptic (the one who'll push back on the purchase).
The five anti-patterns that ruin most interviews
Leading questions
"Don't you think it'd be easier if X?" Of course they'll agree. You've framed it as the only option. Replace with: "When you were doing X, what was the hardest part?"
Future-tense questions
"Would you pay for this?" "Would you use this?" Both produce optimistic lies. Replace with: "Have you ever paid for something similar?" "Tell me about the last tool you actually paid for."
Opinion questions
"What do you think of this idea?" The interviewer is too polite to say "I'd never use it." Replace with: "Walk me through what you'd do tomorrow morning if you faced this problem."
Compliment-fishing questions
"Do you like the concept?" Replace with: silence after their last answer. Wait. The discomfort makes them say something real.
Vague-pain questions
"Is this annoying?" "Is this hard?" Yes-or-no questions about feelings produce useless data. Replace with: "How often does this come up in a typical week?" "What happens when you try to deal with it?"
How to structure the interview itself
Length: 15–30 minutes
Long enough to get past pleasantries, short enough that the interviewee doesn't feel they've been kidnapped. Schedule 25 minutes. Most useful information surfaces in minute 12.
The first 5 minutes: context, not product
Don't describe your product. Don't pitch. Ask about their role, their company, what they're working on this week. You're building rapport and gathering the context you'll use to interpret their later answers.
The next 15 minutes: the five questions
In order. Don't rush. Use silence aggressively — most of what you'll learn comes in the second sentence of an answer, after they've said the polite thing and then started saying the real thing.
The last 5 minutes: ask for next steps, not feedback
Don't ask "What did you think?" at the end — you already know what to think. Ask: "If we built something for this, who else should I talk to?" This filters: people who are genuinely interested introduce you to two other people. People who were being polite say "I'll have a think and let you know."
What to do with the answers
Three patterns to watch for across interviews:
- Three buyers naming the same current workaround → your real competitor is named. Build against it, not against your imagined competitor.
- Three buyers naming the same cost number (within an order of magnitude) → you have a defensible price ceiling. Charge a fraction of it.
- Three buyers naming the same second stakeholder → you have a second interview list, ranked by who's hardest to convince.
Three is the magic number. One interview is an anecdote. Three is the start of a pattern. Ten is a defensible case.
How MoatKit runs the Mom Test playbook
The Mom Test is implemented as a playbook inside the app. You provide three inputs — a customer segment, the current workaround, and the riskiest assumption you're making — and the playbook generates a seven-question interview script tuned to that specific persona, plus a task to book the first three calls and a journal entry template per interview.
Each interview lands as a structured journal entry: signal (what surprised you), workaround (what they actually do), commitment (what they'd pay, in their words), next step. After ten interviews, the journal feeds back into the validation checklist so you can see, across all conversations, whether you have a real signal or three polite ones.
The book is a hundred pages. The script is five questions. The hardest part is the second sentence — the one after the polite answer, when you have to stay quiet long enough that the real answer comes out.