── ── Customer discovery
How to Talk to Customers Without Getting Lied To (The Mom Test)
July 25, 2026 · 5 min read · By Brad Ju
The Mom Test is a rule for customer interviews that gets you the truth: talk about the customer's life, never about your idea. Ask about past behavior, not future intentions, and treat time, reputation, and money already spent as the only evidence that a problem is real.
Most customer interviews are useless — not because founders skip them, but because they run them in a way that guarantees a false positive. You describe your idea. The person is polite. They say "that sounds great, I'd totally use that." You walk away encouraged, build for months, and no one shows up.
The fix is a discipline Rob Fitzpatrick called The Mom Test: ask questions so grounded in the other person's actual life that even your mom couldn't give you a misleading answer.
The one rule: talk about their life, not your idea
The moment you pitch, you've contaminated the data — people compliment ideas to be nice. So don't mention the idea at all. Ask what they're trying to get done and how they do it today.
Which questions get the truth?
| Sounds reasonable | Ask this instead | Why it works |
|---|---|---|
| "Would you use something that does X?" | "When did you last hit this problem? Walk me through it." | Past behavior is data; future intentions are fiction. Everyone says yes to a hypothetical. |
| "Do you think this is a good idea?" | "How are you solving this today?" | Workarounds people already run are proof of pain. Opinions are free. |
| "Would you pay for this?" | "What does this cost you now — in money, time, or hacks?" | Real pain shows up as spend that already exists, not spend that's promised. |
Behind every good question are the three currencies of commitment: time, reputation, and money. "It's annoying" is an opinion. "I built a spreadsheet and pay an assistant three hours a week to maintain it" is a fact.
And anchor every vague claim. "I always struggle with this" becomes "how many times last month?" "I'd pay a lot for this" becomes "what do you pay now?"
How do you know the interview failed?
You left feeling good. That usually means you fished for compliments and got them.
A good interview often feels uncomfortable — you heard that the problem isn't as sharp as you hoped, or that they've never spent a dollar on it. That discomfort is the most valuable thing you'll get all week.
End with a real advance, not a compliment: an intro, a pre-order, a pilot, a follow-up meeting with their data on the table. If they won't spend anything — even time — the enthusiasm was noise.
Why it matters more than it looks
Getting this wrong is one of the quiet reasons startups die: months of building on validation that was never real. Getting it right feeds everything downstream — pricing, positioning, and which go-to-market motion to sequence first.
Run it as a process, not a memory
The Mom Test is one of the mental models we ship as an executable, open-source agent skill — MIT-licensed, free: github.com/deciqAI/knowledge-skills. But you don't need an agent to start. You need to stop pitching and start listening.
FAQ
What is the Mom Test in customer discovery?
A rule from Rob Fitzpatrick's book of the same name: ask questions so grounded in the customer's real life that even your mom couldn't mislead you. Talk about their life and past behavior — never pitch your idea.
Why is "Would you use this?" a bad interview question?
Because everyone says yes to be polite. Future intentions are fiction; only past behavior — what they actually did, spent, or built the last time the problem occurred — counts as evidence.
What counts as real validation in a customer interview?
A commitment in one of three currencies: time (a scheduled working session), reputation (an intro to their boss or team), or money (a pre-order or pilot). Compliments count for nothing.
How many customer interviews are enough?
Keep going until the answers repeat — for a focused segment that's often 10–20 conversations. Count facts, not interviews: real workarounds, real spend, real advances.