Filtering False Signals Out of Customer Interviews — Ask About Actions, Not Compliments

#Why Customer Interviews Invite False Signals

It's not that people are lying to you. People want to make the person across the table feel good, and they're chronically optimistic about their own future behavior. So when they're sitting across from the person who built the thing, what comes out is encouragement, not honest assessment.

The moment you mistake that encouragement for market validation, you start running for months toward demand that doesn't exist. False signals usually show up in three forms. Knowing them in advance lets you catch yourself mid-interview and think, "that's a compliment, not data."

#Three Principles for Digging Out Facts Instead of Compliments

The interview method commonly known as "The Mom Test" comes down to something simple: don't ask people to evaluate your idea — ask about their life and their past behavior instead. In practice, that breaks into three working principles.

1. Ask about past facts, not future opinions — Instead of "Do you think you'd use this?" ask "How are you handling this right now?" People are bad at predicting their own future but reasonably accurate about what they did yesterday.
2. Ask about a specific yesterday, not generalities — "How do you usually handle it?" invites a flattering average. "When was the last time this problem actually hit you? What happened that day?" pulls out a concrete instance.
3. Talk about their life, not your idea — The moment you explain your solution, the other person starts complimenting it and forgets the real problem. Seal your idea away for the entire interview, no matter how tempting it is to bring up.

Hold to these three and the center of gravity in the conversation shifts from "my idea" to "their reality." That's when you start collecting data instead of compliments. If you finish an interview without ever mentioning your product, that's usually a sign you did it right.

#Good Questions vs. Bad Questions

The same topic can come back as a compliment or a fact depending entirely on how you phrase the question. Bad questions are almost always future-tense, hypothetical, or agreement-seeking. Good questions are past-tense, specific, and behavior-tracking.

What the questions on the right have in common is that they ask about something that already happened. If someone has spent money or time in the past working around this problem, that's pain proven by action, not words. On the other hand, if the answer is "no, never," that tells you this isn't a real problem for that person yet. Either answer is useful information.

#Who to Talk To, How Many, and How

At an early stage, a statistically representative sample isn't the goal. The goal is talking to people until a pattern shows up. Usually somewhere between single digits and a dozen or so conversations, the same complaints and the same workarounds start repeating. If three people in a row say the same thing, that's hard to write off as coincidence.

• How many: use 5–15 people per customer segment as a baseline. If answers start converging, that's enough; if every conversation surfaces something new, keep going.
• Who: avoid family, friends, and people who know you. They're the most generous graders you'll ever get. Find strangers who actually experience the problem.
• Where: go where your target customers already gather — communities, open chat groups, industry events, relevant stores. Online, recruit through comments or DMs.
• How to recruit: don't lead with "let me show you my product." Lead with "I want to learn about a problem in this space for 15 minutes." You're there to learn, not to sell.
• How to record: ideally bring two people — one asks questions, the other writes down exactly what's said. Don't summarize. Capture direct quotes.

After the interview, sort what you heard into "real signal" and "noise." One test decides which: did action or resources follow the words, or just talk?

#How to Debrief and a Self-Check Checklist

The value of an interview is decided in how you debrief right after it ends. Memory gets flattering with time, so debrief the same day if you can. The core discipline: preserve what they actually said, not your interpretation of it.

1. Record statements as direct quotes — "I spend two hours every week cleaning this up in Excel," verbatim. A summary like "customer dislikes wasting time" strips out information.
2. Flag behavioral evidence next to each interviewee — check whether they spent money, spent time, or made an introduction or follow-up commitment.
3. Once you have 5+ interviews, cluster the repeated phrases — count how many times the same complaint or the same workaround came up. Frequency is a proxy for the size of the pain.
4. Deliberately look for evidence that your hypothesis is wrong — collecting only confirming evidence is confirmation bias. Keep a separate pile of reasons from the people who said no.
5. Update your next round of questions — fold in whatever new clue came up this time so the next interview digs deeper.

Below is a self-check to see whether you actually ran a real interview. If you answer "no" to most of these, you ran a sales pitch, not an interview.

• Did you ask only about their problem, without explaining your own solution or product?
• Did you get answers in the past tense ("I did") rather than the future tense ("I probably would")?
• Did you talk to strangers who actually experience the problem, rather than family or friends?
• Did you record statements as direct quotes rather than summaries?
• Did you separately flag responses backed by behavioral evidence — money, time, introductions?
• Did you also collect evidence that your hypothesis might be wrong?

Customer interviews ultimately determine what goes into the "problem definition" and "market demand" sections of your business plan. Who feels what pain, and whether you're the right person to solve it (Founder-Problem Fit), is a topic for a separate piece. The stronger the behavioral evidence you gather here, the more those sections read as "evidence" rather than "assumption."

#Frequently Asked Questions (FAQ)

Q. What's the minimum number of interviews to consider it enough?

A. Convergence matters more than a fixed number. Typically, once you've talked to 5–15 people in a customer segment, the same complaints and workarounds start repeating. If a new interview stops surfacing anything new, you've heard enough from that segment. If every conversation gives you a different answer, that's a sign your target is still too blurry.

Q. If I never show my solution, when do I actually get feedback on my product?

A. Split problem-validation interviews and solution feedback into separate stages. First confirm the "real problem" through interviews, then show a prototype or mockup in a later stage and watch the reaction. Mix the two and problem validation gets buried under compliments for the solution — you lose both.

Q. The other person keeps complimenting me. How do I redirect?

A. Acknowledge the compliment, then immediately steer back to a past fact. Something like "Thanks — when was the last time you actually ran into that problem yourself?" pulls the conversation back down to earth with a concrete yesterday. If no example comes up even then, this isn't a real problem for that person yet.

Q. Isn't "I'd buy it if the price is right" a positive signal?

A. A conditional promise is words with no action behind it. A real signal is evidence that someone has already spent money or time on this problem. If you want to test pricing, look at behavior, not opinion — get them to make even a small commitment, like a pre-order, prepayment, or letter of intent. That's far more accurate than what they say.