Customer discovery interview questions that ask about the past

By Meet Patel · 2026-10-03 · 6 min read

Take a founder building inventory software for independent restaurants. She interviews ten owners and asks each one: "Would you use an app that tracks your stock automatically?" Ten of ten say yes, three offer to be beta testers, and one says it sounds like exactly what the industry needs. She leaves with a strong signal and no information. Every one of those answers cost the owner nothing to give, and the polite one is always yes.

Rob Fitzpatrick's book The Mom Test is built around this failure. Its premise is that people will tell even your mother that your business idea is good, so the job is to ask questions that a polite answer cannot satisfy. This post collects the question pairs, an interview script you can use this week, and a way to score your notes afterward so that the result reflects what people said and did.

Why the polite answer is worthless

The summaries of the book sort unreliable answers into three kinds. Compliments are vague praise given to spare your feelings. Fluff covers generic claims ("I always do this"), future promises ("I would definitely buy that") and hypotheticals. Ideas are feature requests that come without any account of the underlying problem. A review at ybrikman.com lists all three, and the book's central observation is that customers mislead you by wanting to please you, with no ill intent.

The mechanism is cost. A prediction about the future is free to make and cannot be checked. A description of what someone did last Tuesday is expensive to invent, easy to probe, and usually true. The best interview questions shift the conversation toward the second kind.

Three rules

The book states three rules, and the same wording appears in several independent summaries, including a book report at mtlynch.io:

1. Talk about their life instead of your idea.
2. Ask about specifics in the past instead of generics or opinions about the future.
3. Talk less and listen more.

Rule 1 means your idea stays in your pocket for most of the interview. Once you describe it, the other person starts judging it, and you have an opinion where you wanted a fact. Rule 2 does the main work in the question pairs below. Rule 3 is a discipline of silence: after an answer, wait a few seconds before the next question, because the second sentence is often the useful one.

Question pairs: weak and better

The first three weak questions come from the book as summarized in the ybrikman.com review, along with the alternatives it suggests. The restaurant examples are my own adaptations.

• Weak: "Do you think my idea for a product that does X is any good?" Better: "How do you currently do X?" The first invites an opinion about you. The second invites a description of their week.
• Weak: "Would you buy a product that did X?" Better: "What products do you use for X, and what problems are you hitting?" Willingness to buy in the abstract is cheap, and current spending is evidence.
• Weak: "How much would you pay for a product that did X?" Better: "How much time and money is that costing you now? Have you paid for a solution before?" Price follows from an existing cost.
• Weak (restaurant): "Is stock counting a pain for you?" Better: "Talk me through the last time you counted stock. What did you do, in what order?" The book's own good questions include "Talk me through last time X happened" and "What else have you tried for X?"
• Weak (restaurant): "Would an automatic alert when you run low on an item be useful?" Better: "When did you last run out of something mid-service, and what did that cost you?" A hypothetical benefit becomes a dated event with a number attached.

A 30-minute interview script

Use this structure, and write the questions on one page before each call.

1. Minutes 0 to 5, context. Ask what they do, how long they have done it, and what a normal week looks like. Say you are researching the area and have nothing to sell.
2. Minutes 5 to 15, the last time. Choose the problem area and ask them to walk you through the last time it happened. Follow up with "why do you bother with that?" and "what are the implications of that?", two of the book's own questions. Stop at each number: how long, how often, how many people.
3. Minutes 15 to 22, what they have tried. Ask how they are dealing with it now and what else they have tried. Anything they built, bought or abandoned is a fact about their priorities.
4. Minutes 22 to 26, money and authority. Ask where the budget for this comes from and who else is involved in a decision. Pricing is a product decision, so these answers shape the product as well as the sale.
5. Minutes 26 to 30, close. Ask "Who else should I talk to?" and "Is there anything else I should have asked?", both from the book's list. Then ask for a commitment (below).

Ask for a commitment before you leave

The book separates the useful outcome of a meeting from a pleasant one by whether the other person gave something up. The summaries name three currencies: time (a follow-up meeting or a trial), reputation (an introduction or a testimonial) and money (a letter of intent, pre-order or deposit). The ybrikman.com review gives examples of each outcome. Weak endings sound like "Looks great, let me know when it launches" and "I would definitely buy that." Strong endings sound like "When can we start the trial?" and "Can you meet with my team?"

Count them. If ten interviews produce ten compliments and no commitments, you have learned that the idea is likeable. If three produce a booked follow-up and one pays a deposit, you have a start. For how those commitments turn into paying customers, see how to get your first 10 customers.

Scoring your notes

After each interview, go through your notes and label every line with one of five tags: F for a specific past fact, G for a generic claim, P for a promise or hypothetical, C for a compliment, and I for an idea. Then keep three counts: the number of F lines, the number of commitments, and the number of times a number appeared (hours, dollars, frequency).

Here is a worked example for one hypothetical interview. A restaurant owner says her last stock count took two people three hours on a Sunday night (F, with a number), that she has tried a spreadsheet and abandoned it after a month (F), that "this would be great" (C), that she would "definitely pay for it" (P), and that it should connect to her supplier's ordering portal (I). That is two facts, two numbers, one idea worth probing with "why do you want that?", and no commitment until you ask for one.

My rule of thumb, with no validated threshold behind it, is to treat a pattern as real when at least half of your interviews contain the same specific fact, such as the same workaround or the same cost. Compliments and promises are not part of that count, however many there are.

Four mistakes to remove from the next round

• Describing your solution early. The moment you pitch, the interviewee switches to judging it.
• Interviewing friends and family. They have the strongest reason to be kind. Ask them for introductions to people who match your segment, and interview those people.
• Taking feature requests at face value. A request is a symptom. Ask "why do you want that?" and write down the answer instead of the feature.
• Mixing segments. Ten interviews across five kinds of customer give you five anecdotes. Choose one segment and interview it until the answers repeat (see also The Audience Trap).

The principle

An interview is a measurement, and its quality depends on what it costs the other person to answer. Questions about the past are expensive to answer falsely, and commitments are expensive to give casually. Build the conversation out of those two, and the notes become something you can count.

Written by Meet Patel — startup operator and growth strategist in Dubai.

Read on themeetpatel.com