Customer interviews
Mom Test Interview Script — Ask About Their Life, Not Your Idea
Rob Fitzpatrick's The Mom Test (2013) is a rule set for conversations that survive politeness. This page is the operational script: what to ask, what not to ask, how to write notes, and how to decide after a handful of talks — before you spend a month building.
Last updated: September 25, 2026
Direct answer
What is the Mom Test?
The Mom Test is a set of interview rules from Rob Fitzpatrick (2013). Talk about the other person's life, not your idea. Ask about the recent past, not a hypothetical future. If your mom would still say “that's wonderful, honey,” the question failed — you learned nothing you can decide on.
Key takeaways
What to remember
- Compliments are not data. “I'd use that” is politeness until you see a recent incident, a workaround, or a real commitment.
- Ask what they did last time, not whether they would. Memory beats prediction.
- One interview is a story. Five to ten matching stories in the same segment is a pattern — still not a market.
- Write their words, not your interpretation. The quote is the evidence; your summary is a guess.
- Decide the kill rule before you book the first call. Otherwise every weak talk becomes “we need one more.”
The rules
Three rules that keep interviews honest
Fitzpatrick wrote The Mom Test after watching founders (including himself) collect compliments and call it validation. The book is short on theory and long on failure modes. The three rules below are the ones that change what you write down after a call. Steve Blank's customer-development work (The Four Steps to the Epiphany, 2005) is the older lineage: get out of the building and talk to people who have the job, not people who like you.
- 1
Talk about their life, not your idea
The moment you pitch, the conversation becomes a courtesy evaluation. They start protecting your feelings. Keep the product in your pocket until you have a specific story from last week.
- 2
Ask about the past, not the future
People are bad at predicting their own behavior. “Would you use this?” produces a yes that costs them nothing. “Walk me through the last time this broke” produces a date, a workaround, and often a receipt.
- 3
Listen more than you talk
If you are explaining, you are not learning. Follow the last concrete detail they gave you. Silence after a short answer is a tool, not an awkward gap.
Question design
Bad questions vs questions that survive politeness
A leading question hands the other person the answer you want. A Mom Test question asks for a recent scene. Use the right-hand column this week.
Don't ask
Would you use an app that plans your team's standups automatically?
Ask instead
Walk me through the last time a standup or status update went badly. What did you do next?
Don't ask
How much would you pay for something like this?
Ask instead
The last time you tried to fix this, what did you spend — time, money, or someone else's hours?
Don't ask
Don't you hate it when reports arrive a week late?
Ask instead
When a report is late, what actually happens? Who notices, and what do they do?
Don't ask
Is this a big problem for people like you?
Ask instead
When did this last cost you a day, a deal, or a customer? What did that cost look like?
Don't ask
If I built this, would you be my first customer?
Ask instead
What would have to be true for you to change how you handle this next Tuesday?
Don't ask
What features would you want?
Ask instead
What have you already tried, and which parts did you keep?
Copy-ready script
A 12-question interview you can run this week
Use this as a spine, not a questionnaire. Skip anything they already answered. Stay on the last incident until you can retell it without inventing details. Questions 13–15 are optional if they still have time.
- 1
What are you responsible for when [this job] has to get done?
Opens on their world. You learn the role, the constraint, and the words they use — before any product language leaks in.
- 2
Tell me about the last time [this area] went badly.
Forces a specific incident. If they cannot name a recent one, the pain may be theoretical.
- 3
What did you do next?
Behavior, not opinion. The workaround is often the real competitor.
- 4
How long did that take, and what did it cost?
Time and money are the first willingness-to-pay clues. Vague “it was annoying” is weaker than a Tuesday and a number.
- 5
Who else got pulled in?
Maps the buying committee or the household. A solo pain and a committee pain need different tests later.
- 6
What have you already tried?
Lists the failed tools, the spreadsheet, the intern. People who have tried nothing may not feel the problem enough to pay.
- 7
What did you keep, and what did you drop?
Shows switching cost and which jobs are load-bearing. Keep asking until you know why the current stack survived.
- 8
If this happened again next Tuesday, what would you do?
Still about their process, not your pitch. Reveals whether they have a default response or still scramble.
- 9
Where do you go when you need help with this?
Names channels you can actually reach later — a Slack, a subreddit, a trade newsletter — without asking “where should I advertise?”
- 10
What would have to be true for you to change how you handle this?
Surfaces the real barrier: trust, data lock-in, a boss, a habit. This is not a feature request list.
- 11
Is there someone else I should talk to who deals with this more often than you?
Warm introductions beat cold outreach. Also tells you whether you are talking to the person who actually lives the job.
- 12
What did I not ask that I should have?
They will often name the constraint you were too close to see. Write it down verbatim.
Three optional questions if they still have time
Use these when the incident is clear and you want artifacts or a purchase story. Do not force them after a tired “I guess.”
- 13
Can you show me the spreadsheet, inbox, or tool you used last time?
Artifacts beat memory. A shared screen often contradicts the polished version of the story.
- 14
When you last paid for a workaround, what triggered the purchase?
Purchase triggers are more useful than list prices. You are looking for a moment, not a number they invented for you.
- 15
What would make you ignore a new tool in this space?
Names the “not now” reasons — security review, another rollout, no owner. Those reasons kill more ideas than competitors do.
After the call
How to take notes you can decide from
Write during the call or in the ten minutes after, not the next morning. Teresa Torres (Continuous Discovery Habits, 2021) teaches interview notes as opportunity snapshots — a quote plus the context — rather than a transcript dump. For a first-time founder, three columns are enough: facts, exact quotes, and your guesses. Never mix the three.
- Date, channel you found them, and the role they actually have (not the role you hoped for).
- The last incident: when it happened, what broke, what they did next.
- Time or money they named. If they named none, write “none named” — do not upgrade “annoying” into a cost.
- Current workaround and which parts they kept.
- Exact phrases. If you cannot quote them, you do not have the note yet.
- Any commitment they offered unprompted (intro, artifact, deposit). If they offered none, write that too.
Three columns, one page
Facts = dates, tools, amounts. Quotes = their words in quotation marks. Guesses = “I think they would pay” and every other inference. If a later interview contradicts a guess, the guess dies; the quote stays.
After 5–10 talks
How to decide after five to ten conversations
Five to ten talks in one segment is a first-round sample, not a census. Customer discovery on this site treats that band as enough to see a pattern or the absence of one. The empirical saturation literature (Guest, Bunce & Johnson, 2006) puts thematic saturation nearer twelve interviews in a single segment — use that as a reason to keep going when stories still diverge, not as a trophy number. Write the decision in one sentence.
Continue — the stories rhyme
Several people describe a recent incident in similar words, already spend time or money on a workaround, and can name what would have to change. Next: a smoke test or a pricing conversation with the same segment — not a six-month build.
Reframe — you found a different job
They have pain, but it is not the pain on your landing page. Rewrite the ICP or the outcome. Do not “add a feature” to keep the original pitch alive.
Stop or park — no recent incidents
You cannot find five people who lived this last month, or every story is vague and compliment-shaped. That is evidence. Do not book twenty more friends to rescue the idea.
Common mistakes
How founders ruin a good script
Pitching in the first five minutes
Once they hear the idea, they evaluate you. You will get “that's cool” and lose the incident. Keep the demo for a later call, after you have notes.
Interviewing people who like you
Friends, family, and coworkers will protect you. Prefer strangers who match the job. Warm intros are fine; loyalty is not a segment.
Counting compliments as demand
“I'd definitely use that” is not a signup, a deposit, or a calendar hold. Fitzpatrick's rule is blunt: compliments are not data.
Asking leading questions
“Don't you think it would be easier if…?” hands them the answer. If you hear yourself selling the frame, stop and ask what happened last time.
Stopping after two encouraging calls
Two stories can rhyme by luck. Five to ten in the same segment is the first honest look. If the first two contradict, you need more, not a pivot tweet.
Never writing a kill rule
If you did not decide what would make you stop, every weak conversation becomes “qualitative insight.” Write the fail line on the same page as the script.
Sources
Where these rules come from
- Rob Fitzpatrick, The Mom Test (2013) — The interview rules, the compliment failure mode, and the “ask about their life” discipline this script follows.
- Steve Blank, customer development — The older instruction to leave the building and talk to people who have the job — The Four Steps to the Epiphany (2005).
- Teresa Torres, Continuous Discovery Habits (2021) — Interview notes as opportunity snapshots: a quote plus context, not a transcript you never reread.
With Yibud
Use the report to pick who to interview
Yibud is a free, no-signup startup idea validator. Scores come from a deterministic rule engine — not from a language model. Optional AI only polishes prose. If the report's weakest dimension is the customer or the assumption you cannot yet name, run this script next, then bring the notes back into the same idea.
Analyze my idea →FAQ
Questions founders actually ask
What is the Mom Test in one sentence?
Ask people about their recent life and behavior — not whether they like your idea — so politeness cannot fake a yes.
Can I mention my idea at all?
Not in the first pass. Mention it only after you can retell their last incident without guessing. A late, short mention is fine if you need to test whether they recognize the job; a pitch deck in minute two is not.
How many interviews do I need?
Plan five to ten in one segment for a first round. If stories still contradict, keep going toward the saturation range discussed on the customer-discovery page — do not average strangers from three different jobs into one “insight.”
Should I pay people for interviews?
A small thank-you is fine when you are asking for 25 minutes of a stranger's calendar. Payment does not make the answers true. What makes them true is a recent incident they can walk through.
Zoom or in person?
Whichever gets you the incident. Video is enough if they can show a screen. In person helps when the job is physical. A survey form is not a substitute for this script.
What if they compliment me?
Write “compliment” in the guesses column and return to the last incident. “Thanks — when did this last go wrong?” is a complete reply.
How is this different from a survey?
A survey asks people to rate statements you wrote. This script asks them to narrate a Tuesday. You cannot follow a surprising sentence on a form.
How does this connect to Yibud?
Run the analyzer first if you do not yet know which assumption is most expensive. The report will not interview anyone for you. It will tell you which risk to take into these conversations.
Know which assumption to take into the next call
Yibud scores the weak dimensions with a deterministic engine, then you pick the experiment — starting with conversations like these.
Generate a Yibud report →