Ask about past behavior, not opinions of your idea

1

The Mom Test filter

A question is useful only if it would still draw an honest answer from your own mother, who loves you and will protect your feelings.

Fitzpatrick starts from the distortion itself. Ask your mother whether your idea is good and she lies to encourage you, because the real question is whether she supports you. The same pull appears with strangers, investors, and friends: most people are polite and want to help rather than tell hard truth.

The filter is mechanical. A good question stays specific to the past, concrete, and about the other person's life, never about your idea. Questions that invite judgment of the idea fail. Questions that force a report of lived behavior pass, because even a loving mother has to answer with facts.

Everything later in the book is a way of engineering conversations that survive this filter. The point is not that people are malicious. Human politeness and optimism about the future systematically corrupt opinion, idea, and hypothetical-purchase questions, so the only defense is to stop asking them and redirect every talk toward documented past.

In practice: An app pitch framed as whether someone would use a meal-planning tool draws a polite yes; asking them to recount the last time they planned meals for the week forces a real behavioral answer.

2

Opinions are worthless

Customers cannot reliably predict their own future behavior, so any answer about what they would do is noise, not signal.

Building on the filter, Fitzpatrick sorts customer talk into three piles: opinions, feelings, and facts. Opinions and compliments sit at the bottom because they cost the speaker nothing and commit them to nothing. Hypothetical future behavior sits barely higher; people are bad at forecasting their own actions.

Facts remain. What the person does now, what they have already paid for, what workaround they built, what they complained about last week. The founder’s job becomes an anthropological dig for evidence that someone has already spent money or serious time on the problem. If not, enthusiasm is cheap and predicts nothing.

Compliments flip from green light to red flag. A genuine signal looks like a customer asking to pay, introducing you to their boss, or giving a real deadline—not saying nice things and leaving. Resist the pull toward opinion-gathering and mine only for costlier signals that correlate with action when money is on the table.

In practice: A budgeting-app founder hears a dozen people say they would love something like that, takes it as validation, then launches to silence because none had ever tracked a budget or paid for a budgeting tool.

3

Past tense, not future

Replacing every future-tense question with a past-tense one is the single most reliable technique for getting honest data.

Fitzpatrick turns the opinion-versus-fact split into a rewriting drill. Founders’ scripted questions are almost all future-tense hypotheticals in disguise: would you use, do you think you’d pay, how much would this be worth. Each invites imagination, and imagination is generous, so answers inflate.

The fix is mechanical. For every question, ask what the person did the last time the problem came up. The past-tense version cannot be faked with enthusiasm because it demands a specific memory. If the person cannot produce one, that absence is itself data: the pain may not have been real enough to generate any behavior yet.

A subtler trap is the leading prompt that smuggles future framing, the wouldn’t-it-be-great-if move. The founder must stay aggressively neutral, letting silence and specific memory work rather than handing the customer a flattering hypothetical to agree with. That past-tense discipline makes the rest of the interviewing advice operational.

In practice: A founder building freelance contract-review tools kept hearing yes-I’d-use-that until switching to the last actual contract dispute; most respondents had never had one, so the pain was theoretical.

4

Money, time, reputation

An interest that has not cost the customer anything is not yet validated interest, so push every talk toward what they have already spent.

Fitzpatrick raises the bar with commitment and advancement. A good conversation should end with the person committing something real; otherwise the meeting is a failure no matter how pleasant it felt.

Commitment takes three tracked forms: money already spent solving the problem, time already invested building a workaround or researching alternatives, and reputational risk such as introducing you to a boss or agreeing to a follow-up with the team present. Ask about specific past expenditures—current tools, internal builds, abandoned fixes.

End every meeting by requesting a small next step: an introduction, a look at their current process, five more minutes of effort. Willingness to spend even that is a cleaner signal than verbal praise. Watch calendars and wallets, not mouths.

In practice: A manager calling enterprise scheduling a huge pain teaches little; discovering the team keeps a shared spreadsheet with three years of manual patches proves the pain was real enough to resource without being asked.

5

Interview before you pitch

Describing your product before gathering facts contaminates the conversation and turns every answer into a reaction to you.

Founders pitch early because pitching feels productive and interviewing feels aimless. Pitching is exactly what triggers the politeness reflex. Once the customer knows what you are building, answers bend toward supporting or gently rejecting that idea instead of describing their actual problem, and the un-primed situation disappears.

The prescribed structure delays the pitch. Open on goals and current process, dig into specific past incidents and workarounds, and only pitch—if at all—at the very end, treated as optional and separate from fact-finding. The founder’s early job is closer to a journalist’s than a salesperson’s: gather the story first, sell later, and never let your solution narrative shape their problem account.

Group meetings add another distortion. In rooms with multiple stakeholders the most senior or vocal opinion often drowns the person who does the daily work and holds the real facts. Seek one-on-ones with the actual process owner rather than treating a meeting with the decision maker as enough.

In practice: A scheduling-tool team withheld their pitch and learned the real bottleneck was approvals from another department—a fact customers would never have surfaced if the meeting had been spent describing the scheduling feature.

6

Narrow the segment

Vague target markets produce vague, unusable data, so learning requires a segment specific enough that its members share the same behavior and pain.

Even clean individual interviews fail when the market is drawn too wide. Facts refuse to form a coherent pattern. Interview restaurant owners and a food-truck operator and a hundred-seat dining room share almost nothing operationally, so the data contradicts itself.

The fix is progressive narrowing. Use early conversations to discover which sub-group actually shares the pain, then deliberately over-sample that sub-group. Chase specificity until each new interview either confirms or breaks a clear pattern, and the founder learns faster whether a real common problem exists.

Pair the narrowing with an explicit list of falsifiable assumptions before each round. Check off which beliefs survived contact with customers instead of walking in with only a vague hope to learn something.

In practice: A team that thought freelance designers were underpaid from poor negotiation narrowed from freelancers, to designers, to designers under two years working with US clients before the same behavior—undercharging on the first three projects—showed up consistently enough to build around.

7

Systematize the learning

Individual great conversations do not build a business; only a deliberate system for logging, reviewing, and acting across many conversations does.

Founders who ask good questions still fail when they rely on impression and memory. Memory favors the most recent or most emotional meeting and quietly smuggles confirmation bias.

Write specific facts down immediately after each meeting. Keep what was said—a fact, a number, a concrete detail—on one line and what you concluded on a separate line so the two cannot blur. Maintain a running list of the riskiest assumptions, the ones that would kill the idea if false, and point the next conversations at those assumptions instead of open-ended chat.

Revisit commitment as a scoreboard. Track, meeting over meeting, whether the ratio of real commitments to polite compliments is improving. That ratio shows whether the idea is getting stronger or the founder is only getting better at hearing what they want. Customer conversations compound only when treated as structured research with explicit hypotheses, not isolated pleasant chats.

In practice: One founder, after every third meeting, updated a simple document listing assumptions tested, what was learned, and what would be tested next, turning scattered anecdotes into cumulative evidence.