Chapter 2: Customer Discovery — Finding Truth in Conversation

The Fundamental Problem: People Lie (Nicely)

Every product team eventually talks to customers. The problem is not getting the conversations — it is getting truth out of them.

People lie in customer interviews. Not maliciously. Not even consciously. They lie because human social wiring makes honesty in these contexts genuinely difficult ^{mom-test cdh}. When someone asks "Would you use this product?", the socially comfortable answer is "Yes, that sounds great." Saying "No, I don't care about this problem" feels rude, especially to someone who is clearly excited about their idea.

The result is that most customer conversations produce data that feels valuable but is actually noise. Teams walk out of interviews feeling validated, build the thing, and then watch nobody use it. The interviews did not fail because the customers were dishonest people. They failed because the questions made honesty almost impossible.

This chapter is about asking questions that make honesty easy.

The Mom Test

Rob Fitzpatrick's "Mom Test" is the most practical framework for honest customer conversations ^mom-test. The name comes from a simple standard: your questions should be so good that even your mother — who loves you and wants to support you — cannot lie to you.

The Mom Test has three rules:

Rule 1: Talk About Their Life, Not Your Idea

Bad: "Do you think this app for tracking expenses is a good idea?" Good: "How do you currently keep track of what you spend?"

The first question invites an opinion about your idea. The answer will be shaped by politeness, by the person's desire to seem supportive, and by whatever they imagine your app might do. The second question asks about their actual behavior. The answer is grounded in reality.

When you talk about your idea, you turn the conversation into a pitch. The customer becomes an audience, and audiences give applause, not data. When you talk about their life, you become an investigator, and they become the expert witness on their own experience.

Rule 2: Ask About Specifics in the Past, Not Generics or the Future

Bad: "How often would you use something like this?" Good: "When was the last time you tried to track an expense? Walk me through what happened."

Future-tense questions invite speculation. People are terrible at predicting their own behavior ^{cdh mom-test}. They overestimate how much they will exercise, underestimate how much they will spend, and confidently predict they will use products they never touch.

Past-tense questions about specific instances anchor the conversation in reality. The person cannot speculate about what happened last Tuesday — they either remember it or they do not. And the details of what actually happened reveal far more than any hypothetical ever could.

Rule 3: Talk Less, Listen More

The person with the most to learn should be doing the least talking. If you are speaking more than 30% of the time in a discovery conversation, you are pitching, not learning. Every minute you spend explaining your idea is a minute the customer is not telling you something you need to know.

Why Direct Questions Fail

The Mom Test rules are not arbitrary. They are countermeasures against specific cognitive phenomena.

The Left-Brain Interpreter

Neuroscience research has identified what Torres calls the "left-brain interpreter" — a module in the brain that compulsively generates coherent explanations for behavior, even when the real cause is unknown or inaccessible to conscious awareness ^cdh. When you ask someone "Why did you switch from Trello to Asana?", their answer sounds logical and confident. But it may be a post-hoc narrative constructed to make sense of a decision that was actually driven by a colleague's offhand recommendation, or a frustration they cannot quite articulate, or simple inertia.

This is not a flaw in the person. It is how brains work. The implication for research: do not ask "why" directly. Instead, ask "what happened" and reconstruct the why from the specifics of their story.

Social Desirability Bias

People want to appear rational, competent, and agreeable ^{prr just-enough}. In an interview context, this means they will:

• Overstate how much they care about the problem you are asking about
• Understate behaviors they consider embarrassing (workarounds, mistakes, laziness)
• Agree with statements you clearly believe
• Offer positive feedback on anything you show them

The only reliable antidote is to ask about behavior rather than attitudes, and about the past rather than the future. People can embellish their motivations, but they have a harder time fabricating specific behavioral details.

The Three Types of Bad Data

Fitzpatrick identifies three categories of information that feel like signal but are actually noise ^mom-test. Learning to recognize them in real-time is one of the highest-leverage interviewing skills.

1. Compliments

"That's a really cool idea!" "I love it!" "You guys are going to do great."

Compliments are the most dangerous form of bad data because they feel so good. After hearing them, you leave the conversation energized and confident. But a compliment contains zero information about whether someone will pay for, use, or even remember your product. Compliments are social currency — they cost the giver nothing and tell you nothing.

How to handle compliments: Deflect them and get back to specifics. "Thanks — but I want to make sure I understand your situation. You mentioned you tried to solve this last month. What happened?"

2. Fluff

Fluff is any generalized, hypothetical, or abstract statement about behavior or preferences:

• "I usually..." (How often is usually? When was the last time?)
• "I would definitely..." (But have you ever actually...?)
• "I think the main problem is..." (Can you give me a specific example?)

Fluff feels substantive because it uses the language of insight. But it is speculation dressed as observation. "I usually go to the gym three times a week" might describe someone who went last in February.

How to anchor fluff into specifics: Use temporal prompts. "You said you usually do X. When was the last time you actually did it? Walk me through that specific instance." The shift from "usually" to "last time" is the shift from fluff to data.

3. Ideas and Feature Requests

"You know what would be great? If you added a calendar integration." "Have you thought about making it work with Slack?"

Feature requests are the customer doing your job for you — and doing it badly. Not because they are unintelligent, but because they have deep knowledge of their problem and shallow knowledge of your solution space and constraints. Their feature request is a symptom. Your job is to diagnose the underlying need.

How to handle feature requests: Ask "Why do you want that?" and then keep asking. "You mentioned a calendar integration — what's the workflow that's breaking down for you right now? Can you show me what you do today?" The request for a calendar integration might reveal a scheduling problem that has a better solution than the one the customer imagined.

See Feature Request Excavation for a step-by-step process.

Story-Based Interviewing

The antidote to all three types of bad data is the same: get the customer to tell you a specific story about a specific instance ^{cdh mom-test}.

Stories are hard to fake. When someone recounts the last time they dealt with a problem, they include concrete details — the tool they opened, the workaround they tried, the moment they gave up, the thing that frustrated them. These details are the raw material of insight.

Torres formalizes this as the core technique of continuous discovery ^cdh. Instead of asking "How do you feel about project management tools?", ask:

• "Tell me about the last time you felt overwhelmed by a project. What was happening?"
• "Walk me through your morning. What's the first work thing you do?"
• "When was the last time you looked for a new tool to solve this? What triggered that search?"

Each of these prompts anchors the conversation in a real event. The customer's response will contain details you could never have anticipated, problems you did not know existed, and workarounds that reveal unmet needs.

See Story-Based Interview Guide for prompt templates.

Primary Market Research as Fuel

Bill Aulet frames customer discovery as "Primary Market Research" and emphasizes three methods that complement interviews ^de:

1. Observation: Watch people in their natural environment doing the thing you care about. What they do often contradicts what they say. A 30-minute observation session in someone's workspace can reveal more than a 60-minute interview.

2. Interviews: Structured conversations following the principles above. Aulet emphasizes that these should happen with potential end users and economic buyers, since the two are often different people with different needs.

3. Ethnography: Extended immersion in the customer's context. This is observation taken further — spending a day shadowing a nurse, a teacher, a sales rep. Ethnography is expensive in time but produces the deepest understanding.

The common thread: go to where the customer is. Do not invite them to your office to look at your prototype. Enter their world first.

The "Never Ask What They Want" Rule

This principle appears across multiple sources ^{just-enough mom-test} and deserves its own emphasis: never ask customers what they want, need, or would pay for.

These questions feel natural and useful. They are neither. "What do you want?" invites the customer to design your product. "What do you need?" invites a socially desirable answer about what they should need. "Would you pay for this?" invites a commitment that costs them nothing to make and means nothing when they make it.

Instead, infer what they want from what they do. What tools do they use today? What workarounds have they built? Where do they spend money? What have they already tried? These behavioral signals are orders of magnitude more reliable than stated preferences.

The strongest signal of all: has this person already spent time or money trying to solve this problem? ^mom-test. If they have, the problem is real and the motivation is real. If they have not, no amount of verbal enthusiasm changes the fact that this problem has never been important enough to act on.

Who to Talk to First

Earlyvangelists

Not all potential customers are equally useful to talk to early. Fitzpatrick and Aulet both emphasize finding people who are already actively seeking a solution ^{mom-test de}:

• They have the problem (not hypothetically — right now)
• They know they have the problem (they can articulate it)
• They have tried to solve the problem (they have spent time or money)
• They are unhappy with existing solutions (they are motivated to switch)
• They have the budget or authority to act (they can actually buy)

These people — sometimes called earlyvangelists, early adopters, or desperate customers — give you the most honest, most detailed, most useful feedback because they are not being polite. They are trying to solve a real problem and they want you to succeed because it serves their interests.

Customer Slicing

Fitzpatrick introduces "customer slicing" as the process of narrowing from a broad market to a specific who-where pair ^mom-test. "Small businesses" is not a customer segment. "Independent coffee shop owners in Portland who have been open less than two years" is.

The narrower your initial segment, the more coherent the feedback, the easier it is to find interviewees, and the more actionable the insights. You can always expand later. Starting broad produces contradictory signals that paralyze decision-making.

See Customer Segmentation Matrix for a structured approach.

Continuous Interviewing

The Weekly Cadence

Torres argues that customer interviews should not be a phase — they should be a continuous weekly practice ^cdh. The product trio conducts at least one customer interview per week, every week, regardless of what phase the product is in.

This cadence has several advantages:

• Small batches of insight prevent the team from drifting too far from reality between research efforts
• Pattern recognition improves when you hear similar stories week after week
• Recruiting becomes routine rather than a heroic effort before each "research sprint"
• The team stays connected to customers even during heads-down building phases

Automated Recruiting

The biggest barrier to continuous interviewing is not willingness — it is logistics. Torres recommends automating recruitment by embedding interview requests into existing customer touchpoints ^cdh: post-transaction surveys, in-app prompts, support ticket follow-ups, onboarding flows. The goal is a steady stream of willing participants so the team never has to choose between building and learning.

See Continuous Interview Setup for implementation details.

Research Questions vs. Interview Questions

One of the most common mistakes in customer discovery is confusing what you want to learn with what you should ask ^{just-enough prr}.

Research questions are what the team needs to know:

• "Is scheduling the primary pain point for our target customers?"
• "Why do people abandon our onboarding flow?"
• "What job is our product hired to do?"

Interview questions are what you actually say to the participant:

• "Walk me through how you planned your last event."
• "Tell me about the last time you signed up for a new tool. What happened?"
• "When was the last time you were frustrated at work? What was going on?"

Research questions are direct and analytical. Interview questions are open, behavioral, and specific. The gap between them is where the skill of interviewing lives. A good interview guide translates each research question into a set of indirect, story-eliciting interview questions that surface the answer without biasing the response.

Lombardo and Bilgen formalize this as the distinction between the "field guide" (what you ask) and the "research plan" (what you want to learn) ^prr. The research plan is shared with stakeholders. The field guide is what the interviewer holds during the session. Conflating the two produces interviews full of leading questions that confirm what you already believe.