Avoiding Bad Data

Chapter 2 catalogs the specific failure modes that make customer conversations produce bad data. Fitzpatrick names three categories. Compliments are the most flattering and the most useless: "I love this idea," "this is so smart," "everyone needs this." A compliment carries zero information about future purchase behavior. The polite move when someone shares an idea is to encourage them; almost everyone will perform that politeness; nothing about it tells you whether they would open their wallet. The discipline is to notice the compliment as a social signal and refuse to count it as data.

Fluff is the next category — vague generalities about what people would do, want, or use in some future scenario. "I would definitely pay for that" is fluff. "I would totally use that all the time" is fluff. These statements feel substantive but they're built entirely on speculation about a future the speaker has not actually experienced. Fitzpatrick's rule: any sentence with a conditional ("would," "could," "should," "might") is fluff and should be steered back to concrete past behavior. The interview's job is to convert imagined futures into described pasts.

Ideas are the third trap, and the most subtle. When a customer offers you a feature idea or product suggestion, the temptation is to write it down and treat it as gold. But customers are bad at solution design — they are good at identifying problems. A customer saying "you should add a button that does X" is information about the underlying problem they're trying to solve, not a specification to build. The chapter argues that founders should listen for the WHY behind every feature request, dig until they reach the underlying problem, and design solutions themselves from that information.

The chapter introduces what becomes a recurring discipline throughout the book: anchoring conversations in the customer's actual life. What did they do last Tuesday? How are they currently solving this problem? How much time did it take them last week? These questions force concrete recall. They produce data founders can act on. Anything that asks the customer to predict their own future behavior or design a product is, in Fitzpatrick's frame, a category error — you are using the customer for the wrong job, and they will helpfully give you the wrong answer.

The unifying discipline across all three failure modes is to steer every conversation away from opinions about your idea and toward concrete facts about the customer's actual life — what they have already done, what they currently pay for, and where they have struggled — because only their past behavior, never their predicted future enthusiasm, carries information solid enough to build a business on. Compliments, fluff, and unprompted feature ideas all feel like progress and all mislead; facts about lived behavior are the only reliable data a conversation can yield.