📖 Business
Three Types of Bad Data
In customer conversations, three types of information consistently feel like validation but are actually noise. Fitzpatrick calls these the three types of bad data: compliments, fluff, and feature requests. Each one triggers a dopamine hit ("they like it!") that makes founders and product managers feel like they are making progress — when in reality they have learned nothing. Recognizing and deflecting these three categories is the core defensive skill of customer development. If you cannot tell the difference between signal and noise in a conversation, every conversation will feel like validation.
2
Minutes
2
Concepts
+45
XP
1
How It Works
  • (1) Compliments — "That is a great idea!" "I love it!" "I would definitely use that!" Compliments are social lubrication. People say them to be nice, to end an awkward conversation, or because they genuinely think the idea sounds cool in the abstract. None of this predicts behavior. Deflection: "Thanks — but I want to understand your situation better. How do you currently deal with this?"
  • (2) Fluff — Generic, hypothetical, or future-tense statements. "I would probably use it." "I generally spend a lot on that." "I always struggle with X." Fluff sounds like data but contains no specifics. "I always struggle with X" could mean it happened once six months ago or it happens every day. Anchor to specifics: "When is the last time that actually happened? Walk me through it. What did you do?"
  • (3) Feature requests — "You should add X!" "It would be great if it could do Y!" Feature requests sound like engaged feedback but they are the customer designing your product based on their limited perspective of what is possible. They are usually describing a symptom, not the root problem. Dig into motivation: "What would that let you do? Why is that important? How are you dealing with that today?"
  • All three types share a common trait: they are about the future, the hypothetical, or the general. Real signal is about the past, the specific, and the concrete
  • The emotional trap: bad data feels better than good data. A compliment makes you feel smart. A specific story about how they already solved the problem without you makes you feel irrelevant. But the second one is infinitely more useful