Avoiding Bad Data

Compliments, fluff, and ideas

Even when you follow The Mom Test rules, bad data can sneak into your conversations. Fitzpatrick identifies three types of bad data that poison customer learning: compliments, fluff (generics, hypotheticals, and the future), and ideas. Recognizing and deflecting these is essential to getting conversations that actually matter.

The Three Types of Bad Data

Customer conversations go wrong not because people lie deliberately, but because certain types of responses feel informative but contain zero useful signal. You need to learn to spot them in real time and steer the conversation back to solid ground.

“Compliments are the fool’s gold of customer learning: shiny, distracting, and entirely worthless.” — Rob Fitzpatrick

1. Compliments

Compliments are the most dangerous form of bad data because they feel so good. When someone says “That’s a really cool idea!” or “I love it!”, your brain lights up with validation. But compliments contain no usable information. They tell you nothing about whether someone will actually change their behavior or spend money.

How compliments sneak in:

How to deflect: Do not accept the compliment. Bring the conversation back to their life. “Thanks — but can you tell me more about how you are handling this currently?”

2. Fluff: Generics, Hypotheticals, and the Future

Fluff is any statement that is not grounded in concrete past behavior. There are three sub-types:

All three sound informative but contain no reliable signal. People are terrible at predicting their own future behavior. The only reliable data is what they have actually done.

How to deflect: Anchor everything in specifics. “You said you usually do X — can you walk me through the last time that actually happened?”

3. Ideas

When customers start suggesting features or solutions, it feels like you are getting valuable product input. But feature requests from customers are almost always bad product direction. Customers are great at identifying problems but terrible at designing solutions.

When someone says: “What you should really do is add a feature that…”

How to deflect: “That’s interesting — what would that let you do that you can’t do now?” This brings the conversation back to the underlying problem, which is what you actually need to understand.

The Deflection Toolkit

Learning to redirect bad data back to useful territory is a core skill. Here are the key deflection patterns.

Deflecting Compliments

The key is to never let a compliment end the conversation. Always follow up with a question that forces specifics.

Deflecting Fluff

Digging Through the Fluff

The deeper problem with fluff is that it can contaminate your entire understanding of the market. If you collect a dozen “I would definitely use that” responses, you might feel like you have validated demand. But all you have is a pile of empty promises.

“The world’s most deadly fluff is: ‘I would definitely buy that.’” — Rob Fitzpatrick

The Fluff-to-Fact Conversion

Every fluffy statement can be converted into a concrete question:

Emotional Signals vs. Intellectual Signals

Not all data is verbal. Pay attention to emotional signals — moments when someone gets visibly frustrated, excited, or embarrassed while describing their situation. These emotional spikes often point to the most important problems.

Real Signals to Watch For

These behavioral signals are far more reliable than any verbal promise.

Key Takeaways

← Previous: Chapter 1 Next: Chapter 3 →