Avoiding Bad Data

The chapter walks through the three categories of bad data that founder-customer conversations produce when conducted poorly: compliments, fluff, and ideas. Each is worse than no data because it produces the false impression of validation.

Compliments are bad data because they tell you only that the customer wants to be polite. The customer who tells you they love the idea has revealed nothing about whether they would actually purchase the product. The customer who tells you their friend would totally love it has revealed nothing about anyone's actual purchasing behavior. Compliments feel like validation but contain no information.

Fluff is the category of generic statements (I usually, I always, I would) that paper over the specifics of actual behavior. Real behavior happens in specific instances with specific contexts. Generic statements about behavior are usually reconstructions or aspirations rather than accurate descriptions. Asking for the specific instance — the last time you encountered this problem, what specifically happened, what you specifically did — surfaces the actual behavior the generic statement obscured.

Ideas from customers are the most dangerous category because they feel like product input. Customers who suggest features are not telling you what they would pay for; they are telling you what they would mention politely if the founder seemed eager. Treating customer-suggested features as a roadmap is one of the most common ways founders waste years building products no one actually buys. The chapter's discipline is to redirect the conversation from the customer's ideas to the customer's problems — and to evaluate whether the problems are worth solving regardless of whether the customer can articulate a solution.