How much of your roadmap should customers actually get to decide?

I try really hard to be the kind of founder that listens to users. The whole idea of building something people want is that you're supposed to be paying attention to them, not building in a vacuum.

But every time I've followed the loudest requests, I've ended up building things almost nobody else wanted. One vocal customer asks for a feature, it feels urgent because they're right in front of me, and three weeks later it's sitting there barely used.

The problem is that customers tell you their problems, not the optimal implementation. They ask for the thing they think they want, and it's my job to figure out the underlying problem, which is often something completely different.

So now I sort of treat requests as data, not instructions. If ten unrelated people describe the same pain in different words, that's good signal. If one person wants a specific button, that's usually just one person.

But I'll admit the line is blurry, and I've talked myself out of good ideas by calling them "just one customer" when I should've listened.

So where do you draw the line: how much of your roadmap do you actually let customers decide, and how do you keep that from turning into building whatever the loudest person asked for last?