How do you choose the next improvement when users want different things?

How do you decide what to improve next when users are asking for different things?

I’m finding that the loudest request is not always the most useful one. One person may ask for a new feature, while several quieter signals point to a confusing step or a task people keep doing manually. It’s tempting to keep a running vote, but that can turn the roadmap into a popularity contest rather than a way to remove real friction.

For founders who have had to make this call, what evidence do you trust most: repeated behavior, frequency of the problem, willingness to pay, or something you learn in a conversation? How do you tell the difference between a request worth building and a request that is better solved with clearer copy or a workaround?

I’ve been thinking about this while building IndieNeed, a directory for indie products (https://www.indieneed.com/submit).