How I stopped building features nobody uses — the question that changed everything

I have a graveyard of features I'm proud of technically that nobody uses.

Beautiful code. Real engineering. Zero real-world impact.

For a long time I blamed users. "They don't understand how to use it." "They need to be educated about the value."

Eventually I stopped blaming users and started looking at my own process.

The problem wasn't that users didn't understand the features.

The problem was that I was building features based on my interpretation of user problems — not the actual problems themselves.

Here's the specific mistake I kept making:

User: "I wish I could do X faster."
Me (in my head): "They need Y feature."
Me (in reality): Building Y feature.

What I missed: X faster might mean a dozen different things. And Y might solve none of them.

The gap between what the user described and what they actually experienced was where all my wasted work lived.

The question that fixed this:

I started ending every user conversation with:

"If this problem was solved, what would you be able to do that you can't do right now?"

This question does something specific: it moves the conversation from describing a problem to describing a desired outcome.

And outcomes are much more useful to build toward than problem descriptions.

Example of how this changed things:

Old conversation:
User: "The reporting is confusing."
Me: builds a redesigned reporting UI

New conversation:
User: "The reporting is confusing."
Me: "If the reporting was better, what would you be able to do that you can't now?"
User: "I'd be able to show my manager our progress in the weekly meeting without having to export to Excel first."

That's a completely different feature than "redesign the reporting UI."

That's a one-click export to a presentation-ready format.

Much simpler to build. Much more directly solving the actual problem.

What I learned:

Users describe symptoms. Outcomes reveal the actual problem.

The feature that solves the symptom and the feature that achieves the outcome are often very different things.

"What would you be able to do?" is the fastest path from symptom to outcome I've found.

I now ask it in every user conversation. It takes 10 seconds. It saves weeks.

What questions do you use to get past feature requests to actual outcomes?