Hire the person who redesigned something they didn't have to
I've learned to look past the hero project for the small thing someone improved because they couldn't leave it alone.
Most resumes are organized around permission. Someone gave you a project, you delivered it, you list it. The bigger the grant of permission, the higher up the page it goes. I read those projects, but they're not what I'm hunting for. What I want is the thing nobody asked for — the internal tool someone rewrote over a weekend because using it three times a day made them wince. The onboarding doc they restructured because watching one new hire get lost was one too many. The export button they moved four pixels because the old position bothered them every single time.
That instinct is the most reliable signal I've found, and almost no interview process is built to surface it.
The hero project tells you less than you think
A flagship launch on a resume tells me a team executed under deadline pressure with a budget and a mandate. That's useful, but it's heavily confounded. There was a deadline, so something shipped. There was a design review, so the work met a bar someone else set. There was a PM defining scope, an org pushing it forward, a launch date that made "good enough" the finish line. I can't separate the candidate's standards from the machinery that surrounded them.
The redesign nobody asked for has none of that machinery. No deadline forced it. No review demanded it. No one would have noticed if it stayed broken. The only thing that moved it was that the person could not look at the broken thing and feel fine. That's the variable I'm actually trying to measure, and it's almost never on the page until you ask for it directly.
What the unrequested fix actually proves
When someone fixes a thing they weren't assigned, three traits show up at once, and you usually can't fake any of them under interview lighting.
- →They noticed. Most people have learned not to see friction; they route around it on autopilot. Noticing is itself rare, and it can't be taught in a sprint.
- →They had a standard the environment wasn't enforcing. The bar came from inside them, not from a rubric, which means it travels to the next job whether or not anyone is watching.
- →They closed the loop. Plenty of people complain about the broken thing in standup. Far fewer quietly fix it and move on without waiting for a ticket.
That third one matters most to me. Taste without follow-through is just opinion, and I have met a lot of people with exquisite opinions and empty commit histories. The fix is the proof that the standard is load-bearing — that it actually changes behavior rather than sounding good in a design critique.
Taste without follow-through is just opinion, and opinions don't ship.
How to interview for it
You can't ask "are you detail-oriented" and learn anything; everyone says yes. So I ask for the artifact directly. Tell me about something you improved that nobody asked you to. Then I push on the part everyone glosses over — not what you changed, but what made you unable to leave it alone.
The answers sort themselves fast. Some people light up. They remember the exact spacing, the specific error message, the way the old flow made them feel slightly stupid every time, and they'll walk you through the fix at a level of detail no one fabricates on the spot. Others go quiet, or reach for a sanctioned project and rebrand it as passion. The quiet isn't a character flaw — plenty of strong people have spent years in places that punished going off-script. But the ones who light up are telling you what their default is when no one is assigning the work, and the default is what you're hiring.
One caution, because the trait has a failure mode. The person who can't leave anything alone, who rewrites a working system because the variable names offend them the week before launch, is the same instinct pointed in the wrong direction. So I ask the follow-up: how did you know it was worth the time? The good answer connects the fix to someone other than themselves — the new hire who got lost, the teammate who kept hitting the same wall. The instinct plus judgment about when to deploy it is the actual hire. The instinct alone is a liability with good intentions.
Why this is the whole job now
The work that survives is increasingly the work no one specs in advance. A model can execute a clear ticket better and faster every quarter. What it can't yet do is feel that something is slightly wrong, decide it's worth the detour, and quietly make it right. The premium is moving toward the judgment about what deserves attention — and the unrequested redesign is the cleanest evidence that a person has it.
Hire the one who fixed the thing nobody was looking at. Not because the fix was important — it usually wasn't, on its own. Because the fix is the only honest proof that their standards don't switch off when the supervision does. Everything else on the resume tells you what they'll do when they're told. This tells you what they'll do when they're not.
