Every hiring mistake I can trace to a bad reference or a thin résumé, I caught. Those are easy — the process is built to catch them. The hires that actually cost me time, team morale, and in two cases a client relationship, were people who cleared every interview cleanly, had references that checked out, and still turned out to be wrong for the role in a way nobody flagged until they'd been in the seat for months.
What I eventually noticed is that a strong interview and a strong hire measure two different things, and I had spent years treating them as the same thing.
Interviews measure performance under a spotlight. The job rarely is one.
A good candidate is, among other things, good at being interviewed — articulate under scrutiny, prepared with a clean narrative, quick to reframe a weakness as a growth story. Those are real skills, but they correlate weakly with the skill the role actually needs, which in most operating roles is something closer to steady, unglamorous judgment exercised without an audience. I hired several people who were excellent in the room and mediocre in the job, and it took me longer than it should have to notice that I was scoring the wrong thing.
The fix, for me, wasn't a smarter interview question. It was insisting on a paid trial project or a structured work sample wherever the role allowed it — something that measured the actual work rather than the description of the actual work. It's more expensive up front. It has been, without exception, cheaper than the alternative.
The reference call that matters is the one nobody offers you
Candidates supply references who will say good things — that's the reference's whole function in the process, and I don't fault anyone for choosing favorably. What I've learned to do instead is ask every reference, including the flattering ones, a version of the same question: "if you were rebuilding your team today, would this person be your first call, or would you be glad to have them but wouldn't go out of your way?" The gap between those two answers, delivered in the same warm tone, has told me more than anything else in the process. It's a distinction most references will draw honestly if you give them a specific enough question to answer.
I've also started, wherever it's ethical and possible, trying to speak with one person the candidate didn't list — a former peer rather than a former manager. Peers see things managers don't, and they have less invested in the story being clean.
What I check for now that I didn't used to
The single change that's made the biggest difference is asking candidates to walk through a specific failure in granular detail — not "tell me about a time you failed," which produces a rehearsed answer, but "walk me through exactly what happened, in order, including the point where you first suspected something was wrong." People who've done real work stumble into specifics naturally. People reciting a prepared narrative stay abstract no matter how hard you push, and that difference is very hard to fake in real time.
None of this makes hiring reliable. It's still, honestly, the part of running a company I trust myself least on. But it's caught mistakes I would previously have made, and it's cost me almost nothing except the discipline to slow down at the exact moment a candidate is making it easiest to speed up.