Why Do Good Interview Processes Still Produce Bad Hires?

Here’s the part nobody wants to hear. A great interview process can still produce a bad hire, and usually does it without a single thing going visibly wrong. Every interviewer was prepared. Every scorecard was filled in. The loop ran on schedule. Then the panel held a meeting and made a decision.

I’ve sat in a lot of those meetings. Nearly all of them were called an interview debrief. Nearly none of them were one.

Five opinions, no new information

You’ll recognize the shape of this conversation if you’ve sat in that room. Everybody on the interview team arrives. Someone goes around the table, and every interviewer gives a recap. Strong yes, really impressive on system design. Yes from me, no concerns. Leaning yes. Fifteen minutes, five opinions, no real disagreement.

However, notice what just happened. Every one of those statements was already in a scorecard that everyone could have read. As a result, the meeting produced no information that didn’t exist before it started. Nobody was persuaded of anything, because nobody argued anything. If no position moves, nobody is deliberating. They’re ratifying. That is consensus hiring, and it passes for rigor precisely because everybody in the loop took part.

Meanwhile the conversation that would have been worth holding never starts: what this person is actually strong at, what they’re demonstrably not, and what happens to your team if they join it.

The debrief that changed a no into a hire

Early in my career I sat in one of those rooms and watched it go differently. The panel was moving toward a no. But the hiring manager had built an environment where a junior voice was expected to speak up, so I did. What I had seen in my own interview didn’t match the conclusion forming around the table, and I said so. What the candidate had actually done, what they’d said, and why I read it the way I did. That changed a no-hire decision into a hire decision. Over the next five years, that person was promoted multiple times.

You are making two decisions, not one

The confusion underneath all of this is that hiring a person involves two separate decisions, and most companies collapse them into a single yes.

The first is whether the candidate clears the bar. That’s an evidence question, answered against a scorecard, and it’s the one the read-out is nominally about. Here’s the uncomfortable finding: it shouldn’t be a group conversation at all. In 2013, Nathan Kuncel and colleagues published a meta-analysis in the Journal of Applied Psychology. It compared two ways of combining candidate data: by a rule agreed in advance, or by letting experienced people weigh everything holistically, the way a panel does. Across 25 samples, for predicting job performance, the rule reached a validity of .44. Holistic expert judgment reached .28, a gap the authors describe as an improvement in prediction of more than 50%.[1] Talking about the scores makes them worse, not better.

The second is what this hire does for the team. Not whether the person is good enough, but where they sit against the people already here, which competencies you are covered on, where you are thin, and whether this hire levels you up or gives you a second copy of a strength you already have.

Two candidates, same bar, different value

Often two candidates can both clear the hiring bar, but aren’t remotely interchangeable. One is an addition to your team that is excellent at something you’re already excellent at. The other covers a gap that has been quietly costing you. Yet a scorecard won’t necessarily tell those two apart. Why? Because a scorecard is designed to measure the candidate against the role, not necessarily the team.

Consequently, answering the team question requires the panel, because the relevant information is distributed across it. Your staff engineer knows where the architecture is fragile. Your product manager knows which conversations keep stalling. The hiring manager knows who is close to being ready for the next level and what a new senior hire would do to that. None of that is in a scorecard, and none of it comes out in a round of verdicts.

Group decision research is blunt about what happens when you don’t ask for it explicitly. In 2012, Lu, Yuan and McLeod pooled 65 studies covering more than 3,000 groups in Personality and Social Psychology Review. They found that groups discuss information everyone already shares far more than information only one member holds, by roughly two standard deviations. Where the right answer depended on pooling what individuals uniquely knew, groups were eight times less likely to find it.[2]

In effect, a read-out is that failure by design. It asks each person for their conclusion, which is the one thing they all have in common, and never asks for the observation only they hold.

What a real debrief covers

An interview debrief worth holding works through four questions, in this order, and none of them is “yes or no.”

Are you having the meeting you think you are?

Of course most hiring managers, asked whether they run debriefs, will say yes. They hold the meeting. It’s in the calendar. That’s exactly what makes this hard to see. The failure doesn’t look like a missing step, it looks like a step that happens every week and produces agreement every time.

Interview architecture gets the attention because it has artifacts: question banks, scorecards, rubrics. That work is genuinely part of the infrastructure worth investing in, and I’d argue for it every time. But a scorecard is an input to a decision, not the decision. If the only thing you ever do with five filled-in scorecards is read them aloud in sequence, you’ve built the instrument and skipped the diagnosis.

So before your next loop closes, ask what your interview debrief is actually for. If the honest answer is confirming what everyone already wrote down, you don’t have one yet.