We once tracked a batch of hires through their first year. One of them had the highest assessment score in the group. Clean reasoning section, strong situational judgement, near perfect on the values questionnaire. On paper, the safest bet in the room. Within eight months, this person was the one the team worked around.

Nothing dramatic had happened. No single incident. But the pattern was clear. They said the right things in meetings and did the wrong things between them. They scored well and performed poorly, and the assessment never saw it coming.

We have seen this enough times now that we no longer treat a high score as good news on its own. We treat it as a question.

What the score actually measured

An assessment measures how someone performs on an assessment. That sounds like a joke. It is not. It is the most important sentence in this piece.

When a candidate sits a test, they are doing a specific thing: reading questions, choosing answers, working under time pressure, in a room or on a screen, with no colleagues, no consequences, and no history. What they are not doing is the job. The job has people in it. The job has bad days, conflicting instructions, and small compromises that add up.

So a strong score tells you the candidate is good at the test. It may also tell you they are good at the job. It does not prove it. The two are related, but the relationship is looser than most hiring processes assume.

The gap between the test and the task

The gap shows up in predictable places. Tests reward tidy answers. Work rewards messy judgement. Tests ask what you would do. Work is what you do when nobody asked.

We have watched candidates who aced situational questions about ethics and then, in the job, took the easiest path whenever no one was checking. We have watched candidates who struggled to articulate a value and then lived it, quietly, every day. The assessment had it backwards both times.

The reasons are not mysterious:

• Tests measure what someone knows, not what they do under pressure.
• Candidates can prepare for tests, and many do, which inflates the score.
• Tests reward the confident and the fluent, not always the capable.
• Answers are given alone, but work is done with others.
• A single sitting cannot capture a pattern that only shows over months.

None of this makes assessments useless. It makes them partial. The mistake is treating a partial signal as a complete one.

Why good scores mislead

There is a second problem, and it is subtler. A good score creates confidence. The interviewer relaxes. The reference checks get lighter. The probation period becomes a formality. Because the candidate scored well, the organisation stops looking.

So the score does not just fail to predict behaviour. It actively reduces the effort to observe it. We have seen managers skip the hard conversations in month one because the assessment said the person was fine. By month six, the pattern is set and much harder to change.

This is the quiet cost. The assessment does not only miss the gap. It hides it, by making everyone feel the question is already answered.

The cost of a confident hire

When a hire comes in with a strong score, everyone relaxes. The manager assumes the hard part is done. The team assumes the new person is capable. The new person, quietly, assumes they are already trusted. None of these assumptions gets tested, because the score seems to have settled the matter.

That is how a poor fit hides for months. The person is not failing at anything obvious. They are failing at the small, unmeasured things: how they treat a junior, whether they own a mistake, how they behave when a deadline slips and the blame is loose in the room. The score never asked about those, so no one watches for them.

We have seen a confident hire damage a team for a year before anyone said the obvious thing out loud. The evidence was there the whole time. It was simply not on the test.

What actually predicts

In our tracking, the things that predicted on-the-job behaviour were rarely on the test. They were in how the person behaved before they were hired, and in what others said about them when asked the right way.

Three signals came up again and again. First, how the candidate handled a real problem in a past role, told in specific detail, not in theory. Second, what former colleagues said when asked not whether the person was good, but how they behaved when things went wrong. Third, how the candidate treated the least powerful person in the interview, the one who could do nothing for them.

None of these fits neatly into a score. All of them are closer to the job than any test.

One question does more work than any score. Ask a candidate to describe the last time they were wrong about something at work, and what they did next. The strong ones give you a real answer, with detail and a cost attached. The weak ones give you a story about someone else, or a polished lesson with no edge to it.

We have found this single question more revealing than an hour of structured testing, because it cannot be prepared for cleanly. It asks the candidate to show you how they behave when they are not at their best, which is closer to the truth of a working week than any multiple choice section.

Assessments that see behaviour

You can keep your assessment. Just stop asking it to do the impossible. Use it to screen for basics, then spend your real effort where behaviour actually shows.

Ask candidates to walk through a past situation until you reach the uncomfortable detail. Give a small, real task and watch how they handle it, including how they respond to a change mid-way. Check references by asking for behaviour, not verdicts. And treat the first ninety days as the real assessment, with someone watching closely on purpose.

The score is a starting point. The floor is the evidence. We keep both, and we trust the second.

A score is a signal, not a verdict

The agent who scored well and performed poorly taught us something simple. A number tells you how someone did in one controlled moment. It cannot tell you how they will behave when the moment is not controlled, which is most of working life.

So we no longer hire on the score. We hire on the pattern, and the pattern takes time to see. If your process ends at the assessment, you are hiring the candidate's best afternoon, not the colleague you will actually get.

Keep the test. Read it carefully. Then put it down and go looking for the behaviour it cannot show you.