Most screening filters test one skill, and it is almost never the job.
An organisation almost never decides that somebody cannot do the job. It decides that somebody did not do well in a room, or on a form, or in the first six weeks, and then it writes that down as a fact about the person instead of a fact about the room. Ability written off as attitude, by filters built for somebody else.
Where the filter came from
Every screening step in an organisation was built by somebody solving a real problem. Too many applicants, not enough hours, a manager who kept hiring badly. So you write a form. Then a panel asks the same person to explain themselves out loud, on a set day, at a set time, to strangers.
Nobody sat down and decided the job required all of that. It accumulated.
Worth remembering when someone tells you the process is neutral. It is not neutral. It is just old.
What it actually measures
An interview measures how well a person performs a conversation about work, under time pressure, with an audience... which is a real skill. It is simply not the same skill as doing the work, and for most roles the overlap is thin.
Probation measures how fast somebody reads a culture nobody has written down. The performance review measures how well they can narrate their own year in the shape the form expects.
None of these are frauds. They are proxies, and proxies drift. Some people are excellent at the proxy and ordinary at the job. Plenty are the other way round. The filter cannot tell you which is which, because the proxy is the only thing it ever saw.
How the result gets renamed
This is where the damage happens. The filter produces a result, and then someone writes the result down in words about the person.
Not a great fit. Didn't seem interested. Attitude problem. Doesn't take feedback.
A measurement of format becomes a description of character, and once it is in a file it stops being a measurement at all.
And it travels. Into the reference, into the next manager's assumptions, and eventually into the person's own account of themselves. The original thing that was measured, which was how somebody handled twenty minutes in a room on a Tuesday, vanishes completely.
What to change first
The cheapest fix is not a training day about difference. It is taking each filter, one at a time, and asking what it is for, then asking whether the format is doing anything the work needs.
Send the questions in advance. It costs nothing... you already wrote them. A candidate who prepares is not cheating. Preparing is the exact behaviour you want from an employee.
Give people a piece of the real work instead of a hypothetical about the real work.
Write down what fit means in your organisation, in specifics, before anybody uses the word about a candidate. If it cannot be written down, it is not a criterion. It is a feeling in a criterion's clothes.
And when a review says attitude, make somebody name the behaviour and the moment it happened in. Half the time the sentence falls apart while they are writing it.
None of this requires you to think differently about anyone. It is admin. Rewrite the question, hand over a real task, take the word fit off the form and put a named behaviour in its place, then count how many people stop failing. Same people, different filter.