What Is a Good Score on a Psychometric Test? Percentiles, Stens and Cut Scores

Almost every question candidates ask after an assessment is a version of the same one. I got 32 out of 40, is that good? I finished 28 of 50, is that a fail? The honest answer is that nobody, including the employer, can tell you from that number alone, because the raw score is the first thing the scoring process throws away.

Why your raw score is discarded

A raw score is meaningless on its own for a reason that becomes obvious once you see it: it depends entirely on how hard the test was. Twenty-eight correct out of fifty is an excellent result on a hard test and a poor one on an easy test, and the candidate has no way of knowing which they sat.

Test publishers solve this by comparing your performance against a group of people who have already sat the same instrument. That group is called the norm group, and the comparison is what gets reported. Your raw score exists for about one step of the calculation and then disappears.

Two consequences follow immediately, and they surprise people. First, there is usually no percentage anywhere in an assessment report. Second, the same raw score can produce different reported results for two candidates, if they were compared against different groups.

The norm group decides everything

This is the single most important idea on this page. A norm group is the reference sample your result is scored against, and publishers maintain several for the same test.

Percentiles, stens, stanines and T scores

Four reporting scales dominate, and they all express the same underlying comparison with different granularity. Learning to read them takes about two minutes and saves a lot of misinterpretation.

The rough conversions are worth memorising, because reports frequently mix scales. A sten of 7 is around the 77th percentile. A stanine of 7 is around the 89th. A T score of 60 is around the 84th. If a report gives you a band rather than a number, that is deliberate: publishers band results to stop people over-reading small differences that are within measurement error.

Cut scores, and why nobody will tell you the pass mark

A cut score is the threshold an employer applies. Three things about it explain most of the confusion around assessment results.

The employer sets it, not the publisher

Test publishers supply the scale. Employers decide where to draw the line, and they do it against their own hiring needs. This is why a page claiming “you need the 70th percentile to pass at company X” should be treated as unsourced folklore.

It moves between intakes

A cut score is often set to produce a manageable number of candidates for the next stage. In a year with twice the applications, the same performance can fall below a line it cleared last year. Nothing about you changed.

Sometimes there is no cut score at all

Plenty of processes use assessment results as one input among several, or as a ranking rather than a gate. In those processes the question “did I pass?” has no answer, which is why recruiters sometimes seem to be dodging it.

Why an adaptive test feels hard

An adaptive test selects your next item based on how you answered the last one. Get one right, get a harder one. Get one wrong, get an easier one. Over a sitting this converges on the difficulty level where you are answering roughly half the items correctly.

Which means: on a well-functioning adaptive test, everybody finds it hard, and finding it hard tells you nothing about your result. A candidate who breezed through an adaptive test has usually been converged downwards. A candidate who left feeling battered by the last five items has usually been converged upwards.

Adaptive designs also break the arithmetic candidates try to do afterwards. Your item count is not comparable with anybody else's, because you did not sit the same items. Aon, Korn Ferry Talent Q, Assessio's Adaptive Matrigma and Sova's reasoning tests all use adaptive or semi-adaptive designs, and practising them is largely about getting used to that feeling.

What you can actually find out

• Ask for your report. In the UK and EU you generally have a right to personal data held about you, and in several markets that includes assessment reports. Publishers write feedback reports precisely so they can be shared.
• Ask which norm group was used. A recruiter often will not know, but the question sometimes reaches someone who does, and it is the only number that makes a percentile interpretable.
• Do not ask for the pass mark. Employers treat cut scores as commercially and legally sensitive, and asking rarely gets an answer.
• Do not compare with friends. Different norm groups, often different item sets, sometimes different tests with the same name. The comparison is not measuring anything.

The practical upshot

Stop trying to reconstruct your score. It is not recoverable from what you remember, and even if it were, the raw number is not what anyone is looking at. The controllable things are the ones that come before the sitting: knowing the format, working at a steady rate, and knowing whether the test you are on penalises guessing. Those change your percentile. Post-hoc arithmetic does not.

If you want to know how much of the outcome preparation actually moves, the companion piece on whether practising psychometric tests works gives the honest version.