A "good" score is one above the employer’s cut-off — and that cut-off is a percentile, not a percentage. Numerical reasoning tests compare you to a norm group of similar candidates; getting 70% of questions right might put you in the 40th percentile on an easy test or the 90th on a hard one. The percentile is the score.
Your answers produce a raw score (questions correct). The publisher converts that to a percentile against a chosen norm group — for example, "UK graduate applicants" or "managerial candidates." The employer then applies a cut-off: candidates below it are screened out, candidates above it progress. You typically see none of this — at most a pass/fail email — which is why preparing against honest, comparable scores matters.
Adaptive tests (Korn Ferry Talent Q, SHL Verify Interactive) complicate raw scores further: the questions get harder as you succeed, so two candidates with the same number correct can score very differently. The percentile principle, though, is the same.
Employers rarely publish cut-offs, so treat any precise claim with suspicion. The honest picture from the assessment industry: volume screening commonly cuts somewhere between the 30th and 50th percentile; competitive graduate schemes and analytical roles set higher bars. Practically, that makes the 70th percentile a sensible target — clear of almost any screen, with margin for a bad day.
On speed tests like the CCAT (50 questions, 15 minutes) and Wonderlic (50 questions, 12 minutes), almost nobody finishes — average raw scores sit around the low-to-mid 20s out of 50. A score that sounds low as a percentage can be strong: on these tests, comparing yourself to the average matters more than to the maximum. Details on each test’s scale are in our provider guides.
Our free diagnostic gives you a real raw score, names the exact areas you missed, and reports a percentile only against a clearly labelled baseline with a visible sample size. We never invent percentiles or population claims — a prep site that fakes its numbers can’t teach you to beat a test built on real ones. Full method on how scoring works.
A score alone changes nothing — the named weak areas attached to it do. Whatever you scored, the route up is the same loop: diagnose → fix the pattern → train at real pace → re-test.
Get your score nowOne above the employer's percentile cut-off, which you usually won't be told. Common cut-offs sit between the 30th and 50th percentile for volume screening, higher for competitive analytical roles — so aiming for the 70th percentile or above is a sensible safety margin.
Raw score is questions correct (e.g. 18/24). Percentile is how that compares to a norm group (e.g. better than 65% of graduate candidates). Employers decide on the percentile, because raw difficulty varies between test versions.
On most numerical reasoning tests, no — there's no negative marking, so an educated guess beats a blank. Adaptive tests (like Talent Q) work differently: difficulty adjusts to your answers, and the sustained level matters.
Usually not within the same application — most employers allow one attempt, sometimes with a re-test to verify. Across different employers you simply take each one's test fresh, which is why practising beforehand matters.