The short answer: diagnose which question patterns you get wrong, learn the method for each one, drill it with fresh questions until it’s fast, then practise under your real test’s time pressure. That loop — not general maths revision — is what moves scores.
Decades of learning research point at two techniques that outperform everything else: retrieval practice (testing yourself rather than re-reading) and spaced practice (revisiting a skill days later rather than cramming it once). Numerical reasoning is unusually well suited to both, because the tests reuse a finite set of patterns — percentage change, ratio splits, table lookups, series — and each pattern has a small number of named mistakes that cause most wrong answers.
Start with a mixed, scored test, not a textbook. You need to know which of the twelve areas leak marks for you — most people are wrong about their own weaknesses. The free 12-question diagnostic names them in ten minutes.
For each weak area: read the method (two or three lines, one worked example — see the method guides), then immediately drill that one pattern with feedback after every question. When you get one wrong, don’t just note the right answer — name the mistake. "I used the new value as the base" is fixable; "I got it wrong" is not. Blocked practice like this is how a method becomes automatic.
Accuracy without pace fails these tests. Once a pattern holds untimed, practise it — and then mixed sets — at your actual test’s pace, which varies enormously: roughly 75 seconds per question on SHL, 22 on the CCAT, 14 on the Wonderlic. Use the format simulators to train under the real timing and calculator policy, and learn the skip reflex: on speed tests, a question you can’t start in a few seconds should be skipped without guilt.
A pattern fixed on Tuesday lapses by the weekend unless it comes back. Re-test fixed areas with fresh questions a few days later — short, mixed sessions beat marathons. Daily 10–15 minute sessions are the most reliable structure: enough to keep every fixed skill warm, short enough to actually happen.
Two weeks out: the full loop above — diagnose, fix the two or three weakest areas, daily short sessions, simulator runs in week two, a full timed mock two days before.
One week out: diagnose today; fix only the two weakest areas; every other day in your test’s real format; mock at day five.
Two days out: skip new learning. One diagnostic, one review of the named mistakes, one simulator run at real pace each day, and sleep — pacing discipline and rest move more marks at this point than any new method.
Start with the free diagnosticMeaningful gains come in days, not months. Most candidates improve fastest in the first week of focused practice, because the biggest early wins are pattern recognition and pacing — both quick to train.
Match your real test. SHL, Saville and cut-e allow calculators; CCAT, Wonderlic and PI don't. Training in the wrong mode builds habits that hurt you on test day.
Review beats volume. One question whose mistake you understand and re-test is worth more than ten you rush past. The exception is pacing work, where volume at real speed is the point.
Not measurably for these tests. The reliable gains come from practising the actual question patterns under the actual time constraints — specificity wins.