Skip to main content

Interview preparation

Numerical Reasoning Tests: What They Cover and How to Pass

·7 min read

Numerical reasoning tests sit near the front of most graduate and professional application pipelines, and they reject more candidates than any interview does. That sounds brutal, and in one sense it is: the sift exists to shrink the pile. But the test itself is narrower and more coachable than its reputation suggests.

The open secret: the maths is GCSE level. Percentages, ratios, averages, currency conversion, reading tables and charts. Nobody fails a graduate numerical test because the mathematics is beyond them. They fail on speed, accuracy under pressure, and unfamiliarity with the question grammar, and all three of those respond quickly to practice.

What the tests actually contain

A typical test gives you a table or chart (sales by region, headcount by year, costs by product) and asks four or five questions per exhibit, roughly 60 to 90 seconds each. The recurring operations:

  • Percentage change, in both directions: what rose 12 percent, and what a fall from 480 to 415 is in percentage terms.
  • Ratios and proportions: splitting totals, scaling recipes of numbers up and down.
  • Weighted averages: the average that accounts for group sizes, not the average of averages (a favourite trap).
  • Multi-step chains: convert the currency, then take the percentage, then compare. Each step easy; the chain is the test.
  • Cannot say questions: whether the data supports a conclusion at all. Read exactly what the axis and units say, not what you expect them to say.

Why people actually fail

Time, not difficulty. Untimed, most graduates score highly. The clock changes everything: it punishes deriving methods from scratch and rewards having them ready. The candidate who recognises this is a weighted average inside two seconds has bought twenty seconds of calculation time.

Calculator fumbling. Know whether the test allows a calculator (most do) and practise with the one you will use. Percentage change as (new minus old) over old should be a finger pattern, not a decision.

Reading the wrong number. Under pressure, eyes jump to the biggest figure or the wrong year. Train the habit of touching the row and column headers before lifting a number, every time.

Panic cascades. One bad question bleeds into the next three. The recovery skill is deliberate: skip, breathe once, next question clean. A skipped question costs one mark; a rattled candidate loses five.

A two-week preparation plan

Days 1 to 3: rebuild the toolkit. Percentages, ratios, weighted averages, currency conversion, with no clock. Speed built on shaky method collapses in the test.

Days 4 to 10: timed sets, one or two a day. Short, frequent, always timed. After each set, review every error and classify it: method, reading, or arithmetic. The classification tells you what to fix; most candidates discover their errors are 70 percent reading.

Days 11 to 14: full tests in exam conditions. Same time of day as your booking if you can, phone in another room, one sitting. This trains the endurance and the pacing, not just the maths.

Aim your practice at the publisher your employer uses if you can find out (the invitation email usually names SHL, Talent Q, Cubiks or similar); the question grammar differs slightly and familiarity is worth marks.

On the day

Do a warm-up set an hour before, so the first live question is not your first question of the day. Check the rules: negative marking is rare, so unanswered beats unattempted only when guessing is penalised, and it almost never is. If the test is unsupervised, resist the temptation to get help: most employers retest a sample at the assessment centre, and a score you cannot repeat is worse than a modest honest one.

Numerical tests are the least personal stage of the whole pipeline: no judgement, no chemistry, just preparation made visible. That makes them the stage where two weeks of deliberate work moves your outcome the most.

Once you are through, the pipeline turns human: video interviews, competency questions and assessment days, where our practice interviews do for your answers what the timed sets did for your arithmetic. And if verbal reasoning is on your list too, the companion guide is here.

Ready to put this into practice?

Answer one real interview question out loud and see how you would score, with a stronger answer to learn from. No sign-up needed.

Keep preparing