Numerical reasoning

What the framework is

Numerical reasoning tests measure whether someone can pull the right figures out of data and reason with them correctly under mild time pressure. They are deliberately not maths exams: the arithmetic stays at percentages, ratios, rates and simple trends. What is being tested is reading the data properly and choosing the right operation.

Tests of this kind are among the better predictors of performance in analytical and commercial roles, because most workplace numeracy failures are interpretation failures, not calculation failures.

How the questions are built

Each question presents a small data set inline — a table of sales by quarter, a set of costs, a simple series — and asks a single question with four answer options and exactly one correct answer.

Distractor options are chosen to catch specific mistakes: using the wrong row, computing change instead of percentage change, or inverting a ratio. That is why a wrong answer can be informative — it often shows *which* step went astray.

Questions are tagged easy, medium or hard, and a pack draws a set number of them so every candidate faces a comparable test.

Timing

The section carries an optional time limit set on the pack. A limit adds mild pressure and keeps candidates comparable; without one, the score reflects accuracy alone.

A rough guide is 60 to 90 seconds per question. Anything tighter starts to measure reading speed rather than reasoning.

How to interpret a result

The score is the percentage of questions answered correctly, with the number correct shown alongside. On a short test, each question moves the score several points, so treat 70% and 80% as broadly the same result.

Read the score together with the number of questions. Six out of eight is a much rougher measurement than twenty-four out of thirty.

Look at the per-question breakdown when a score is mid-range. Consistent near-misses on percentage change point to one fixable gap; scattered errors suggest rushing.

Compare against the demands of the job, not against other candidates. Most roles need competence, not a top score.

Common mistakes

Using a numerical score as a ranking device between candidates who are within a few points of each other.

Ignoring context: unfamiliarity with the test format, language, or plain nerves all depress scores in ways unrelated to the job.

Testing numeracy for a role that does not need it. If the job never involves interpreting figures, the section is noise.

Never following up. Asking the candidate to talk through one question they got wrong is often more revealing than the score.