A numerical reasoning test is a timed, multiple-choice assessment that measures how quickly and accurately you can interpret numerical data (tables, charts, graphs and financial figures) and draw the correct conclusion from it. It is not a maths exam. The arithmetic rarely goes beyond percentages, ratios, averages and simple algebra, and most tests allow a calculator. What is really being measured is data handling under time pressure: typically 45 to 75 seconds per question. Employers use it as a proxy for how well you will reason with numbers on the job. The skill is trainable, which is why structured practice moves scores.
Find out which publisher your employer uses, practise that exact format under a timer, memorise six formulas (percentage change, percentage of total, ratio, mean, margin, conversion), read the question before the data, estimate before you calculate, never spend more than about 75 seconds on a question, and answer every single item because there is normally no negative marking.
What Is a Numerical Reasoning Test?
You are shown a data set (a sales table, a line graph of quarterly revenue, a pie chart of market share, a currency conversion table) and asked a question with four or five answer options. You must extract the right figures, apply a basic calculation, and pick the option that matches. The data set usually stays on screen for several questions, so the first question against a new table costs you the most time and the following ones cost the least.
Numerical reasoning sits inside a broader cognitive battery. Most employers pair it with a verbal reasoning test and a logical or inductive reasoning test, and increasingly with a situational judgement test. Scores are almost always reported as a percentile against a norm group (other graduates, other managers, other applicants for the same role) not as a raw percentage. Scoring 70% correct can put you in the 60th percentile in one norm group and the 85th in another, which is why raw marks alone tell you very little.
What the Test Actually Measures
- Data extraction: can you find the right row, column or data series without misreading the axis or the units?
- Basic numeracy under pressure: percentages, ratios, averages, indices, rates of change. Nothing beyond secondary-school level.
- Attention to detail: units in thousands versus millions, values in a footnote, a year that is missing from the chart.
- Speed-accuracy trade-off: whether you keep your accuracy when the clock is visible. This is the single biggest differentiator between candidates.
- Resilience to distractors: wrong answer options are engineered from the most likely mistakes, so the answer that 'feels right' is often the trap.
Most numerical reasoning tests allow a calculator, and on-screen calculators are standard. The challenge is never the maths itself. It is finding the right numbers fast and not falling for engineered distractors.
The 6 Question Types You Will Meet
- Table interpretation: multi-column tables where you compare totals, trends or subsets. The most common type by a distance.
- Graph and chart analysis: bar charts, line graphs, stacked columns and pie charts. Read the axis labels and the legend before anything else.
- Percentage calculations: percentage change, percentage of total, percentage point differences. Expect these in a large share of questions.
- Ratio and proportion: splitting a total in a given ratio, or comparing two ratios across periods.
- Rates and conversions: currencies, units, per-hour and per-employee figures, and time-period conversions such as monthly to annual.
- Estimation: questions designed so that the exact calculation is slower than a sensible approximation, and the answer options are far enough apart that approximation is enough.
Essential Formulas to Memorise
These six cover the overwhelming majority of what a numerical reasoning test asks. Memorise them until you never have to reconstruct them mid-question.
| Formula | Expression |
|---|---|
| Percentage change | ((New - Old) / Old) x 100 |
| Percentage of total | (Part / Total) x 100 |
| Reverse percentage (find the original) | New / (1 + rate) |
| Ratio split | Total x (Share / Sum of shares) |
| Average (mean) | Sum / Count |
| Profit margin | (Profit / Revenue) x 100 |
| Currency or unit conversion | Amount x Rate |
| Compound growth over n periods | Start x (1 + rate)^n |
Mental-Math Shortcuts That Save Seconds
Every second you do not spend typing into a calculator is a second you spend reading the next question. These shortcuts are the ones worth drilling until they are automatic.
| Formula | Expression |
|---|---|
| 10% of a number | Move the decimal one place left |
| 5% of a number | Take 10%, then halve it |
| 15% of a number | 10% + 5% |
| 1% of a number | Move the decimal two places left |
| Percentage increase of one third | Roughly +33%; two thirds is roughly +67% |
| Doubling check | An increase from 40 to 80 is +100%, not +80% |
| Approximate division | Round both numbers to one significant figure first, then refine |
| Index numbers | A rise from index 100 to 118 is exactly +18% |
Sit a timed numerical reasoning test built on the same formats employers use. The free preview gives you 4 questions per test plus 5 AI coach messages, no payment details needed.
Start the free test9 Strategies That Move Your Score
- 1Read the question before the data. Knowing what you are looking for turns a 20-second scan of a table into a 5-second lookup. Never read a data set 'to understand it' first.
- 2Check the units before you calculate. Values in thousands, a chart axis that starts at 40 rather than 0, a footnote that excludes one region: this is where most careless errors come from.
- 3Estimate first, calculate second. If the options are 12%, 28%, 44% and 61%, a rough estimate identifies the answer without a single keystroke. Only calculate precisely when the options are close together.
- 4Use the answer options as information. Options are engineered from predictable mistakes. If your answer is exactly ten times one of the options, you have a units error, not a calculation error.
- 5Do the second and third question against the same data set immediately. The reading cost is already paid. Skipping back and forth between data sets is the most expensive habit in the test.
- 6Practise with the calculator you will use. If the test provides an on-screen calculator, practise with an on-screen calculator. Muscle memory for a physical keypad does not transfer.
- 7Set a hard per-question ceiling. Around 75 seconds. When you hit it, guess, flag it and move on. A question you spend three minutes on costs you two other questions.
- 8Answer every question. Negative marking is rare in graduate-level numerical reasoning. A blank is worth nothing; a guess between four options is worth a 25% chance.
- 9Review errors by category, not one by one. Sort every mistake into misread data, wrong formula, calculation slip, or ran out of time. Whichever bucket is biggest is your entire training plan for the next week.
Timing: How to Manage the Clock
Time management is the most trainable part of the test and the part most candidates neglect. Work out your per-question budget before you start: divide the total time by the number of questions, and subtract ten percent as a buffer.
- SHL-style test (18 questions, 25 minutes): about 83 seconds per question. Comfortable. Accuracy should be your priority.
- Korn Ferry-style test (20 questions, 20 minutes): about 60 seconds per question. You cannot afford to reread a table twice.
- cut-e scales-style test (roughly 15 seconds per item): pure speed. Reading the whole data set is impossible by design; you jump straight to the tab that holds the number you need.
- Last two minutes: stop solving. Fill in every remaining option, including the flagged ones. An unanswered question scores zero with certainty.
Publisher Differences: What to Expect
The single highest-return preparation step is finding out which publisher your employer uses and practising that format. A cut-e test and an SHL test both call themselves numerical reasoning, and they demand almost opposite strategies. Your invitation email normally names the platform.
| Publisher | Questions | Time | Difficulty | Notes |
|---|---|---|---|---|
| SHL Verify | 18 questions | 25 minutes | Medium-Hard | Data-heavy tables, several questions per data set, calculator allowed |
| Korn Ferry | 20 questions | 20 minutes | Hard | Adaptive, very time-pressured, used heavily for graduate and management roles |
| Talent Q Elements | 12 questions | Adaptive | Adaptive | Difficulty adjusts to your answers; accuracy matters more than speed |
| Cubiks Logiks | 20 questions | 25 minutes | Medium | Mix of data interpretation and straightforward arithmetic |
| Aon cut-e scales | Tabbed data sets | About 15 seconds per item | Very Hard | Extreme time pressure by design; data is split across tabs so you never read it all |
| Saville Swift | Combined battery | Roughly 6-10 min per section | Medium-Hard | Numerical, verbal and diagrammatic in one sitting |
Default to SHL-style practice (it is the most widely used format and the closest to a neutral middle ground) then add one session of extreme-speed practice so a cut-e-style test cannot shock you on the day.
A Worked Example, Step by Step
A table shows regional revenue for a company. Europe: 2024 revenue EUR 4.2m, 2025 revenue EUR 5.1m. Asia: 2024 revenue EUR 3.6m, 2025 revenue EUR 3.9m. The question: which region grew faster, and by how many percentage points did the growth rates differ?
- 1Read the question first. You need two growth rates and their difference. You do not need any other row in the table, so you ignore everything else.
- 2Apply percentage change to Europe. (5.1 - 4.2) / 4.2 = 0.9 / 4.2 = 21.4%.
- 3Apply it to Asia. (3.9 - 3.6) / 3.6 = 0.3 / 3.6 = 8.3%.
- 4Subtract. 21.4 - 8.3 = 13.1 percentage points. Europe grew faster.
- 5Sanity-check against the distractors. An option of 2.6 would come from comparing absolute growth (0.9 versus 0.3) instead of rates. An option of 13.1% (rather than percentage points) tests whether you know the difference between a percentage and a percentage point. Both are the engineered traps.
Your 14-Day Preparation Plan
- 1Days 1-2: Baseline. Sit one full timed test without preparing. Record your score, your time per question, and which of the four error categories dominates.
- 2Days 3-5: Formulas and shortcuts. Drill the two tables above until every formula is automatic. Do untimed question sets so you build correctness before speed.
- 3Days 6-8: Format-specific practice. Switch to the publisher your employer uses. Practise timed, with the same calculator type you will get on the day.
- 4Days 9-11: Speed layer. Cut your per-question budget by 15% deliberately. Working slightly faster than the real test makes the real test feel slow.
- 5Days 12-13: Error review only. Redo only the questions you got wrong, and write one sentence per question explaining the actual cause.
- 6Day 14: Full rehearsal, then stop. One complete timed test in the morning, then rest. Cramming the night before reliably lowers scores.
MakingMoves.ai covers 50+ test categories with 113,000+ practice questions, and the AI coach explains every answer step by step rather than just marking it wrong. The free preview gives you 4 questions per test and 5 AI coach messages. Paid plans are GROW at EUR 19,95 per week, PRO at EUR 49,95 per month, and MAX at EUR 79,95 per month.
Mistakes That Cost the Most Marks
- Confusing percentage and percentage point. A rise from 20% to 25% is a 5 percentage point rise and a 25% increase. Test writers love this.
- Ignoring the units row. 'Figures in thousands' at the top of a table changes every answer in it.
- Averaging averages. The mean of two regional averages is not the overall average unless the regions are the same size.
- Reading a truncated axis as if it started at zero. A bar that looks twice as tall may represent a 6% difference.
- Chasing one hard question. Every extra minute on a question you will probably get wrong costs you two you would have got right.
- Practising a format you will not sit. Generic practice builds the foundations, but format-specific practice is what changes your percentile. Pair numerical work with verbal reasoning and logical reasoning if your employer uses a full battery.
Our numerical reasoning category mirrors SHL, Korn Ferry, Talent Q, Cubiks and Aon cut-e formats, with AI explanations on every question and timed mock tests that match the real durations.
Explore numerical reasoning practiceFrequently Asked Questions
Scores are reported as percentiles against a norm group, not as raw percentages. Competitive graduate employers typically shortlist from the upper part of their norm group, so aim to be comfortably above the median rather than to hit a specific raw mark. Employers rarely publish their cut scores, and they differ by role and intake.
In most graduate-level tests, yes, usually an on-screen calculator, sometimes your own basic calculator. Speed tests such as Aon cut-e scales are deliberately built so that a calculator gives you little advantage. Always check the instructions on the invitation email.
Two focused weeks is enough for most candidates: a few days on formulas, a week on format-specific timed practice, and a final rehearsal. Spaced daily sessions beat one long cramming session, because the bottleneck is retrieval speed, not knowledge.
The maths is not hard. It is secondary-school arithmetic. The difficulty comes from the clock and from answer options engineered around common mistakes. Candidates who fail usually run out of time, not out of ability.
That depends on the employer. Many allow a retake after a waiting period, typically six to twelve months, and some do not allow one at all within the same application cycle. The invitation email or the employer's FAQ is the only reliable source.
Yes, usually as a sift. At high-volume graduate employers the online tests are the first hard filter after the application form, which is why the test is the cheapest stage to prepare for and the most expensive one to fail.
Read our assessment tips hub for timing drills, test-day checklists and preparation plans across every test category.
Read the assessment tipsIngmar van Maurik is the founder of MakingMoves.ai and Assessment-Training.com. With 10+ years in psychometric assessment design, he has helped over 1 million professionals prepare for job assessments.