Free Numerical Reasoning Test

What a numerical reasoning test actually measures

A numerical reasoning test is not a maths exam. It does not ask you to recall a formula, to factorise anything, or to remember what you were taught at sixteen. It asks a narrower and more practical question: when you are handed a set of numbers you have never seen before, how quickly and how reliably do you extract the meaning from them? That is why the arithmetic on this page is deliberately gentle. Every division works out, every percentage lands on a sensible figure, and no question needs more than two lines of working. The difficulty is in seeing what to calculate, not in the calculating.

The skill matters because most working decisions arrive in exactly this shape. A quarterly revenue table, a page of sales figures, a spreadsheet with four regions and four columns — the person who can look at it and say "the east grew twenty-five per cent while the south shrank" within a few seconds is genuinely more useful than the person who needs ten minutes and a colleague to confirm it. Employers test for it because it is one of the few desk skills that predicts on-the-job performance without needing any prior knowledge of the industry.

The two question families on this test

Ten of the eighteen items are number series. You are given five or six values and asked for the next one. The underlying rules vary on purpose: constant multipliers with an offset (each term doubled and then reduced by one), second-order differences (the gaps themselves follow a pattern), interleaved sequences where two independent series alternate position by position, factorial-style growth where the multiplier increases each step, and additive rules where each term is the sum of the two before it. Because the families are mixed, no single trick carries you through; you have to diagnose each series before you solve it.

The other eight items are data interpretation questions built on two small tables — one of quarterly revenue by region, one of a single day's trading at four stores. These are the questions that most closely resemble a graduate assessment. They ask for percentage change, whole-year totals, the ratio between two figures, one item's share of a combined total, and which row moved most in absolute terms. Several of them deliberately reward reading the question carefully: a row with the largest percentage increase is not always the row with the largest absolute increase, and a store with the highest unit sales is not always the store with the highest revenue.

The three routines that decide your score

Almost every data question on a test like this reduces to one of three small routines, and the people who score well are simply the people for whom those routines are automatic. The first is percentage change: the difference divided by the original value, never by the new one. Going from 120 to 210 is a rise of 90 on a base of 120, which is 75 per cent — a mistake here costs you the whole question even when the reading was right. The second is ratio reduction: find the largest factor common to both numbers and divide it out, so 495 to 660 becomes 3 to 4. The third is share of a total: add the parts once, carefully, then divide. Most of the errors we see on this format are addition slips in that first step, not conceptual failures.

How your score is calculated

Your raw number of correct answers is converted to the familiar 100-centred scale, where 15 points is one standard deviation. Eighteen items is a short test, so treat the number as a band of roughly ten points rather than an exact value. The table below is the standard reading.

How to work through it under the clock

• One minute per question is the budget. If a series has not revealed itself in forty seconds, guess from what you can eliminate and move on — the next item may be free.
• On a series, check the differences first. If the differences are irregular, check the ratios. Only after both fail should you look for an interleaved or additive rule.
• On a table question, read the row and column labels before the numbers. Half of all errors on this format come from answering about the right value in the wrong quarter.
• Write the intermediate figure down. Holding a four-digit total in your head while you divide is where the arithmetic slips happen.
• Answer every question. There is no penalty for a wrong answer, so a blank is strictly worse than a guess between two remaining options.

How this differs from a general IQ test

A general intelligence test samples several abilities — verbal, spatial, logical and numerical — and averages them, so a weakness in one area is partly hidden by strength in another. This test does the opposite: it isolates the numerical component and pushes on it for eighteen minutes. That makes the result narrower but more diagnostic. If your score here sits well below your score on a mixed test, the gap is almost always a fluency problem rather than a reasoning one, and fluency is trainable in a way that raw reasoning speed is not. If it sits well above, you have found the kind of work you should probably be applying for.

Practising without wasting your time

The least useful form of practice is retaking the same test until you remember the answers; your score will climb and your ability will not. The most useful form is slower and duller. Take one real data table — a quarterly report, a league table, an energy bill — and force yourself to state three things about it out loud within thirty seconds: which figure grew fastest in percentage terms, which grew most in absolute terms, and what share the largest item holds of the total. Do that for a fortnight and the routines in the section above stop consuming any working memory at all, which is exactly what a timed assessment rewards.