You applied to a graduate scheme, the recruiter sent a friendly email, and then a link landed in your inbox: an SHL numerical reasoning test, to be completed in the next few days. You clicked through to a practice page, saw a bar chart and a countdown timer, and felt your stomach drop. That is roughly where I was a while ago, and I remember thinking the same thing you probably are: how hard can reading a chart really be?
Harder than it looks. Not because the maths is difficult, but because of the clock, the data density, and the way the wrong answers are built to catch you. I took the test, and this is the honest account I wish I had read first: what the questions actually feel like under time pressure, what the score really means, and how I would prepare if I had to sit it again tomorrow.
Here is the thing almost nobody tells you upfront. The SHL numerical reasoning test is not testing whether you can do maths. It is testing whether you can read data fast, under pressure, without making a careless slip.
The underlying arithmetic is GCSE level. Percentages, ratios, fractions, a currency conversion or two. If you sat and did those sums with no clock, you would score close to perfect. The difficulty is entirely manufactured by three things working together: a tight timer, tables and charts crammed with numbers you have to locate first, and answer options designed around the mistakes people typically make. Miss that framing and you will prep the wrong thing.
What the questions actually feel like
Each question puts a data set in front of you: a pie chart, a bar chart, a line chart, a table of figures, sometimes a small paragraph of context. Then it asks something like “what was the percentage increase in revenue from Q2 to Q3 for the European division?”
Simple, until you notice the table has six divisions, four quarters, two currencies, and a footnote saying the figures are in thousands. Now you are not doing maths, you are hunting. You have to find the European row, find Q2 and Q3, confirm the units, and only then do the calculation. The hunting is the hard part, and it is where the seconds vanish.
A few things that surprised me the first time:
- Every question is independent. The data set changes each time, so there is no momentum to build. You reset and re-read from scratch, over and over.
- The charts are deliberately busy. Extra series, extra categories, a legend you have to cross-reference. The relevant number is in there, but it is surrounded by numbers meant to slow you down.
- The wrong answers are not random. If the correct answer is a 12 percent increase, one option will be what you get if you divide by the wrong base value, another will be the decrease instead of the increase, another will be the raw difference instead of the percentage. Every distractor is a mistake someone has actually made. That is what makes it feel unfair: you can do the calculation correctly and still tick a wrong box because you read one figure off the wrong row.
The test does not punish you for being bad at maths. It punishes you for reading the chart one row too fast.
A worked example, the way it actually plays out
Let me make this concrete, because the abstract description undersells how the trap springs. Picture a table titled “Regional revenue (in thousands of GBP)” with rows for six divisions and columns for Q1 through Q4. The European row reads 420, 480, 540, 510. The question: “By what percentage did European revenue rise from Q2 to Q3?”
The maths is trivial: (540 minus 480) divided by 480, which is 60 divided by 480, which is 12.5 percent. Ten seconds with a calculator. But look at the answer options you are actually given: 12.5 percent, 11.1 percent, 6.25 percent, and 60. Every wrong option is a real mistake waiting for you. The 11.1 percent is what you get if you divide by 540 instead of 480 (wrong base). The 6.25 percent is the Q2-to-Q4 jump, (510 minus 480) divided by 480, which is what you calculate if your eye slid one column too far and read Q4 instead of Q3. The bare 60 is the raw difference, forgetting the “percentage” in the question. And the whole thing is priced in thousands, so if the follow-up asks for an absolute figure you have to remember to multiply back up.
I got a question shaped almost exactly like this wrong on a practice run, not because I could not do the sum, but because I divided by the end value. That is the entire test in one question: correct arithmetic, wrong reading, wrong box.
If you want to feel this before test day, the most useful thing I did was work through timed, section-by-section examples rather than a single long mock. This section-by-section SHL walkthrough breaks each reasoning type into worked questions with the answer logic explained, which is exactly the kind of “why is this the right answer” review that fixed my careless errors faster than raw repetition did. The paid practice platforms are good at giving you a high volume of questions to grind through, and that has its place, but volume alone never taught me why I was picking the wrong option. Understanding the distractor logic did.
The formats and timing, so nothing surprises you
Which version you get depends on the role and the employer, but these are the common ones you will run into. It is worth knowing which you are facing, because the pacing math is different for each.
| Version | Questions | Time limit | Roughly per question |
|---|---|---|---|
| SHL Verify Interactive (numerical) | up to 10 | 18 minutes | ~1 min 48 sec |
| SHL Verify (numerical reasoning) | up to 18 | 25 minutes | ~1 min 23 sec |
| SHL Verify (numerical ability) | up to 16 | 20 minutes | ~1 min 15 sec |
The interactive version uses a drag-and-drop style instead of plain multiple choice, and it is gradually replacing the older format, so it is the one most people now take. The difficulty is the same either way; only the way you input the answer changes.
Just over a minute per question sounds fine until you realize half of it is spent finding the right numbers in the chart. That was the single biggest gap between how I imagined the test and how it actually went.
Most online versions give you a basic on-screen calculator. Some supervised in-person versions do not, so always read the instructions on the intro screen. Even with a calculator, reaching for it on every sub-step slows you down, so a bit of mental arithmetic still pays off.
What your score actually means (it is not raw correct answers)
This is the part I understood only afterwards, and it changes how you should think about the whole thing.
SHL uses norm-referenced scoring. Your raw number of correct answers is not what the employer sees. Instead, your performance is compared against a norm group of previous test-takers, and your result is reported as a percentile or a standardized score. Score in the 70th percentile and it means you did better than 70 percent of that comparison group, not that you got 70 percent of the questions right.
Employers then set their own cutoff, and these vary a lot by firm and role, so treat any specific number as a rough guide rather than a rule. From what I saw cited while I was prepping, a general commercial role might accept somewhere around the 50th percentile, while graduate consulting or investment-banking streams are commonly said to want the 70th to 90th. Because the bar is relative, speed and accuracy both matter: everyone is being ranked against everyone else, so leaving questions blank or rushing into careless errors both drag your percentile down.
Two practical notes that genuinely helped my nerves: there is no negative marking for a wrong answer on SHL tests, so a considered guess on a question you cannot finish is strictly better than a blank. And you generally want to complete the whole test, since not finishing can count against you. If a question is eating your clock, estimate, tick, and move.
The reasoning skills SHL measures here overlap heavily with the other big cognitive-aptitude tests, which is why so much of the same practice transfers. I had already read up on how these assessments work across tech hiring in this look at cognitive aptitude tests beyond the CCAT, and the mental model carried over almost one-to-one to the SHL numerical section.
How I would prep if I had to sit it again tomorrow
I did not prep well the first time. I treated it like a maths refresher, which was the wrong instinct. If I were doing it again, this is the order I would work in.
- Fix your data-reading before your arithmetic. Do a handful of questions with no timer at all, but force yourself to name the units and the exact row or column before you touch the calculator. Read the chart title first, always: it tells you what the data measures and in what units before you look at a single number. Most of my early mistakes were unit slips, not sum slips.
- Estimate before you calculate. Before hitting the calculator, guess the answer’s rough range in your head. It takes three to five seconds and it catches the single most common error: a calculator input mistake that spits out a wildly wrong number. If your calculated answer is nowhere near your estimate, you keyed something wrong.
- Then add the clock. Once your reading is clean, practise timed, so the pressure stops being a surprise on the day. A drill that isolates the numerical question type is far more useful than one long mixed mock; this numerical reasoning drill guide has worked examples and a plan built around exactly that, and it flags traps like confusing “percentage change” (divide by the old value) with “percentage of the total” (divide by the whole), which is a mistake that quietly costs people several questions per test.
- Review why, not just whether. For every question you get wrong, work out which distractor you fell for and why. That review loop, not raw repetition, is what moved my accuracy.
If you want a broader sense of how tough these timed cognitive tests are before you commit hours to prep, I wrote up my honest take on the difficulty of a related test in how hard the CCAT really is, and most of that reality check applies to SHL too.
FAQ
How hard is the SHL numerical reasoning test really?
The maths is GCSE level, so the calculations themselves are easy. The difficulty is the time pressure plus data-heavy charts plus answer options engineered around common mistakes. Most people who struggle are not bad at maths; they are reading the data too fast and slipping on units or rows.
Do I need advanced maths to pass?
No. Percentages, ratios, fractions, and basic conversions cover almost everything. Speed at reading tables and charts matters far more than mathematical ability.
Can I use a calculator?
Usually yes on the online version, where a basic on-screen calculator is provided. Some supervised in-person sittings do not allow one, so check the instructions on the intro screen. Either way, estimate first and use the calculator to confirm, not to think.
What is a good SHL percentile?
It depends entirely on the employer’s cutoff, so any figure is a rough guide, not a fixed pass mark. Many graduate roles are commonly said to look for around the 60th to 70th percentile, while top-tier consulting and banking streams reportedly want the 80th to 90th. Because scoring is norm-referenced, you are ranked against other candidates rather than against a set score.
Should I guess if I run out of time?
Yes. There is no negative marking for wrong answers, and completing the test matters, so a reasoned guess beats a blank every time. If a question is draining your clock, estimate, answer, and move on.
What I would actually do next
Here is the honest summary. I went in thinking the SHL numerical reasoning test was a maths check and got humbled by a data-reading speed test. Once I understood that the real skill is locating the right numbers fast and not falling for engineered distractors, everything about how to prepare changed. Read the chart title first, confirm the units, estimate before you calculate, practise timed by section, and never leave a blank.
I build software for a living now, not test-prep, but I keep an interest in this space because I genuinely took these tests and remember how opaque they felt. When people ask me where to practise, I point them at PrepClubs, because it does the thing that actually moved my scores: timed drills by question type with the answer logic explained, rather than another wall of untimed questions. Start with the free material to find your weak spot, then decide whether systematic paid practice is worth it for the role you are chasing. That “free first, then pay only if it earns it” order is how I would approach any of these assessments, and it is the honest advice I would give a friend the night before their test.