TD
Key Takeaways
- Tests usually sit immediately after the application form and before any human review, so they are a genuine filter.
- Numerical reasoning is the most trainable and the one candidates most often underestimate.
- Situational judgement tests measure fit against the employer's stated values, not general common sense.
- Gamified assessments cannot be revised for in the usual sense, but familiarity with the format removes a real disadvantage.
- Practising under proper time pressure is what improves scores; untimed practice builds the wrong skill.
Where tests sit in the process
For most UK graduate schemes the sequence is: online application form, then tests, then video interview, then assessment centre.
The tests are automated and often the largest single cut in the process. A large share of applicants are removed here, before any person has read a word of the application they spent an hour on.
That makes preparation for this stage unusually high-value. It is also the stage candidates most often skip preparing for, because it feels like an administrative formality rather than an assessment.
Numerical reasoning
Data in tables and charts, questions requiring percentages, ratios, currency conversion, growth rates and simple interpretation. Time is tight — often around a minute per question.
The maths itself is not advanced. The difficulty is speed under pressure with unfamiliar data layouts.
What actually improves your score:
Practise with a timer from the start. Untimed practice builds accuracy you already have and not the speed you need.
Know the standard calculations cold. Percentage change, percentage of a total, ratio comparisons, reverse percentages. These recur constantly and hesitation on them is what costs you questions.
Read the question before the table. Tables are deliberately dense and contain information you do not need. Knowing what you are looking for prevents wasted scanning.
Watch the units. Thousands versus millions, and figures already expressed as percentages, are the most common trap.
Use the calculator you are allowed. Most tests permit one. Practise with the same kind you will use.
This is the most trainable of the tests. Candidates who practise for a few hours routinely improve their results.
Verbal reasoning
A passage and statements to judge as true, false or cannot say. The passages are usually dry business or policy text.
The recurring failure is answering from world knowledge rather than the passage. "Cannot say" means the passage does not contain enough information to decide, even if you personally know the answer.
Practical approach: read the statement first, then scan for the relevant part of the passage, and answer strictly on what is written. If you are inferring, the answer is usually "cannot say".
Situational judgement tests
A workplace scenario and several possible responses, which you rank or rate.
The important thing to understand: these are scored against the employer's competency framework, not against universal common sense. Two employers can score the same answer differently because they value different things.
So preparation is research rather than practice. If the firm emphasises client focus, answers prioritising the client score well. If it emphasises collaboration, answers involving consulting colleagues score well. Many employers publish their values or graduate competencies, and reading them before the test is the single most useful preparation available.
Two general patterns hold almost everywhere: escalating immediately to a manager without attempting anything yourself scores poorly, and so does acting unilaterally on something significant without consulting anyone. The middle path — attempt, involve appropriate people, escalate if genuinely warranted — is usually the intended answer.
Answer as the person the employer describes wanting, not as your most cynical self.
Gamified assessments
Increasingly common, particularly at large graduate employers, and genuinely different from the traditional tests.
What they are. Short interactive tasks — inflating a balloon for points, matching patterns, reacting to prompts, memory exercises. Ten to thirty minutes, and they measure things like risk appetite, processing speed, attention and learning rate.
Can you revise for them? Not in the usual sense. There is no syllabus and no technique to learn.
Familiarity still matters. A candidate meeting the format for the first time performs worse than one who has seen it, purely because the first two minutes are spent understanding the interface rather than playing. Most providers offer practice versions; doing one removes that penalty entirely.
Read the instructions properly. Sounds trivial, and these games have specific rules — whether points bank automatically, whether a wrong answer costs you — and misunderstanding one wastes the whole exercise.
Do them alert and undistracted. Reaction and attention are literally being measured. Doing these at midnight after a long day depresses the score for no good reason.
Do not try to fake a personality. Risk-appetite games in particular are hard to game, and inconsistent play patterns are visible.
Practical conditions
- Do them properly. A quiet room, a full battery, a stable connection, no interruptions.
- Have paper and a calculator. For numerical tests, working things out on screen is slower.
- Do not let someone else take them. Many employers re-test at the assessment centre under supervision, and a large discrepancy is noticed.
- Watch the expiry. Test invitations often expire within days and land in email that is easy to miss during a busy application season.
- Check whether scores carry over. Some employers share test results across applications within a period; others do not.
How this fits the wider application
Worth putting in proportion, because candidates either dismiss these or over-invest in them.
The UK graduate process is a funnel with four stages, and each removes people for a different reason. The application form removes generic answers. The tests remove people below a threshold, mechanically. The video interview removes people who could not deliver evidence to a camera. The assessment centre decides among those who remain.
Tests are the only stage where the outcome is a number and there is nothing to argue with. That makes them uniquely frustrating and uniquely fixable — unlike every other stage, there is no judgement involved, and three weeks of the right practice reliably moves you above the line.
It also means they deserve exactly the effort required to clear the threshold and not a minute more. Your test score does not appear at the assessment centre. Nobody at the final interview knows what you got. It is a door, and once you are through it the only thing that carries forward is the evidence you built for the other three stages.
Numerical reasoning, in detail
The most trainable of the tests and the one candidates most often underestimate, on the assumption that arithmetic is arithmetic.
What it actually is. A table or chart, a question, four or five options, and roughly sixty to ninety seconds. The mathematics is GCSE level. The difficulty is entirely the time and the format.
The five calculations that cover most questions:
Percentage change. New minus old, divided by old. Asked constantly, and candidates lose marks by dividing by the wrong figure.
Percentage of a total. Part over whole. Watch whether the "whole" is the column total or the grand total — this is the most common trap in the entire test.
Ratios. Usually simplification or applying a ratio to a total.
Reverse percentages. "After a 20% increase the figure is 96 — what was it before?" Divide by 1.2, do not subtract 20%. Candidates get this wrong under pressure with great reliability.
Compound growth over two or three periods. Multiply the factors; do not add the percentages.
Four habits that raise scores immediately:
Read the question before the table. You will otherwise absorb data you do not need. This single change saves fifteen seconds per question, which across twenty questions is enormous.
Watch the units. Thousands versus millions, and figures already expressed as percentages. Tables are deliberately labelled to catch this, and a units error produces an answer that matches one of the wrong options — which is not an accident.
Estimate first, then check against the options. The options are usually far enough apart that a rough calculation identifies the right one, and this is faster than computing precisely.
Move on. Most of these tests reward answering more questions rather than perfecting a few. Spending three minutes on a hard question costs you two easy ones.
Use the calculator you are permitted, and practise with that one. Fumbling an unfamiliar calculator is a real and avoidable loss.
Verbal reasoning, and the trap in it
A passage and a statement, and you choose True, False, or Cannot Say. It looks easier than the numerical test and candidates score worse on it.
The entire difficulty is "Cannot Say". You must judge the statement only against the passage — not against what you know, not against what is obviously true in the world.
If the passage says a company opened offices in three cities, and the statement says the company is expanding, the answer is very often Cannot Say, because the passage did not say why the offices opened. Applying real-world reasoning here is exactly the error being tested for.
The mechanical fix: for every statement, ask "can I point to the specific sentence that establishes this?" If you cannot point at it, it is Cannot Say. Your knowledge, inference and common sense are all out of bounds.
Watch the qualifiers. "All", "some", "may", "always", "the majority". A statement saying "all" when the passage says "most" is False, not True.
Read the statement first, then scan the passage for the relevant part.
Reading the whole passage carefully first is too slow at these time limits.
Situational judgement, and what it is really measuring
The most misunderstood test, because candidates try to guess what sounds impressive.
What it is. A workplace scenario, then several possible responses to rank or rate. There is no arithmetic and no verbal trick.
What it measures. Fit against this employer's stated values, calibrated against how effective employees at that organisation actually behaved. It is not a general test of common sense, which is why the same answer can score well at one employer and poorly at another.
Which makes the preparation obvious and rarely done: read the employer's published values and competency framework before sitting the test. If a firm emphasises collaboration and client focus, responses that consult colleagues and prioritise the client score better than decisive independent action. If a firm emphasises ownership and pace, the weighting shifts.
Patterns that generally score well: gathering information before acting; consulting the right person rather than escalating over someone's head or going it alone; addressing a problem directly with the person involved before raising it upward; being honest about a mistake early.
Patterns that generally score badly: doing nothing and hoping; going straight to a senior person over your manager's head; acting unilaterally on incomplete information; and any response that avoids the difficulty entirely.
Answer as your best professional self, not your actual self and not a fictional hero.
Candidates who try to game these usually score worse, because the scoring is derived from real behavioural data rather than from what sounds good.
Where tests sit, and why that matters
Understanding the position of these tests in the process explains why they deserve real preparation and also why they do not deserve more than that.
They come immediately after the application form, usually before any human has read a word you wrote. At the largest graduate employers, this is the first genuine cut, and the volumes are such that a substantial share of applicants are removed here.
The threshold is usually a fixed score, not a ranking against that year's cohort — though some employers do band by percentile. Either way, no human is weighing your test result against your excellent written answers. Below the line, the application ends.
Which is why "I'm bad at tests" is an expensive belief. Numerical and verbal reasoning are trainable within a few weeks, and the training is mostly about speed rather than mathematics. Candidates who accept a weakness here lose access to a large part of the graduate market for a fixable reason.
And why over-preparing is also a mistake. These are a gate, not a differentiator. Nobody gets an offer for a good test score. Once you are consistently clearing the time limit, further practice has almost no return, and the same hours spent on the competency stories or the assessment centre affect outcomes far more.
Retakes are usually not permitted within a cycle, and some employers share results across their own multiple applications. Some also allow a retake the following year. Treat the first attempt as the real one — practise before you open the link, rather than treating an employer's live test as practice.
Accessibility and adjustments
Rarely mentioned and worth knowing about.
Adjustments are available and normal. Extra time, alternative formats, screen-reader compatible versions, or a different assessment entirely. Dyslexia, ADHD, visual impairments and a range of other conditions all qualify at most large employers.
You have to ask, and you have to ask early. Adjustments are arranged before the test is issued, not applied retrospectively to a poor score.
Ask the recruitment team directly. A short email stating what you need is sufficient; you are generally not required to disclose a diagnosis in detail, though some employers ask for evidence.
It does not disadvantage your application. Adjustment requests are handled separately from assessment and are not visible to the people scoring you at later stages.
If your first language is not English, verbal reasoning is where that shows most. Extra practice on that section specifically is usually more effective than a general adjustment request, though it is worth asking what is available.
Common Mistakes
- Not practising numerical reasoning. The most trainable test and the most commonly underprepared.
- Practising untimed. Builds accuracy rather than speed, when speed is the constraint.
- Answering verbal questions from general knowledge. "Cannot say" exists for a reason.
- Treating SJTs as general common sense. They are scored against a specific employer's values.
- Taking tests in poor conditions. Interruptions cost questions on a tight clock.
- Letting the invitation expire. They are time-limited and easy to lose in a busy inbox.
Practical conditions, and how to practise
The most avoidable losses in this whole stage have nothing to do with ability.
Do the test on a laptop or desktop, not a phone or tablet. Tables render badly and mis-taps cost questions.
Wired connection where possible, and close everything else. A dropped connection mid-test is sometimes recoverable and sometimes not.
Have paper, a pen and the permitted calculator on the desk before you start.
Do it at a time when you are actually sharp. These arrive with a multi-day window and get taken at 1am. The performance difference is not small.
Somewhere quiet, one sitting, no interruptions. Most tests cannot be paused.
How to practise, in order of value:
With a timer, from the very first practice question. This is the single most important instruction on the page. Untimed practice builds accuracy, and accuracy is not what is being tested — speed under constraint is. Candidates who practise untimed and then sit a timed test are frequently shocked.
Twenty to thirty practice sets is enough for numerical and verbal. Beyond that the returns flatten sharply.
Review your wrong answers properly. Understanding why you picked the wrong option teaches more than another set. Most errors cluster into two or three recurring types, and naming yours is what fixes them.
Do the employer's own practice test where one is offered. Providers differ in interface and question style, and matching the real thing is worth more than generic practice.
Then stop. These tests are a filter to clear, not a skill to master. Once you are consistently clearing the time limit, your remaining preparation time belongs to the competency stories and the assessment centre, which decide far more outcomes.
A three-week plan
Enough to move most candidates from unprepared to comfortably clearing the threshold, at about forty minutes a day.
Week one — diagnose. Do one full practice set of each type, timed, without preparation. The score does not matter; what matters is finding out which of the three is your weak one and what kind of mistakes you make. Nearly everyone discovers their errors cluster into two or three recurring types rather than being spread evenly.
Week two — drill the weakness. If numerical, spend the week on the five core calculations until reverse percentages and compound growth are automatic. If verbal, spend it exclusively on Cannot Say questions, which is where essentially all verbal marks are lost. If situational, read the employer's values and work through scenarios asking which value each response expresses.
Week three — full sets under real conditions. Timed, laptop, calculator you will actually use, no interruptions, at the time of day you intend to sit the real one. Three or four full sets is enough. Review every wrong answer properly.
Then sit the real test, alert, in one sitting, on a wired connection.
What not to do: fifty practice sets. Returns flatten hard after about twenty-five, and the hours are worth far more spent on the stages that actually differentiate candidates rather than merely filter them.
If you fail one
Common, recoverable, and worth knowing the rules before you assume the year is lost.
Ask whether a retake is permitted. Policies vary — some employers allow one, some allow none within a cycle, some allow reapplication the following year. The rejection email usually says, and the recruitment team will answer if it does not.
Results are sometimes shared across an employer's own roles.
Applying to a second position at the same firm may reuse your score rather than giving you a fresh attempt. Worth knowing before you burn a second application.
They are not shared between employers. A poor score at one firm has no bearing anywhere else, whatever the rumour suggests.
Diagnose before you retry. Failing a numerical test because you ran out of time and failing it because you misread every units label are different problems with different fixes.
And apply elsewhere in the meantime. Plenty of good UK employers — smaller firms, mid-tier professional services, the public sector, most technology companies — either do not use these tests or weight them lightly. Concentrating exclusively on the employers with the heaviest testing is a choice, not a requirement.
Frequently Asked Questions
Can I improve my score with practice?
Yes for numerical and verbal, substantially. Situational judgement improves with research into the employer rather than practice. Gamified assessments improve mainly through familiarity with the format.
Are the tests the same across employers?
Several providers supply most graduate tests, so formats recur. Situational judgement is the exception, because scoring reflects each employer's own competency framework.
What happens if I fail?
You are usually rejected for that cycle, and many employers impose a waiting period before you can reapply. This is why preparing before your first attempt matters more than it seems.
Do I need to declare a disability or condition?
If you need adjustments, request them before taking the test. Employers routinely provide extra time and other accommodations, and asking afterwards is much harder.
Can I use a calculator?
Usually yes for numerical reasoning, and the test will say. Practise with the same type you intend to use.
Do scores transfer between companies?
Generally no, though a few employers reuse results across their own applications within a period. Assume each application requires its own attempt.
What is your resume scoring right now?
Scan it against a job description and get your ATS match score in about a minute.
Drop your resume here or choose a file
PDF only. Max 2 MB.
We never share your data or use it to train AI models.
Was this guide useful?
Be the first to rate it.
