Psychometric testing is the widest filter in finance recruiting and the least prepared-for. At most large banks, an application to a spring week, summer internship or graduate scheme goes: online form, then tests, then a recorded video interview, and only after all three does a person look at you. The tests are where the majority of the applicant pool exits. That is not a reason to fear them — it is a reason to treat them as a preparable stage with known formats, known providers and a preparation method that reliably moves scores. Here is what each test actually measures, who builds them, and what to do about it.
Where the tests sit in the funnel
Banks receive tens of thousands of applications for a few hundred seats, and the economics force automation at the top of the funnel. The standard sequence at a large UK or European bank is: application form, psychometric tests (sent within hours or days of submitting, sometimes immediately), a recorded video interview, then an assessment centre. US processes use testing less uniformly, but the largest firms and almost all quant employers test early and hard.
The consequence worth internalising: at the testing stage nobody has read your CV yet. A first from a target university with a strong internship fails the same way a weaker application does if the numerical score lands below the cut. This is why testing rewards preparation so disproportionately — it is the one stage where your background buys you nothing and practice buys you everything.
Numerical reasoning: what is actually tested
Numerical reasoning tests do not test mathematics beyond roughly GCSE level. What they test is speed and accuracy on a narrow band of operations, read out of charts and tables under time pressure. The typical format is 15-25 minutes for 15-25 questions — close to a minute per question, including reading the data, and an on-screen or physical calculator is usually permitted.
The operations repeat across every provider: percentage change and reverse percentage change, ratios and proportions, currency and unit conversions, weighted averages, and multi-step chains of the above ("revenue grew 12% then fell 8%; costs are 60% of revenue; what happened to profit?"). The data arrives as bar charts, line graphs, pie charts and dense tables, often with a deliberate distractor — a footnote changing the units, a figure in thousands next to one in millions, a chart showing percentages where the question asks for absolutes.
- The failure mode is rarely the arithmetic. It is misreading the exhibit: wrong row, wrong year, wrong units, or answering the question you expected rather than the one asked.
- Timed conditions change performance more than candidates expect — an untimed 90% scorer commonly drops to 60-70% under the clock until they have practised the format specifically.
- Adaptive versions (common with some providers) get harder as you answer correctly, so feeling stretched mid-test is usually a good sign, not a bad one.
Situational judgement tests
Situational judgement tests (SJTs) present short workplace scenarios — a colleague takes credit for your work, a deadline conflict, a client asks for something outside policy — and ask you to rank or rate possible responses. They are scored against the firm's own competency framework, which is why the "right" answer can differ slightly between firms.
SJTs are less coachable than numerical tests but far from random. The patterns that score well across finance employers are consistent: take ownership before escalating, escalate genuine risk and integrity issues immediately, communicate directly with the person involved before going around them, and prefer team outcomes over individual heroics. The answers that score badly are also consistent: doing nothing, going straight over someone's head for a routine problem, and any option that compromises accuracy or compliance for speed.
The one preparation step that materially helps: read the firm's published values and competencies before sitting the test, because that document is usually the scoring key in paraphrase.
Game-based assessments
A growing share of early-stage screening, particularly for spring weeks and first-year programmes, uses short games instead of traditional question sets. Common formats include balloon-style risk tasks (inflate for reward, lose everything if it pops — measuring risk calibration and learning from feedback), memory and sequence tasks (working memory and processing speed), rapid sorting or matching tasks (accuracy under time pressure and consistency), and simple resource-allocation or trading games (decision quality as information arrives).
These are harder to practise for in the traditional sense because they measure traits and consistency rather than knowledge. But familiarity still matters: knowing that the risk games reward calibrated risk-taking rather than maximal caution, that consistency of response is itself measured, and that the games typically run 20-30 minutes on a phone or laptop removes the noise that comes from encountering the format cold. Play any available practice versions once or twice; do not grind them.
The providers you will actually meet
A handful of assessment firms build most of the tests in finance recruiting, and knowing which provider a bank uses tells you the format before you open the link.
- SHL — the most widely used traditional provider. Timed numerical and verbal reasoning with chart-and-table exhibits, plus an interactive variant where you drag values and build answers rather than pick from options.
- Korn Ferry / Talent Q — adaptive tests where question difficulty adjusts to your performance, so per-question time limits are short and the test feels harder as you do better.
- Cappfinity — strengths-based, blended assessments mixing numerical items, SJT-style scenarios and video responses into a single flow; common in UK spring week and early-talent processes.
- Arctic Shores — fully game-based assessment; a sequence of short tasks measuring risk calibration, processing speed and persistence rather than curriculum knowledge.
- Pymetrics (now part of Harver) — a battery of short behavioural games measuring attention, memory, risk and fairness traits, matched against profiles of successful employees at the hiring firm.
The pass-mark reality
Scores are typically reported as percentiles against a norm group — other applicants or a graduate population — and banks cut hard. At the most oversubscribed programmes the effective bar sits high enough that the majority of test-takers are eliminated at this single stage, and the cut is applied mechanically: below the line, the process ends, regardless of everything else in the application.
Two practical implications. First, most firms allow one attempt per application cycle — there is no retake if you sit it cold and underperform, which is why sitting a test "just to see what it's like" at your top-choice firm is a serious error. Sequence your applications so early sittings happen at firms you care less about. Second, some banks share providers, and a strong score with one provider occasionally carries between firms — but plan on sitting every test fresh.
Preparation that actually moves the score
Numerical reasoning improves quickly with the right practice and barely at all with the wrong kind. The wrong kind is untimed, casual, and unreviewed. The right kind has three components.
First, timed practice in the real format: full-length practice tests under strict time, ideally matched to the provider you will face. The point is not the content but the pacing — learning what a minute per question feels like and when to abandon a question and move on.
Second, an error log. After every practice test, write down each miss and classify it: misread exhibit, units error, arithmetic slip, ran out of time, genuinely did not know the method. Most candidates discover that two categories account for almost all their misses, which turns vague practice into a targeted fix. This is the single highest-leverage habit and almost nobody does it.
Third, daily mental arithmetic. Percentage changes, fractions to decimals, quick multiplication and division — five to ten minutes a day rebuilds the speed that calculators have atrophied, and it compounds into the assessment centre and interview stages too, where no calculator is offered. Our mental math drill in L3vlup Labs is free, timed, and built for exactly this.
Stage context: where the tests bite hardest
Testing weight varies by stage, and it is highest exactly where candidates are least prepared. Spring weeks test hardest relative to the rest of the process: applicants are first-years with near-empty CVs, so the tests and video interview effectively are the selection process. Summer internships test almost as hard — volumes are at their peak and the test cut does most of the funnel's work before assessment centres. Graduate programmes still test, but with more of the decision weight sitting in interviews and the assessment centre, because the pool is smaller and more experienced.
Off-cycle internships and boutique processes frequently skip formal testing altogether — one of several ways they reward candidates who approach firms directly. And quant employers are the extreme case in the other direction: their online tests are harder, more mathematical, and even more decisive than banking's, with the majority of applicants eliminated by a timed maths or coding assessment before any human contact.
Frequently asked questions
What psychometric tests do banks use in recruiting?
Most large banks use some combination of numerical reasoning (timed chart-and-table calculations), situational judgement tests scored against the firm's competency framework, and increasingly game-based assessments. The main providers are SHL, Korn Ferry/Talent Q, Cappfinity, Arctic Shores and Pymetrics, and which provider a firm uses determines the format you will see.
How hard are numerical reasoning tests for finance?
The mathematics is roughly GCSE level — percentages, ratios, conversions and weighted averages. The difficulty is time pressure and data extraction: about a minute per question including reading charts and tables that often contain deliberate distractors like unit changes. Untimed ability does not predict timed performance until you have practised the format.
What score do you need to pass a bank's online tests?
Banks score against a percentile norm group and set their own cut-offs, which they do not publish. At oversubscribed programmes the bar is high enough that most test-takers are eliminated at this stage. Because the cut is applied before a human reviews the application, a strong CV does not compensate for a below-threshold score.
Can you retake a psychometric test if you fail?
Generally not within the same application cycle — most firms allow one attempt per application. The practical consequence is sequencing: sit your first tests at lower-priority firms so the format is familiar before you face your top choices.
How should you prepare for finance psychometric tests?
Three things move scores: full-length timed practice in the real provider format, an error log that classifies every miss so you fix the two or three failure patterns causing most of them, and daily mental arithmetic drills to rebuild calculation speed. Casual untimed practice barely helps.
Do spring weeks have harder tests than internships?
The tests themselves are similar, but they carry more weight at spring week stage because first-year applicants have little else to be judged on — the tests and video interview effectively are the process. Summer internships test almost as decisively due to application volume; graduate schemes shift more weight to interviews and assessment centres.
Related guides
Want this applied to your recruiting?
Reading is the easy part. For practitioner feedback tailored to your situation, work 1:1 with Suro.