Online Assessment Practice
60% of candidates fail before the interview
SHL Verify G+, Watson-Glaser, Pymetrics, McKinsey Solve, HackerRank, on the exact interface and timer your target firm runs. Timed and percentile-scored against modelled graduate norms, with section-by-section coaching that tells you which question type is dragging your overall score.
Free, no signup, no credit card
Choose your test
12 questions per test. Timed. Instant results with explanations. No signup required.
Every provider
Each firm picks a different provider. You only pass when you practise on the right one. Generic online assessment practice on the wrong format wastes your prep time.
The bulge-bracket standard. Numerical, verbal and inductive reasoning, plus the OPQ32 personality questionnaire. Used by Goldman Sachs, Barclays, Deutsche Bank and most asset managers.
Twelve neuroscience-derived games (Balloon Analogue Risk Task, digit span, flanker, Iowa, etc.). Builds a behavioural profile rather than a cognitive score.
The law-firm critical thinking standard. Inference, assumptions, deductions, interpretation and evaluation of arguments. 40 questions in 30 minutes. Pass mark sits at the 65th to 75th percentile depending on the firm.
McKinsey's proprietary game-based assessment (formerly the Problem Solving Game). Ecosystem, redrock and case scenarios. Tests data interpretation, hypothesis-building and decision-making under time pressure.
Bain's combined situational judgement and online case. Tests practical judgement plus structured problem-solving on a consulting-style case prompt.
Numerical, verbal and abstract reasoning combined with a situational judgement test. Used by Oliver Wyman and several mid-market consulting firms.
Modular reasoning suite (the scales tests) plus the scales personality inventory, all built around short, sharp adaptive timers. Used by Morgan Stanley and UBS in their markets and junior-talent tracks.
The dominant graduate immersive: a blended, scenario-based assessment combining situational judgement with numerical and verbal exercises inside a single business case, usually followed by a job simulation. Used by HSBC, Deloitte and EY (KPMG pairs it with an Arctic Shores game).
A numerical, verbal and logical reasoning suite paired with a work-style assessment. A long-standing screen across several US bulge brackets.
Adaptive numerical and verbal Elements tests, often paired with a behavioural survey such as Plum. Used by Citi and several global banks.
BCG's proprietary interactive chatbot case (the Casey assessment): structuring, data selection, quick maths and a written recommendation, plus a one-way video pitch.
The engineering and Strats screen for most banks. Two to three medium-to-hard algorithm problems plus CS multiple choice and quantitative maths. 90 to 120 minutes. Goldman's configuration is regarded as among the most difficult on Wall Street.
A task-based, game-style behavioural assessment that measures how you work rather than what you know. KPMG pairs it with the Cappfinity immersive; it produces a strengths profile from in-game behaviour, with no pass or fail in the traditional sense.
The basics
A online assessment is a structured cognitive evaluation used by graduate employers to filter applicants before any interview stage. It typically arrives within 24-72 hours of your application clearing the resume sift, delivered via a third-party platform (SHL, Cubiks, Pymetrics, Watson-Glaser, McKinsey Solve, HackerRank) embedded into the firm's application portal.
The standard structure combines three to five timed sections: numerical reasoning (~20 questions in 20 minutes), verbal reasoning (~30 questions in 20 minutes), logical / inductive reasoning (~18 questions in 24 minutes), a situational judgement test (~10-15 scenarios), and an untimed personality questionnaire (typically 30-45 minutes). Engineering candidates skip the cognitive battery and face a 90-120 minute coding assessment instead.
Scoring is comparative, not absolute. Your performance is benchmarked against a norm group, usually all applicants from target universities within the same recruitment cycle. Competitive thresholds sit around the 80th percentile on the numerical section for front-office banking, 70th percentile on verbal and logical. Aggregation is non-compensatory: an exceptional verbal score cannot rescue a weak numerical.
Most firms enforce a strict no-retake policy within the same cycle. The score is final the moment you submit, and if it falls short you wait twelve months for the next intake. Practising on the exact format ahead of time is how you avoid spending your one real attempt learning the interface.
Section breakdown
Every multi-section assessment is built from the same five section types in different combinations. Understanding what each one tests is half the prep.
What it tests. Data interpretation, multi-stage percentage and ratio calculations, currency translation, identifying the relevant variables inside a busy financial table.
Common trap. Distraction columns and units-of-measurement mismatches (thousands vs millions, foreign currencies). Most candidates lose points to careless misreads, not the underlying maths.
What it tests. Evaluating True / False / Cannot Say statements against a 150-200 word passage on a regulatory, economic or operational topic.
Common trap. Bringing outside knowledge into the evaluation. If the passage does not explicitly state it, the answer is Cannot Say, even when the statement is clearly true in the real world.
What it tests. Non-verbal pattern recognition: identifying rules across geometric shapes, matrix transformations, and spatial sequences.
Common trap. Trying to read the whole shape at once. Top candidates isolate one variable at a time (rotation, shading, side count) and eliminate violators of that single rule.
What it tests. Practical workplace decision-making. You rank or pick the best and worst actions in realistic analyst dilemmas (prioritisation conflicts, data errors, compliance issues).
Common trap. Choosing the "solo hero" answer where you fix major problems alone without informing the team. Firms score upward communication and transparency, not lone wolves.
What it tests. Behavioural and work-style profile. Most blocks ask you to pick one statement as Most like me and one as Least like me from a group of four.
Common trap. Gaming the test with an extreme persona. Internal consistency checks shuffle similar items throughout, so contradictory answers flag the profile and can trigger an automatic rejection.
Feedback that helps
Every practice run is percentile-scored, not raw-marked - the metric your target firm actually uses, with the modelling stated.
Your score for each section, placed on a modelled UK or US graduate norm curve for that test format. So you know whether 70% is excellent or worrying.
"You are spending 90 seconds on multi-stage currency calculations" beats "you scored 68%". Specific advice on which question type is dragging your overall score.
Item difficulty mirrors the real provider's band, and adaptive formats like Talent Q step up and down as the real test does.
Why candidates fail
Specific failure patterns documented across past applicant accounts. Most are avoidable with structured preparation on the right format.
Mismanaging the numerical time budget
Spending 3-4 minutes on one complex question leaves the final five rushed or guessed. Discipline: ~60 seconds per question or move on.
Bringing outside facts into verbal evaluations
If the passage does not explicitly state it, the answer is Cannot Say. Reading widely actually hurts here because you instinctively bring in real-world knowledge.
Trying to read the whole geometric pattern at once
Inductive section: cognitive overload. Top candidates isolate one variable (rotation, shading, count) and eliminate violators before moving to the next rule.
Gaming the personality questionnaire
Selecting an extreme "perfect candidate" persona triggers consistency-check flags. Answer honestly, with the role context in mind.
Engineering candidates writing brute-force code
O(N squared) solutions pass small test cases but time out on the large hidden ones. Write a working brute force first, then optimise.
Random guessing under negative marking
Some sections penalise wrong answers (+5 correct, -2 incorrect, 0 unattempted). Strategic skipping protects the score; random guessing destroys it.
Compare your options
The honest comparison. Intervyo is calibrated against the live test interfaces each firm actually deploys. Generic libraries give you volume without firm fidelity.
| Feature | Intervyo | Generic aptitude site | Free practice PDFs | Real assessment |
|---|---|---|---|---|
| Simulates the real provider format | partial | |||
| Percentile and stanine reporting | ||||
| Adaptive formats simulated (Talent Q) | ||||
| Section-by-section coaching | ||||
| Covers SHL, Watson-Glaser, Pymetrics, McKinsey Solve | partial | |||
| Fresh questions every sitting | partial | |||
| Built for graduate finance | ||||
| Single attempt counts |
How it works
Goldman Sachs runs SHL Verify G+. Linklaters runs Watson-Glaser. JPMorgan runs Pymetrics. Pick the firm and Intervyo loads the right test architecture.
The exact interface, the exact timer, adaptive difficulty calibrated to the firm. Same pressure as the real attempt, with infinite retries.
Per-section percentile against the live candidate pool, coaching by question type, and the specific patterns dragging your score down.
Keep going
FAQ
A online assessment is a structured cognitive evaluation graduate employers use to filter applicants before any interview stage. Most are timed, multi-section assessments covering numerical reasoning, verbal reasoning, logical / inductive reasoning, and a situational judgement test. Some firms also include a personality questionnaire (such as SHL OPQ32) or a custom game-based assessment (Pymetrics, McKinsey Solve).
SHL is used by Goldman Sachs, Barclays, Deutsche Bank, PwC and most bulge-bracket banks, while HSBC, Deloitte, EY and KPMG run Cappfinity immersive assessments. Pymetrics powers JPMorgan; Bank of America uses IBM Kenexa and Citi uses Korn Ferry Talent Q. Aon cut-e screens Morgan Stanley and UBS. Watson-Glaser is the law-firm standard (Linklaters, Clifford Chance, Freshfields, A&O Shearman). McKinsey runs its proprietary Solve game and BCG its Casey online case. Bain uses SOVA plus its own online case. Engineering and Strats roles at most banks (including Goldman) use HackerRank instead.
Most firms send the assessment link with a hard 5-7 day completion window from the timestamp of the invitation. The test itself runs 60-120 minutes total once you start, split across timed sections (numerical ~20 mins, verbal ~20 mins, logical ~24 mins, plus untimed personality). High-performing candidates complete within 48-72 hours of receipt to stay competitive on rolling tracks.
Firms do not publish exact thresholds, but the practical floor for competitive online assessments sits around the 80th percentile on the numerical section (the highest-weighted), and the 70th percentile on verbal and logical. Front-office investment banking and Markets tracks demand the 85th-90th percentile. Scoring is non-compensatory: a strong verbal cannot rescue a weak numerical.
Yes for the SHL numerical reasoning test at home. The platform also provides an on-screen basic calculator. However, the questions are designed to test quick estimation under time pressure, not long-form arithmetic, so calculator dependence often slows candidates down rather than helping them. The Watson-Glaser, McKinsey Solve, and most game-based assessments do not allow calculators.
Intervyo is calibrated against the firms graduates actually apply to, not generic aptitude libraries. Practice runs match the exact SHL Verify G+ interface Goldman uses, the Watson-Glaser timer Linklaters runs, and the Pymetrics game architecture JPMorgan deploys. Adaptive difficulty, percentile scored against the live UK / US candidate pool, with section-by-section coaching that tells you which question type is dragging your overall score.
15-30 hours of structured practice over 2-3 weeks is the band where most candidates who pass on the first attempt sit. Engineering candidates targeting HackerRank-style screens at Goldman or JPM need 40-60 hours covering data structures, algorithm patterns and probability. Practice in full-length timed simulations, not untimed drilling.
Three failure modes account for the majority of fails. First, mismanaging time on the numerical section: spending 3-4 minutes on one complex question leaves you guessing the final five. Second, bringing outside knowledge into the verbal "True / False / Cannot Say" evaluations - if the passage does not state it, the answer is Cannot Say. Third, gaming the personality questionnaire by selecting an extreme persona that triggers consistency-check flags.
No. Most firms enforce a strict one-attempt-per-cycle policy. Once submitted or once a section timer hits zero, the score is final. If your application is rejected based on the assessment, you wait 12 months until the next annual recruitment cycle to apply again. The practice runs in the Pack exist so the real attempt is not the first time you see the format.
Practise the exact provider your firm uses, scored against the live candidate pool percentile. Fix what is weak before the one attempt that counts.
Start practising online assessmentsFree, no signup, no credit card