Which online assessment does your firm send?

One page with the exam formats (80 in 8, 80 in 8 typed, 50 in 5, 60 in 6) where two digit products, decimals, fraction conversions, percentages and missing operands appear at the difficulty candidates report from the real papers and a wrong answer costs a point, and the ranked drill, which is the four operations one question at a time with the next appearing the instant you are right. The ranked drill runs 30s, 60s, 120s or 300s; the 120 second game with all four operations is the one that scores.

• Optiver
• Flow Traders
• IMC
• Cappfinity

What it tests

Raw arithmetic throughput, and on the exam formats, whether you know when to skip. Your ranked score is how many you clear, not a percentage, so hesitation is the thing that costs you.

Where you will meet it

The opening test at most prop and market making firms (Optiver, Flow Traders, IMC and similar), and a fair proxy for the numerical reasoning tests banks use.

How to get better

Nothing is deducted for a wrong keystroke, the question only clears when the answer is right, so type immediately and correct as you go. Learn times tables to 20 and get comfortable dividing by 7, 11 and 13.

A fraction appears with five decimal options and you pick the right one. Values are checked to three decimal places, so 7/16 is 0.4375 and 3/8 is 0.375. One wrong pick voids the question: the right answer still moves you on, but it no longer scores.

• Optiver
• IMC

What it tests

Recall rather than calculation. Trading tests assume you already know these and are timing something else.

Where you will meet it

Inside mental maths rounds, and in market making questions where prices move in halves, quarters, eighths and sixteenths.

How to get better

Memorise the conversions from halves through twentieths. Anything with 8, 16, 32 or 64 underneath should be instant, not worked out.

A run of terms appears and you type the next one. Questions come from a verified bank and include fractions and powers, not only arithmetic steps.

• Optiver
• Flow Traders
• Suited

What it tests

Pattern recognition under time pressure, plus the discipline to abandon a wrong idea quickly rather than forcing it.

Where you will meet it

Sequence rounds in trading assessments, including the Optiver, Flow Traders and Maven style tests, and the numerical sections of graduate aptitude tests.

How to get better

Work through it in a fixed order: first differences, then second differences, then ratios, then two interleaved sequences. If nothing lands in a few seconds, skip and take the next one.

The same drill with no warm-ups and nothing from the easy or standard tiers: three-step operation cycles, second-order recurrences, hidden common denominators, prime skips, paired divisor tracks, fraction chains with geometric denominators and the rest. Scores go on a separate board, so a hard run is never ranked against a normal one.

• Optiver
• Flow Traders

What it tests

Whether the fixed-order method still works when the first differences and ratios tell you nothing, and how quickly you give up on a question that is not going to land.

Where you will meet it

The later questions of the Optiver and Flow Traders number-logic rounds, which are the ones that separate the field.

How to get better

Expect to skip more than usual. Try two interleaved tracks and a repeating cycle of operations before anything else; on fractions, put every term over one denominator first.

Twelve questions in eighteen minutes, ninety seconds each. Every question shows a table of three to five rows (regions, offices, product lines, funds, quarters) and asks one thing: which row is largest, what share of the total, the percentage change, the figure per head, next year's figure at a stated rate, a conversion into dollars, or a two-step margin question. One point per correct answer, nothing deducted, skipping allowed, no feedback until the end.

• SHL
• Cappfinity
• Talent Q
• Suited
• Saville

What it tests

Data interpretation under time: reading the right row and the unit in the column header, choosing the right base for a percentage, and estimating well enough to rule most options out before doing the arithmetic. Every wrong option is the number a specific mistake produces.

Where you will meet it

SHL's Verify numerical tests (Barclays, Deutsche Bank, Jefferies), the numerical module in Cappfinity's immersive assessments (HSBC, Lloyds) and Korn Ferry's Talent Q Elements numerical (Citi, UBS).

How to get better

Read the units and column headers before the question, work out exactly what is being asked before touching the numbers, estimate first to kill three of the five options, and skip anything that needs more than two minutes.

Six tabs (Income, Costs, Market share, Employees, Return on equity, Outlook) about one fictional company, and a statement under them to judge true, false or cannot say. Thirty seven statements in twelve minutes, about twenty seconds each, roughly a third of each answer. One point per correct answer, nothing deducted, skip allowed. The tab you had open is recorded for the readout at the end but never scored, and every miss is shown with its answer, its tab and the figures that settle it.

• Aon
• Nomura
• UBS
• Morgan Stanley

What it tests

Finding the right tab fast and computing one figure from it: a comparison, a percentage change, a share of a total, a ranking, or a figure that needs two tabs. And the discipline to answer cannot say when the fact is not there, however plausible the statement sounds.

Where you will meet it

Aon's scales numerical (formerly cut-e), the numerical reasoning test Nomura, UBS, Morgan Stanley and BNP Paribas send for spring weeks and internships, and the numerical half of what Nomura sends alongside Motion.

How to get better

Read the six tab names before the first statement so you know where each fact lives. The statement usually names its tab. If the fact is on no tab the answer is cannot say, and twenty seconds means you never compute more than one figure.

Six tabs (Structure, Strategy, Principles, Products and services, Locations, Executive board) of prose about one fictional company, and a statement under them to judge true, false or cannot say. Forty nine statements in twelve minutes, about fifteen seconds each, with about a fifth cannot say, which is the share students who sat the real test report. One point per correct answer, nothing deducted, skip allowed. The tab you had open is recorded for the readout at the end but never scored, and every miss is shown with its answer, its tab and the sentence that settles it.

• Aon
• Nomura
• UBS
• Morgan Stanley

What it tests

Finding the right tab fast and reading one sentence exactly: whether it says the thing, says the opposite, or says nothing about it. And the discipline to answer cannot say when a likely-sounding statement is not actually there.

Where you will meet it

Aon's scales verbal (formerly cut-e), the verbal reasoning test in the same family as the numerical test Nomura, UBS, Morgan Stanley and BNP Paribas send, and the true, false or cannot say verbal tests of most graduate batteries.

How to get better

Read the six tab names first: a person is Executive board, a city is Locations, a plan is Strategy. Find the one sentence the statement is about and ask whether it settles the point. Plausible is not true and unlikely is not false; if the tabs are silent, it is cannot say.

Eight sequences, easiest first, once a day. Up to 30 minutes with the clock as the tiebreaker, a daily board, a shareable result and a streak that resets if you miss a day. New challenge at midnight.

What it tests

Consistency more than ability. It exists so that prep is a habit instead of a panic the night before a deadline.

Where you will meet it

Not a test format in itself. It is the daily rep that makes the rest of this page work, and the easiest way to compare yourself with people on your course.

How to get better

Do it at the same time every day. Ten minutes a day for a term beats a weekend of cramming the week applications open.

Four sections back to back on their own clocks, each with a brief and a thirty second countdown before it. Fifty typed arithmetic questions in five minutes, thirty sequences in six minutes with letters read as alphabet positions mixed in among the numbers, eighteen five-option probability questions in thirty minutes where a wrong pick costs a point and a skip is free, and one quantity to estimate in two minutes with an 80 percent confidence interval. A point a question, 99 in all, no feedback until the end and no going back.

• Maven Securities

What it tests

Whether the arithmetic holds up when every answer has to be typed and the clock is six seconds a question, whether a letter pattern reads as quickly as a number one, whether you leave the probability questions you cannot narrow down, and whether you can put honest bounds on a number you do not know.

Where you will meet it

The online assessment Maven Securities sends for its trading spring week, summer internship and graduate programmes, as candidates report it: mental maths, sequences and probability at these counts and clocks, with reports from 2026 adding typed answers and the estimation question. Maven publishes none of the counts, so the shape here is the reported one.

How to get better

Sit it once end to end before doing anything else, and read the section cards rather than the total: the maths section is where practice moves the number, the probability section is where guessing loses it. On the estimation question, say your reasoning out loud, because the real one has you explain it to the camera.

You quote a bid and an ask on the sum of some dice, and a counterparty trades against you. Part of that flow already knows the settlement value and only trades when your quote is wrong; the rest walks if your quote is too wide. One more die is revealed after every round.

• Optiver
• IMC
• SIG
• Jane Street

What it tests

Quoting under adverse selection. Tight quotes get picked off by the informed side, wide quotes get no fills at all, and the width that survives both grows with how much is still hidden.

Where you will meet it

The market making exercise trading firms run in interviews and assessment centres, at Optiver, IMC, SIG and Jane Street among others.

How to get better

Start wider than feels comfortable while most of the dice are hidden, then tighten as they come up. The game scores your quote against the exact fair value, so a good quote that happens to lose still scores well.

Four question types, all answered as a whole percentage: break-even equity, minimum defence frequency, bluff frequency and outs to equity. Two minutes, and the numbers stop dividing nicely in the back half.

• Trading interviews

What it tests

The indifference arithmetic behind any risk and reward decision, done in your head and to the nearest percent rather than exactly.

Where you will meet it

Poker questions in trading interviews, which come up far more often than the poker content suggests, and any question asking what price makes a bet break even.

How to get better

Break-even equity is risk over risk plus reward. Learn that one shape and three of the four kinds fall out of it. Answers are accepted in a band, so estimate rather than dividing precisely.

Some dice are face up, some hidden. A dealer quotes a bid and an ask on the total. You buy if the ask is too low, sell if the bid is too high, and pass if the quote is fair. Difficulty presets change how many dice are hidden, how tight the quotes are and how long you get per round.

• Prop firms

What it tests

Expected value under uncertainty, and the discipline to pass. Trading firms care as much about whether you trade a fair market as whether you spot a bad one.

Where you will meet it

Market making rounds and trading games at prop firms, and the on-the-spot pricing questions in trading interviews.

How to get better

A hidden die is worth 3.5. Add the visible faces, add 3.5 for each hidden one, then compare with the quote. Passing on a fair market is a correct answer, not a wasted round.

A dealer quotes a bid and an ask on the sum of five cards, some face down. Unlike dice, cards are drawn without replacement, so what has already been revealed changes what the hidden cards are worth.

• Prop firms

What it tests

Expected value with conditional information, and updating a price when new information arrives mid round.

Where you will meet it

Card based trading games in prop firm assessments and superday market making exercises.

How to get better

Look at what is already showing before you price the rest. If the visible cards are high, the unseen ones are worth less than the deck average, and the fair value moves with them.

Six minutes of probability, one question at a time and no going back. Two warm-up questions, then the families candidates report from the real rounds: cards and dice counts, Bayes, biased gambler's ruin, waits for a run of heads, random walks round a circle, reroll games, coupon collecting, the birthday problem and totals of many dice. Five options, +1 for a correct answer, 1 off for a wrong one, and skipping is free.

• Optiver
• SIG

What it tests

Whether the standard families are automatic rather than worked out from scratch, and whether you can tell a question you can do from one you cannot before the clock decides for you.

Where you will meet it

The probability sections of trading assessments, Optiver's Beat the Odds among them, and the SIG Problem Solving Assessment at a slower pace.

How to get better

Guessing at random loses marks on this scoring, so skip anything you cannot see a route through in a few seconds. Every wrong option is the answer to a specific mistake, so if one looks obviously right, check which mistake makes it right.

Thirty questions, each on its own strict 90 second clock that skips the question when it runs out, no going back and nothing marked on screen until the end. Five coarse options and you pick the one closest to the true value, because the exact answer is usually not on the list. +1 for a correct answer, 1 off for a wrong one, skipping is free. No warm-ups: coin pairing, grid escapes, two tokens meeting on a polygon, the martingale bankroll, stopping sums, the overshoot of a running total, random number differences and going broke in a fixed number of rolls, mixed with the hard families from the main drill. Scores go on a separate board.

• Optiver

What it tests

Whether you can estimate to the nearest option under a per-question clock rather than compute exactly, and whether you skip when the shortlist is longer than two.

Where you will meet it

Optiver's Beat the Odds, as reported by 2025 and 2026 candidates, and the harder end of other trading firms' probability rounds.

How to get better

Order of magnitude first, then only as much precision as the gap between the options needs. A blind guess across five options is worth minus 0.6 on average, so skip anything you cannot get down to two.

Eight questions, 45 seconds each by default, drawn from a 500-question bank modelled on the estimation questions trading interviews actually use: markets, cities, packing problems and classic Fermi counts. For each one you give a low and a high that you are 90% sure contains the answer. Capturing the true value scores, and the tighter the interval the more it scores. Miss it and the question is worth nothing.

• Consulting
• Optiver

What it tests

Structured estimation and calibration: whether you actually know how much you know.

Where you will meet it

Consulting case maths, trading interview brainteasers and the estimation questions Optiver style firms use to see how you reason out loud.

How to get better

Break the question into factors you can each get within a factor of three, then multiply. Keep intervals honest: tight when you know the answer, wide when you are guessing. Hedging everything scores badly.

Ten questions, 40 seconds each by default, one of fifteen kinds a McKinsey, BCG or Bain case asks for out loud: a percentage of a big number, a percentage change, market share, compound growth, the Rule of 72, a CAGR estimate, break-even volume, a margin, a weighted average, a market-sizing chain, payback, a perpetuity, big multiplication and division, daily to annual. Exact where the case is exact; an estimate within consultant rounding counts where an estimate is the point. Ten points for a right answer and up to five more for speed, with the method a strong candidate says out loud shown after each.

• McKinsey
• BCG
• Bain

What it tests

Whether the standard moves are automatic under a clock, and whether you can say the method before the number.

Where you will meet it

Every MBB case interview, the arithmetic steps of BCG's Casey chatbot case, and the data questions in McKinsey Solve's Redrock study.

How to get better

Say the approach first, do the percentages before the big multiplication, keep the units, and round only when the question says an estimate will do. The speed bonus rewards a confident answer, not a reckless one.

Ten exhibits from a bank of twenty-four, 60 seconds each by default. Each is a titled table with a line of context and a footnote, in the shape of the slides in the public MBB sample cases: a P&L over two years, share by competitor, a funnel, a cost structure, a payer mix, a price test. One question per exhibit of the three kinds a case asks: compute the number, name the insight, or judge what to do next. Four options, one point each, no penalty for a wrong answer, and the reasoning shown after every one.

• McKinsey
• BCG
• Bain

What it tests

Reading a table for the comparison that matters (a ratio, a gap, a trend) and saying what it means for the client's question, rather than reading the numbers aloud.

Where you will meet it

The exhibit questions of every MBB case, Bain's exhibit-heavy cases in Singapore, and BCG's Casey, which shows charts and asks for the implication.

How to get better

Read the title and the footnote before the numbers: the trap is usually in one of them. Say what the exhibit shows in one line, then the so-what, then what you would look at next.

Two rows of four symbols, what went in and what came out, and an operator is the four digit code that rearranged them. The job is to name the operator, not to work out the output. Later levels chain two operators and hide one, and the hardest give you the second stage so the first has to be solved backwards.

• Aon
• BNP Paribas

What it tests

Deductive reasoning against a clock, and specifically reading a transformation backwards, which is a different skill from applying one forwards.

Where you will meet it

Aon's switchChallenge, used by BNP Paribas among other banks, and the deductive sections of graduate aptitude batteries.

How to get better

Read the output row first and work out where each symbol came from. Checking only the first digit is exactly what the near-miss options are built to punish, so verify all four before committing.

A four by four or five by five grid. Some squares hold a shape, the rest are blank, and one holds a question mark. No shape repeats in any row or column, and that rule is all you get. You pick the missing shape from the row under the grid, and you can pencil shapes into the blank squares as working. Six minutes, as many grids as you can, harder as you go: the easy ones give way to one line, the hard ones need two or three other squares settled first.

• Aon
• Rolls-Royce

What it tests

Deductive reasoning: applying a rule you have been given, in one step and then in several, under a clock. Aon calls it deductive-logical thinking.

Where you will meet it

Aon's gapChallenge, the gamified version of cut-e's scales lst, which a Rolls-Royce applicant in the 2026/27 cycle reported as the timed puzzle in that firm's assessment, and which Aon's other graduate clients can send in place of the switchChallenge.

How to get better

Start at the question mark. List what its row is missing, strike out what its column holds, and most grids are done. If two shapes remain, ask where else each could go in that line. Fill other squares only when that fails, and never guess: a wrong answer is deducted.

A row of shapes goes in, one or more operators run on it in order, and a row comes out. A panel says what each operator does. One part of the flow is missing: the output at first, then an operator, then the input, which means running the chain backwards. The letters are reshuffled every question so there is nothing to memorise.

• Saville
• IMC

What it tests

Holding several freshly given rules in working memory while applying them, and inverting a process rather than just following it.

Where you will meet it

The diagrammatic and abstract reasoning sections of graduate aptitude batteries, including the Saville tests IMC uses, and the logical reasoning parts of consulting and banking screens.

How to get better

Read the whole panel before you look at the shapes, and check one attribute at a time: order first, then shape, then fill. On the backwards questions, take each option and run it forwards rather than trying to undo the chain in your head.

Nine tiles of abstract shapes. Eight obey a rule nobody tells you and one does not, and the job is to pick the one that does not, twenty times in five minutes. Early items hang the rule on one property you can see at a glance. Later ones tie two together, so the rule is something like solid whenever the count is even.

• Aon
• UBS

What it tests

Inductive reasoning: generating candidate rules from examples and discarding them quickly, which is a different skill from applying a rule you have been given.

Where you will meet it

Aon's scales ix, which UBS sends as its inductive reasoning test, and the abstract or diagrammatic sections of most graduate aptitude batteries.

How to get better

Check one property at a time rather than staring at the grid: shape, colour, count, solid against hollow, then the marks inside. Eight against one on any of them is the answer. If none splits cleanly the rule pairs two properties, and fill against count is the usual pair.

A six by six grid with a ball, an exit, three to six blocks and a scatter of fixed walls. Slide any piece along a row or column to clear the ball's route and bring it to the exit. A move is one slide, however far it travels, and every move counts: there is no reset and the fewest-moves figure is never shown. Six minutes, forty puzzles all in one difficulty band, two points for a solve and a third for doing it in the fewest moves the puzzle allows. Every board looks easy and needs ten to fourteen moves in the right order, which is what students who sat the real test describe.

• Aon
• Nomura

What it tests

Planning: holding a route in your head and ordering the moves that clear it before you touch anything. Aon labels it complex planning capability, and with no reset a wrong first move is paid for.

Where you will meet it

Aon's motionChallenge, which Nomura sends with its spring week and internship online assessment, and the Aon smartPredict batteries other employers use.

How to get better

Trace the ball's path back from the exit, count what sits on it, and find each blocker's parking space before you move anything. The whole plan has to be right before the first drag, so when one dies, unpick it by hand for the solve points or skip rather than shuffle.

Four abstract figures, each following on from the last, and five candidates for the fifth. Every figure is set by its shape, how many of it, solid or hollow, colour, and on harder items a dot walking round the corners, each following its own rule. Fifteen items in eighteen minutes, easiest first: one attribute changes at first, then two, then three or more with the dot always among them.

• SHL
• Talent Q
• Plum

What it tests

Inductive reasoning in its commonest form: reading several independent rules off a short series and applying each one step further, under a clock that allows about a minute an item.

Where you will meet it

SHL's Verify inductive reasoning test, which Barclays sends inside Verify Interactive G+ and Deutsche Bank and Jefferies send in the classic form, Korn Ferry's Talent Q Elements logical test used by Citi and UBS, and the problem solving sections of Plum's Discovery Survey used by Citi and Bloomberg.

How to get better

Check one attribute at a time, always in the same order: shape, count, fill, colour, dot. Say the rule for each before you look at the options, even when it is "stays the same". Then throw out each option on the first attribute it gets wrong; one wrong attribute is always enough.

Two, three or four statements about a small group (all analysts work late, no one who works late takes the train, Priya is an analyst), then a further statement to judge as always true, sometimes true or never true taking the statements as given. Sixteen items in ten minutes, easiest first, one point each and nothing deducted. Every answer is computed by checking the statement in every arrangement that fits the premises, and every item you miss is explained at the end.

• Suited
• PJT
• Evercore

What it tests

Deduction under a clock: following a chain of alls and nos, reading some as at least one and nothing more, and recognising when the statements simply do not settle the question, which is the verdict most often missed.

Where you will meet it

The logical reasoning section of the Suited assessment, which PJT Partners, Evercore, Houlihan Lokey and Guggenheim send to investment banking applicants, and the syllogism items in most graduate reasoning batteries.

How to get better

Draw the circles. All puts one inside another, no pulls two apart, some is a single cross in the overlap. Then try to make the question false while keeping every statement true, and try to make it true; if you can do both it is sometimes, and bring in nothing you know from outside the statements.

Fifteen scenarios from a spring week or an internship, each with four things you could do, ranked by tapping them in order. Ten points a scenario when your order matches the key, fewer the further from it you are, nothing for the key reversed. Untimed in the real test but speed scored, so the time on each scenario is shown and the run is capped at fifteen minutes.

• Cappfinity
• HSBC
• Lloyds
• SHL

What it tests

Judgement and ownership rather than knowledge: whether you keep the right person informed, own a mistake straight away, escalate without going over heads, and do the right thing when nobody is watching. The skill is the ranking between the two plausible answers.

Where you will meet it

Cappfinity's Situational Strengths Test and the strengths based judgement inside its immersive assessments for HSBC, Lloyds, NatWest and the Big 4; close to the SHL situational judgement tests at Deutsche Bank and the Barclays Business Insight stage.

How to get better

Rule out the two you would not do first, then ask which of the remaining two shows judgement and ownership. Escalating early is usually right, hiding a mistake is always wrong, and doing nothing is rarely the top answer.

Three parts in the real order. Twelve scenarios over a day at a fictional advisory firm, ranked from the thing you would most likely do to the thing you would least likely do, or answered as a most and least pick of four. Twenty pairs of opposing statements with a slider between them, unscored. Then twelve chart and table questions answered by picking an option, typing the number or ranking the rows.

• Cappfinity
• Rothschild & Co
• HSBC
• Lloyds
• AtkinsRéalis

What it tests

Judgement against the strengths Rothschild says it looks for, consistency across the behavioural pairs, and unhurried numeracy: reading a chart or table correctly and calculating without a trick.

Where you will meet it

Rothschild & Co's online assessment for spring insight, summer and graduate programmes, as candidates report it, the Cappfinity strengths and immersive assessments at HSBC, Lloyds, NatWest and the Big 4, and the untimed scenario and chart questions of the Amberjack blended assessment AtkinsRéalis is reported to send.

How to get better

Sit it once end to end for the shape, then use the readout: the scenarios are where points move, the sliders are where a contradiction shows, and the numbers should be near full marks after a few runs of Numerical Reasoning.

Three parts in the reported order. Twelve scenarios from a week with an audit team at a fictional professional services firm, ranked from the thing you would most likely do to the thing you would least likely do, or answered as a most and least pick of four. Twenty pairs of opposing statements with a slider between them, unscored. Then twelve chart and table questions answered by picking an option, typing the number or ranking the rows, with a calculator allowed.

• Cappfinity
• SHL
• EY

What it tests

Judgement against the eight skills EY says it hires on (proactive, doing the right thing, curious, building relationships, agility, thinking critically, embracing technology, motivated), consistency across the behavioural pairs, and unhurried numeracy with the time recorded.

Where you will meet it

The Cappfinity strengths assessment and job simulation EY sent for Discover EY, its first-year insight programme, and for its internships and graduate schemes up to the April 2026 intake, as candidates report them. EY's privacy notice names SHL as the platform for the 2027 cycle, with the same question formats listed on its FAQ: choose from options, interpret a chart or table, drag items into position, spot a pattern, record a video.

How to get better

Rule out the two you would not do, then ask which of the rest keeps the right person informed and owns the problem. Answer the sliders as yourself and consistently. The written and video parts of the real one are not here: before you sit it, draft a three-line client email and a two-minute answer to why EY and why this service line.

Fifty multiple choice questions in twelve minutes, which is about fourteen seconds each. Verbal analogies, vocabulary, disarranged sentences and proverbs; arithmetic word problems, number series and value comparisons; logical deduction, date reasoning and spatial questions. All mixed together, and getting harder as the paper goes on, so the questions that need the most time arrive when you have the least. You can skip but you cannot go back. Scored out of 50, and every question is shown with its answer at the end.

• Wonderlic
• Criteria CCAT
• Lazard

What it tests

General cognitive ability under heavy time pressure, and the judgement that comes with it: whether you can tell in a couple of seconds that a question is going to cost you a minute, and move on. That call is worth more marks here than being clever on any single item.

Where you will meet it

This is the shape of the Wonderlic cognitive ability test and of Criteria's CCAT, which Lazard sends for its London Spring Insight, internship and analyst applications: the same fifty mixed questions, there with fifteen minutes. Beyond that it is here for the speed rather than the badge: the timed sections of SHL, Watson-Glaser and the numerical and verbal batteries banks send all reward the same habit of deciding fast and skipping without regret.

How to get better

An average score on this format is about 20 out of 50, so do not read a first attempt in the twenties as a bad sign. Two rules move the number more than anything else: never spend more than about twenty seconds on one question, and answer rather than skip when you can rule out two options, because nothing is deducted for a wrong pick. Read the review afterwards, which is where the actual learning is.

Reach a target from four numbers, pick the code that matches exactly, deduce a hidden figure from how many properties you matched, pump a balloon and bank it before it bursts, and click inside a window that keeps shrinking. About nine minutes in total, with a breakdown per game at the end.

• Arctic Shores
• HireVue
• Optiver

What it tests

Switching between unrelated tasks under time, and the specific cost of arriving cold: the first thirty seconds of every game is spent working out the rules, and that is paid five times rather than once.

Where you will meet it

The gamified batteries trading firms screen with, Optiver's Zap-N among them, and the game based assessments banks buy from Arctic Shores and similar.

How to get better

Read the brief before you press start, because the clock does not wait while you work out the rules. On the balloon, expected value peaks about halfway up the range, so bank earlier than feels brave.

Thirty nine balloons in three colours, one at a time. Each pump adds five cents to the balloon; collect and it is banked, pump once too often and it bursts and that balloon's money is gone. Each colour bursts somewhere different and nothing on screen says where. No clock, about four minutes, and a readout per colour at the end.

• pymetrics
• Blackstone
• JP Morgan

What it tests

Risk tolerance and learning: how far you pump the balloons you collect, whether a burst makes you pump that colour less, and how steady your stopping point is. The real game feeds a trait profile rather than a score, so the practice is in having a stop rule per colour before the first balloon.

Where you will meet it

The pymetrics (Harver) battery Blackstone sends for its analyst and summer analyst programmes before its video interview, and that JP Morgan and BCG send too. It is the Balloon Analogue Risk Task from the psychology literature, shortened to thirty nine balloons.

How to get better

Spend the first balloon or two of each colour finding the range, then collect at about half of it and keep to that. The burst pump is evenly spread, so the expected value of collecting after k pumps peaks halfway up the range, and a steady rule beats nerve every time.

Towers (Tower of London, fewest moves), Cards (four decks, eighty draws), Keypresses (two ten second rounds), Easy or Hard (effort against a shown chance of payout), Lengths (long or short, learnt from feedback), Magnitudes (dots and fractions), Stop (press on red, withhold on green), Money Exchange (the trust game, both seats) and Faces (photographs of expressions, ten words, some with a story, seven or thirty seconds each). Every game has a brief screen before it, and none of them pauses once running. About twenty minutes, with a points total and a trait-style readout per game at the end.

• pymetrics
• JP Morgan
• Blackstone

What it tests

Planning, learning from outcomes, effort, impulse control, number sense and trust: the traits the pymetrics battery reports rather than a mark. The real thing cannot be passed or failed, so what practice buys is knowing every game's rules before its clock starts.

Where you will meet it

The pymetrics (now Harver) battery JP Morgan sends for spring weeks and summer analyst roles before its HireVue, and that Blackstone and BCG use. Balloons, Digits and Arrows from the same battery are Balloons, Memory Digits and the Arrow Game.

How to get better

Read each brief properly and sit it at a normal pace. Plan Towers before the first move, follow the balance rather than the win size on Cards, pick hard at 88% on Easy or Hard, and on Stop let accuracy beat speed. Money Exchange has no right answer, so send what you would send.

Balloon, Arrows, Sequence, Tickets, Security Door, Token Search, Power Generator, Order, Balance, Team Sales, Patterns and Prediction, each with a brief before it and a readout after. About twenty five minutes in all. Pumping and banking, flanker arrows with a reversal rule, light sequences past ten, task switching at a fast pace, a dial to stop on a number, a spatial search with no marks, hidden button rules to deduce, figure and scale puzzles, a pricing game against a hidden opponent, odd one out, and a prediction task whose odds shift.

• Arctic Shores
• Barclays
• HSBC
• KPMG

What it tests

The traits the behaviour tests read: risk appetite, attention under distraction, working memory, task switching, learning from feedback, planning, cooperation and calibration. Arctic Shores publishes no marking, so the points here are ours, and the readout says what each task reads as.

Where you will meet it

Arctic Shores' behaviour based assessments, reported at Barclays, HSBC, Citi, KPMG, PwC, Deloitte, Metro Bank, Unilever and Maven Securities, and in the format of the game based screens other providers sell.

How to get better

There is no rehearsal of the real interface beyond one practice task in the invitation, so the value here is knowing each task's shape before it starts. Take the brief seriously, bank early on the balloon until you have seen a few bursts, and work the search tasks in a fixed order so you never revisit.

Respond when a letter repeats the one two back, judge when a marker reaches a line after it disappears, respond under a rule that keeps flipping, and respond to every digit but one. Around eight minutes, with a breakdown per task.

• BrainsFirst
• IMC
• Arctic Shores

What it tests

Updating working memory, anticipation, switching between rules, and whether accuracy holds once responding has gone automatic. The four abilities the neuroscience style assessments are built around.

Where you will meet it

Game based assessments from BrainsFirst, used by IMC among others, and the cognitive batteries banks buy from Arctic Shores and Pymetrics.

How to get better

Every task scores hits less false alarms, so pressing on everything scores zero. The gain from practice here is small and it is all in the first two minutes: knowing the shape of the task means you are not learning it on the clock.

60 seconds of arrows. Answer the direction of the centre arrow while ignoring the ones either side of it, and give the opposite answer when the arrows are red. Correct answers score +1, wrong answers score -1.

• Arctic Shores
• pymetrics

What it tests

Response conflict and inhibition: whether you can follow a rule that fights your first instinct, and how fast you recover after a mistake.

Where you will meet it

Arctic Shores style gamified assessments used across banking and trading, where a flanker task is close to standard.

How to get better

Wrong answers subtract, so accuracy beats speed here. Say the rule to yourself on red rounds, and reset immediately after an error instead of chasing the lost point.

The screen turns green after an unpredictable delay and you click as fast as you can. Five rounds, averaged. Clicking early restarts the round, so there is no gaming the wait.

• Gamified batteries

What it tests

Simple reaction speed, the baseline every timed game is layered on top of.

Where you will meet it

Reflex sections of gamified assessments and trading aptitude screens, and as a warm up before anything timed.

How to get better

Somewhere around 200 to 280ms is a normal adult range. Rest your finger on the button, watch the whole screen rather than one point, and do not try to predict the delay.

Digits flash one at a time, then you type the sequence. It starts at three digits and adds one every time you get it right.

• HireVue
• BAE Systems
• Trading screens

What it tests

Working memory, specifically digit span: how much you can hold while something else demands attention.

Where you will meet it

The digit span game in HireVue's game based assessment, which BAE Systems sends with its video questions, the Digits game in the pymetrics battery, cognitive sections of trading firm assessments, and indirectly any numerical test where you have to carry figures between screens.

How to get better

Chunk the digits in groups of three and rehearse them out loud. Most people plateau around seven to nine digits, and chunking is what gets you past it.

Squares light up one after another and you click them back in the same order. Every level adds one more step.

• Arctic Shores
• pymetrics

What it tests

Sequential and spatial working memory, and attention across a longer stimulus.

Where you will meet it

Gamified assessment batteries of the Arctic Shores and Pymetrics type, and cognitive screens at trading firms.

How to get better

Store the pattern as a shape or a route across the grid rather than a list of positions. Shapes survive in memory far better than lists.

A set of tiles flashes for just over a second, then you click every one of them. Get it right and the grid grows; you have three lives.

• Gamified batteries

What it tests

Spatial recall of information presented all at once, which is a different skill from remembering a sequence.

Where you will meet it

Memory sections of gamified assessments, and the pattern recall parts of graduate aptitude batteries.

How to get better

Take the whole grid in with one soft focus instead of scanning tile by tile. Scanning uses the time you need for encoding.

Numbers are scattered across a grid and vanish once you click the first one. You click the rest in ascending order from memory. Each round adds a number, and three strikes ends the run.

• Gamified batteries

What it tests

Fast encoding of position and value together, under a very short exposure.

Where you will meet it

Working memory sections of cognitive assessments, and it is the single best warm up for the memory games above.

How to get better

Lock in the first four positions before you go hunting for the rest. Getting past nine or ten reliably already puts you ahead of most first attempts.

Multiple choice questions across accounting, valuation, M&A, markets, brainteasers and commercial awareness, with an explanation after every answer and a readiness score per topic.

What it tests

Technical knowledge and recall, not speed.

Where you will meet it

First round and assessment centre interviews, plus the written technical tests some banking and PE processes run.

How to get better

Ten questions a day in whichever topic shows red. The readiness bars are there so you stop practising what you are already good at.

A run of numbers appears and you give the next term. It sits alongside the 80 in 8 rather than replacing it, and the pressure is thinking time rather than raw speed.

Length and marking

Reported at around 15 questions in 25 minutes. Unlike Beat the Odds you can move between questions, so the usual advice is to bank the ones you see immediately and come back.

What we cannot confirm

Counts and clock vary between accounts and years, and Optiver publishes neither. Treat 15 in 25 as a working figure, not a promise. No marking scheme is published for this section.

Closest practice here

Sequences runs the same format against a clock.

Probability under time pressure: expected value, conditional probability and Bayes, binomial questions and the occasional Markov chain. Once you leave a question you cannot return to it.

Length and marking

Reported at about 90 seconds a question, five options, and you pick the closest value rather than computing it exactly. Wrong answers are widely reported to deduct marks, which makes skipping better than guessing when you are stuck.

What we cannot confirm

Three different lengths are in circulation: 10 questions in 15 minutes, 30 minutes at 90 seconds each, and 30 questions at 90 seconds each. They cannot all be right and Optiver publishes none of them. The most recent and most detailed candidate accounts (2025 and 2026) say 30 questions, 90 seconds each, +1, 1 off and 0 for a skip, which is what Probability (Hard) runs. Treat the count as reported, not confirmed.

Closest practice here

Probability (Hard) runs the reported format question for question: 30 questions, a strict 90 second clock on each, five coarse options where you pick the closest, no feedback until the end, with the families candidates describe (coin pairing, grid and polygon random walks, martingales, stopping sums, overshoot). The six minute Probability drill is the gentler place to learn the families first.

An hour of short cognitive games rather than one test. Reported games include a barbecue timing task, a Number Box where you hit a target using four numbers and the four operations, digit recall, task switching and pattern guessing.

Length and marking

About 60 minutes across roughly nine games, each running two to fifteen minutes. It produces a profile rather than a mark, so there is no pass score and no negative marking to play around.

What we cannot confirm

The game list comes from candidate reports and changes year to year, so expect a different set from any list you read, including this one. No scoring, weighting or pass mark has ever been published.

Closest practice here

Battery runs five games of these kinds back to back. Memory Digits is the closest analogue to the digit recall game.

A personality questionnaire, reported at around 150 forced choice pairs where you pick the statement that describes you better. It is profiling, not an ability test.

Length and marking

Untimed. No right answers, no negative marking, no score to improve.

What we cannot confirm

The item count is a candidate figure. Optiver publishes nothing about how it is weighted or whether it is scored at all.

Closest practice here

Nothing to practise and we will not build a drill for it. Rehearsing a personality test mostly produces inconsistent answers, which is the one thing these instruments do reliably detect.

Long form quantitative problems rather than a speed test. Continuous probability, expected value, combinatorics, Bayes, distributions, Markov chains, single variable calculus and logic. Pen, paper and a calculator are allowed.

Length and marking

Reported at 9 to 15 questions in 60 minutes, with the ability to move back and flag. A shorter Quantitative Evaluation variant is reported at 12 to 16 questions in 20 minutes. You get about a week to start once the link lands, and switching tabs can end the attempt.

What we cannot confirm

Which of the two you sit varies by role, region and year. Both counts are ranges in every source we found, not fixed. No negative marking is reported, but none is confirmed either.

Closest practice here

Practice Questions covers the brainteaser end. Nothing here reproduces a 60 minute paper you can work on paper.

Saville aptitude subtests. Numerical reasoning from tables and charts, plus diagrammatic reasoning where a panel tells you what each operator does and you work out either the missing operator or the output.

Length and marking

Saville subtests run 6 to 15 minutes each. A combined numerical and diagrammatic sitting is commonly reported at about 40 minutes. The Swift Analysis version is 18 minutes split evenly across verbal, numerical and diagrammatic.

What we cannot confirm

Prep providers note the battery changes by region and year, so the version you sit may match none of these numbers. We could not verify whether IMC's configuration is adaptive, and no negative marking is published.

Closest practice here

Operators runs the diagrammatic half. Mental Maths covers the arithmetic underneath the numerical section.

A set of neuroscience based games sat alongside the aptitude tests. BrainsFirst describes four games measuring how you process information, learn and focus, and names IMC as a client on its own site.

Length and marking

About 45 minutes in total, each game running roughly 10 to 15 minutes depending on how you play. Scored as a cognitive profile, so there is no pass mark and no negative marking.

What we cannot confirm

BrainsFirst does not publish the mechanics and asks candidates not to. We are not reconstructing them from forum posts, so this entry stays deliberately thin.

Closest practice here

The cognitive games here build the same underlying skills. Chimp Test is the warm up.

Two timed sections: rapid arithmetic, then number sequences. The arithmetic is two and three digit multiplication, percentages and fraction conversions at roughly six seconds a question.

Length and marking

Reported at 60 questions in 6 minutes for the maths and 25 sequences in 26 minutes. Candidates commonly quote a competitive maths score around 72.

What we cannot confirm

The sequences timing comes from prep providers rather than Flow Traders, and the scale behind that score of 72 is never stated, so it is folklore until someone publishes the denominator. Negative marking is not confirmed.

Closest practice here

Mental Maths has this format as its 60 in 6 mode, and Sequences covers the second half.

Typed mental maths with no answer choices, then sequences, then a coding round. Questions arrive one at a time and you can neither skip nor go back, so a question you cannot do costs you the clock as well as the mark.

Length and marking

Reported at 80 questions in 8 minutes typed and 24 sequences in 12 minutes. The coding round that follows is reported at roughly 70 minutes for 4 to 5 problems mixing code and probability.

What we cannot confirm

Reports say the question banks were split by track, so a trader, a quant researcher and a developer see different mixes. No penalty rate is published for the maths.

Closest practice here

Mental Maths has this exact format as its typed 80 in 8 mode.

Maven's Emerging Talent page lists four assessments and says which you sit depends on the programme and business area: a numerical test of mental maths ability (Maven points candidates at Zetamac to practise under timed conditions), a probability test, a HackerRank coding test for technology roles, and an Arctic Shores task-based assessment. For trading candidates the reported sitting is three timed tests: mental maths with no calculator or paper, number and letter sequences, and multiple choice probability on expected value, combinatorics, conditional probability and classic puzzles. Maven publishes one example, an 8-sided die rolled until two 8s come up in a row, expected rolls, with five options.

Length and marking

Reported at 50 maths questions in 5 minutes, 30 sequences in 6 minutes and 18 probability questions in 30 minutes, with several accounts of minus one for a wrong probability answer and nothing for a blank. Candidates from August 2026 describe a newer version: typed answers rather than multiple choice, a shorter probability block of about 20 questions in 10 minutes, and an added estimation question, one unknown quantity in two minutes with an 80 percent confidence interval and the reasoning explained aloud to the proctoring camera.

What we cannot confirm

Maven publishes no counts, clocks or marking; every figure above is a candidate report, the accounts disagree on whether the maths and sequences are typed or multiple choice and on whether wrong answers are deducted, and the older format was still being reported in late 2025, so prepare for both. Which programmes get the Arctic Shores stage, and in what order, varies by cohort and location in the reports. Maven's FAQ says you can apply to one Emerging Talent role a year and reapply the next cycle if unsuccessful.

Closest practice here

The Maven-Style Assessment runs the whole reported sitting in one go: typed maths, letter and number sequences, five-option probability with the deduction, and the estimation interval. Mental Maths has the 50 in 5 as a standalone mode, Sequences and Probability cover the middle sections, Fermi Estimation the last, and Arctic Shores-Style Tasks the game stage.

Jane Street's own description is short: multiple choice and short answer questions. They also say their interviews are designed to need no preparation, which is a statement of intent rather than a reason to sit it cold.

Length and marking

About an hour, per Jane Street's careers site. They publish no question count, no marking scheme and no pass mark.

What we cannot confirm

Candidate reports add specifics: HackerRank hosting, roughly ten questions across four modules, heavy probability and expected value. Jane Street confirms none of it, so it is on this page as hearsay and nothing more.

Closest practice here

Practice Questions for the probability and brainteaser end, Dice Markets for expected value under uncertainty.

Not an aptitude test. Citadel Securities runs Datathons through Correlation One as a route into data and quant tracks: a qualifying quiz first, then teams work a large dataset and present findings to judges.

Length and marking

Citadel Securities publishes a 60 minute qualifying quiz. The team stage runs over the event itself, and no marking scheme is published for either part.

What we cannot confirm

This is the firm we could least verify. Figures circulating for a Citadel quant online assessment come from aggregator sites with no primary source behind them, so they are deliberately absent here.

Closest practice here

No drill here, and a Datathon is not a format you can drill anyway.

One mixed quantitative paper rather than separate sections: mental maths, probability, combinatorics, geometry and some integration.

Length and marking

Reported at 19 questions with roughly two to three minutes each. Marking not published.

What we cannot confirm

This rests entirely on prep site write ups. The same 19 question figure repeats across sources that may all share one origin, so treat it as indicative rather than checked.

Closest practice here

Practice Questions for the probability and geometry, Mental Maths for the arithmetic.

Not a maths test. Algorithmic coding problems in Python or C++, auto graded on hidden test cases, with trading flavoured problem statements over what is really an engineering exercise.

Length and marking

Reported at 3 to 4 problems somewhere between 70 and 150 minutes depending on the role. Auto grading means partial credit comes from tests passed, so there is no penalty scheme to manage.

What we cannot confirm

The time range is wide because different roles sit different papers, and none of it comes from HRT directly.

Closest practice here

Nothing here runs this format. It is an engineering round, so practise on a coding platform with hidden test cases, and rehearse the edge cases graders always include: the empty input, the single element, the case where nothing works.

Aon's scales ix, sat as part of the UBS cognitive assessment alongside a numerical test. Nine tiles of abstract shapes, eight obeying a hidden rule and one breaking it, and you pick the one that breaks it. Items get harder as you go.

Length and marking

Aon's format is most consistently described as 20 items in 5 minutes, about fifteen seconds each, and you are not expected to reach the end. UBS sittings have also been reported at 6 minutes.

What we cannot confirm

UBS publishes neither the item count nor the clock, and the accounts disagree: 20 in 5 minutes, 6 minutes, and 12 in 12 minutes all appear, the last of which is probably the numerical test rather than this one. Nobody agrees on whether wrong answers are deducted either. Aon's own guidance describes answering as many correctly as you can and mentions no penalty, so our version deducts nothing and says so.

Closest practice here

Inductive Reasoning is built on this format, and Data Statements on the numerical test sent alongside it.

Aon (cut-e) tests sent after the application for EMEA and Asia Pacific programmes, spring weeks included. A numerical reasoning test where you judge statements against a set of data tabs as true, false or cannot say, a logical reasoning test, and for some programmes the motionChallenge: a six minute sliding puzzle that measures planning. A recorded video interview follows for those who pass. Applicants in the Americas skip the online tests and go straight to interviews.

Length and marking

The numerical test is most consistently reported at 37 items in 12 minutes across six data tabs, and the logical test at 12 items in 12 minutes. The motionChallenge is six minutes for as many puzzles as you can, scored on puzzles solved and the moves each one took; students who sat it in the 2026/27 cycle describe no reset button, no easy levels and boards of three or four blocks that still need about a dozen moves. Aon's invitation typically gives about five days to complete the set.

What we cannot confirm

Nomura publishes none of the counts or clocks; they come from prep sites and candidate reports. Which tests you get depends on the programme, region and year: prep sites describe the motionChallenge as sent to some candidates, and students applying in the 2026/27 spring week cycle reported sitting it. The 12 in 12 figure may describe the deductive test rather than the inductive one, and nobody publishes a marking scheme for any of the three.

Closest practice here

Motion is built on the motionChallenge format and Data Statements on the numerical test. Verbal Statements covers Aon's scales verbal, which some candidates report alongside the numerical test. Switch covers Aon's deductive test and Inductive Reasoning its inductive one.

Aon (cut-e) numerical and logical reasoning tests for spring weeks and summer analyst roles in EMEA, then a HireVue video interview. The numerical test is the six tab true, false or cannot say format; the logical test is reported as scales ix (inductive) for some divisions and the deductive switchChallenge for others.

Length and marking

Numerical reported at 37 statements in 12 minutes, logical at 12 items in 12 minutes or 20 in 5 depending on which Aon test the division sends. No marking scheme is published.

What we cannot confirm

Morgan Stanley publishes none of it. Which logical test you get, and whether there is one at all, varies by division and year in candidate reports.

Closest practice here

Data Statements is built on the numerical format. Inductive Reasoning covers scales ix and Switch covers the switchChallenge.

SHL Verify Interactive General Ability (G+): one adaptive sitting that mixes numerical reasoning from charts and tables, inductive figure sequences and deductive scheduling puzzles, answered by dragging and setting values rather than picking from a list. A Mindset questionnaire follows, then the Business Insight stages with a situational judgement flavour.

Length and marking

Reported at 24 questions in about 36 minutes for the cognitive test and 105 statement pairs in 20 minutes for the Mindset questionnaire, 60 to 90 minutes in total. No negative marking is published.

What we cannot confirm

Barclays has used both SHL and Cappfinity in recent cycles and prep sites disagree on which stage carries which. The counts are candidate figures. The interactive answer format cannot be reproduced here without copying SHL's interface, so the drills use the multiple choice version of the same questions.

Closest practice here

Numerical Reasoning for the numerical section, Figure Series for the inductive one, Logical Statements and Operators for the deductive one, Situational Strengths for the insight stages.

One untimed sitting of about 69 questions in three blocks: situational judgement scenarios set at the firm, answered by ranking the responses from the one you would most likely do to the one you would least likely do (older cycles picked the most and least likely of four); about forty behavioural pairs of opposing statements with a slider between them; and twelve numerical questions on charts and tables, one per page. Rothschild sends it for spring insight, summer and graduate programmes, and one assessment is reused across applications in the same year.

Length and marking

Rothschild says it is not timed and takes 30 to 45 minutes; the instructions candidates quote say allow 45 minutes in one sitting. Cappfinity records the time on every answer and it feeds the result alongside accuracy. Candidates in the 2025 and 2026 cycles call the numerical part easy and the situational and behavioural parts the hard ones. No marking is published.

What we cannot confirm

Rothschild does not name the provider. The untimed, slider-based format is Cappfinity's and candidates compare it to Cappfinity; some prep sites still describe a timed SHL numerical test from an older cycle, so the provider is reported, not stated. The split of the 69 questions and whether any verbal or logical items appear vary by account.

Closest practice here

Capp-Style Assessment runs all three blocks in one sitting. Situational Strengths for more scenarios on their own, Numerical Reasoning for the numbers on their own.

Discover EY, the first-year insight programme students call the EY spring week, has three steps: a short application form with no CV, the online assessment, and the programme itself, where the final day is assessed for a fast track to a summer internship interview. Up to the April 2026 intake the online stage was Cappfinity's: a strengths assessment of workplace scenarios ranked from the thing you would most likely do to the thing you would least likely do, or answered as a single pick, with sliders between two statements about how you work; then a job simulation of about 14 items built on emails, a news article about a fictional company and a few charts, mixing ranked scenarios, short written replies to a client or manager (about 1,000 characters), chart and table questions with a calculator, and two or three recorded video answers on your motivation and strengths, each with two or three minutes to prepare and two to answer, no retakes.

Length and marking

Untimed with the time recorded, in one sitting EY recommends doing on a laptop in Chrome; candidates report 45 to 90 minutes across the two parts, with the video answers the only clocked items. EY's FAQ for the 2027 cycle says the assessment will have a time limit stated in the invitation and lists the formats: choose from options, interpret a table, chart or graph, drag items into position, spot a pattern or apply a set of rules, record a video response. A score is produced for each assessment and mapped against set criteria; below the pass score the application is rejected without human review, and the video answers are scored by people. Adjustments, including up to 25 per cent extra time, must be requested before the assessment is started.

What we cannot confirm

EY never publishes the assessment. Everything above the FAQ's list of formats is candidate and prep site reporting of the Cappfinity version, and EY's early careers privacy notice, updated in September 2026, names SHL as the platform for the cycle now open, with a realistic job preview, a first online assessment covering teamwork and learning, and a behavioural skills test. No first-hand account of the SHL version exists yet, so the counts, the split between parts and whether first years get the numerical questions are reported, not stated. The student version in 2025 is reported without a separate numerical section; graduate versions add one and, for some service lines, a critical thinking block. One application per six months, and a rejection means a six month wait.

Closest practice here

EY-Style Assessment runs the scenarios, the sliders and the chart questions in one sitting, with the scenarios set at a professional services firm and the key argued from EY's eight skills. Situational Strengths for more scenarios on their own, Numerical Reasoning for the charts, Figure Series for the pattern items the FAQ lists, and Firm Prep for the written and video answers, which no drill here scores.

One sitting of about 35 to 40 minutes in three parts, and the same package for Spring Insight, summer internship, analyst and industrial placement applications in London. A 15 minute aptitude test with no calculator, which candidates and prep sites identify as Criteria's CCAT: 50 mixed verbal, numerical, logical and spatial questions that get harder as the paper goes, with nothing deducted for a wrong answer. An untimed behavioural trait questionnaire of about ten minutes, rating how well each statement describes you (Criteria's Employee Personality Profile, about 140 items, with a Workplace Insights Report emailed to you afterwards). Then a recorded video interview of about three questions on your motivation, Lazard and the markets, with 15 to 30 seconds to prepare and up to two minutes to answer each, and no retakes.

Length and marking

Lazard says the aptitude test is 15 minutes, the behavioural part untimed at about ten minutes, and the whole sitting 35 to 40 minutes in one go, and tells you to guess and move on when stuck because nothing is deducted. On the CCAT the average candidate gets about 24 of 50 right and under one percent finish, so 30 or more is a strong run. Spring Insight then goes straight to one 30 minute online interview with a banker and someone from HR; there is no assessment centre. Anyone who sat the aptitude or behavioural tests in the last 12 months is not asked to retake them on the analyst, intern and placement tracks, only the video. The 2027 Spring Insight window is 19 October to 15 November 2026, reviewed as applications arrive.

What we cannot confirm

Lazard names Criteria for technical support but never says which Criteria tests it buys; the CCAT and the personality profile are the candidate and prep site reading, backed by the format and the Workplace Insights Report. Lazard publishes no question count and no pass mark. Some 2025 reviews mention situational judgement, which no current account describes and is most likely the behavioural questionnaire; older walkthroughs describe a pre-2024 format with scenarios and a written task that no longer applies. Video question counts and preparation times vary by track and year. Lazard allows one London application a year, and Spring Insight or Industrial Placement, not both.

Closest practice here

Wonderlic-Style Test is the same shape, fifty mixed questions against a clock, three minutes faster than the real paper. Mental Maths and Sequences for the arithmetic and number series without a calculator, Logical Statements for the deduction items, and Firm Prep for the recorded interview.

A Cappfinity immersive assessment: you take on a role and work through a business scenario that mixes strengths based situational judgement with embedded numerical, verbal and inductive questions, followed by a job simulation and a written or recorded task. Lloyds and NatWest run Cappfinity assessments of the same family.

Length and marking

Untimed but speed scored: Cappfinity records how long each answer takes and feeds it into the result alongside accuracy. The numerical part is reported at about 12 chart and table questions. Nothing about the marking is published.

What we cannot confirm

The scenario itself changes by programme and year. The strengths key behind the situational questions is never published, so any practice key, ours included, is a judgement about what the frameworks reward rather than HSBC's answers.

Closest practice here

Capp-Style Assessment for the situational and numerical core in one sitting, Situational Strengths for the judgement questions on their own, Numerical Reasoning for the numerical ones, Figure Series for the inductive ones.

SHL ability tests (numerical, verbal and inductive or deductive reasoning) with a behavioural questionnaire, and at Deutsche Bank a situational judgement test that candidates report as the deciding stage for spring weeks.

Length and marking

SHL Verify numerical is reported at 10 questions in 18 minutes in its interactive form, inductive at up to 15 in 18 minutes, verbal at 30 in 19 minutes with true, false or cannot say answers. Neither firm publishes which tests a given programme gets or the marking.

What we cannot confirm

Both firms change the mix by programme and year, and Deutsche Bank's spring week is widely reported to lean on the situational judgement test rather than the ability tests. The verbal test is not reproduced here.

Closest practice here

Numerical Reasoning, Figure Series and Situational Strengths cover three of the four parts. Logical Statements is the nearest thing to the deductive items.

Two parts in one sitting. First an untimed behavioural section where you pick the behavioural skills that best describe how you work. Then a timed reasoning section built from three SHL tests: deductive reasoning (scheduling and logic puzzles), inductive reasoning (figure sequences and patterns) and numerical reasoning (calculations from tables and charts), in the interactive format where answers are dragged and set rather than picked from a list. Technology applicants then get a two-problem Coding Simulation in a browser IDE, and everyone who passes records the Workplace Scenarios video interview: four workplace scenarios, one question each, one minute to prepare and two to answer.

Length and marking

Standard Chartered's early careers FAQ publishes the clock: the behavioural section is untimed and takes 15 to 20 minutes, the reasoning section is exactly 36 minutes and cannot be paused, and the whole thing wants about an hour. The Coding Simulation runs 45 to 85 minutes depending on the role. The reasoning score is compared with a level set in advance and you progress if you meet it; the behavioural section produces a feedback report sent to everyone, pass or fail. Invitations give a few days to complete it.

What we cannot confirm

Until the 2026 intake Standard Chartered sent the twelve pymetrics games and a Valued Behaviours Assessment, and most prep sites, and an older page of the bank's own, still describe that process. The bank's current apply page and FAQ name the Skills Insight Assessment as the first stage and SHL as its provider, so expect SHL. The FAQ warns of minor variations by market and programme, and the question count in the 36 minutes is not published. The interactive answer format cannot be reproduced here without copying SHL's interface, so the drills use the multiple choice form of the same questions.

Closest practice here

Numerical Reasoning, Figure Series and Logical Statements cover the three reasoning tests, and SHL's own free practice tests (the Behaviour-Based Assessment and the General Ability Test, Mobile-Enabled, Interactive) are the ones the FAQ points at. Situational Strengths is the nearest thing to the behavioural section, and Firm Prep covers the Workplace Scenarios questions.

Twelve short behavioural games played in the browser, then a HireVue video interview. The games are Balloons, Keypresses, Digits, Arrows, Lengths, Cards, Towers, two Money Exchange rounds, Easy or Hard, Stop and Faces. Blackstone and BCG send the same battery; Standard Chartered did until it moved to SHL's Skills Insight Assessment.

Length and marking

About 20 to 25 minutes in total, two minutes or so a game. There is no pass mark: the games produce a trait profile that is matched against the firm's model, and a result is kept for twelve months and shared with any other employer using pymetrics.

What we cannot confirm

Pymetrics publishes the game list but not the traits each game feeds or how the match is scored, so no practice can tell you what a good profile is. What practice does is take the rules of each game off the clock, which is the only part of the score you control.

Closest practice here

Pymetrics-Style Games runs nine of the twelve in one sitting, Faces included, on stock photographs of posed expressions. Balloons is the balloon game on its own, Memory Digits the digit span, Arrow Game the arrows.

Talent Q Elements numerical and logical reasoning for the spring week: adaptive tests where a right answer makes the next item harder, with a clock per question rather than per test. Some Citi roles send the Plum Discovery Survey instead, a 25 minute mix of personality items, two short figure series sections and workplace situational questions.

Length and marking

Elements is reported at 12 items per module with roughly 75 to 90 seconds each and a 15 minute cap. Plum is reported at about 25 minutes with around 7 puzzles in each problem solving section, untimed but speed recorded.

What we cannot confirm

Citi publishes neither which programme gets which test nor any marking. Elements' adaptivity means two candidates see different items, so a count of questions says little about difficulty.

Closest practice here

Numerical Reasoning and Figure Series cover Elements and Plum's puzzles. Situational Strengths is the nearest thing to Plum's workplace section; its personality sections cannot be practised.

The Suited assessment used by several elite boutiques for insight and summer programmes: timed logic (given two or three statements, is a further statement always, sometimes or never true), number series, a checking section (do two strings match), and a behavioural questionnaire.

Length and marking

Candidates report about 15 minutes for the timed sections, with an untimed example before each. Suited publishes no counts, no clock and no marking, and says the result is one data point alongside interviews.

What we cannot confirm

Which boutiques use Suited changes year to year; PJT, Evercore, Houlihan Lokey and Guggenheim are the ones candidates report most. Moelis is reported to use Test Partnership instead. None of it comes from the firms.

Closest practice here

Logical Statements is built on the logic section. Sequences covers the number series and Battery's Code Compare the checking section.

No aptitude test for the front office spring weeks. Both send a HireVue recorded interview: three to five questions, about 30 seconds to prepare and 90 seconds to two minutes to answer each, on motivation, teamwork and a market or problem solving prompt. Bank of America's version is reported to add situational judgement and written case modules for some divisions.

Length and marking

Around 20 minutes end to end. Marked by recruiters watching the recordings; neither firm publishes the criteria.

What we cannot confirm

Some Bank of America tracks (technology, wealth) are reported to sit SHL or HackerRank tests instead, and Goldman's Americas process differs from EMEA. Treat the question list as typical rather than fixed.

Closest practice here

Firm Prep rehearses HireVue questions by firm with marking. Situational Strengths covers the judgement modules.

Aon's deductive reasoning test, also called scales sx. Two rows of symbols, what went in and what came out, and a four digit operator that rearranged one into the other. You name the operator, not the output, and the later items chain two operators and hide one.

Length and marking

Time boxed rather than a fixed question count, and adaptive: your score is what you get right inside the window, so two people can see a different number of items. Most accounts say six minutes, some say five.

What we cannot confirm

Accounts vary on the clock and on whether wrong answers are deducted. Aon does not publish BNP Paribas's configuration, so our version picks one and says so on the page.

Closest practice here

Switch is built on this format.

Three assessments inside one Aon chatAssess sitting, sent after the application form (there is no CV): a situational judgement exercise answered as messages to colleagues, a work style questionnaire, and one timed logical puzzle. Rolls-Royce's own guide says the set depends on the programme and that engineering and business routes differ; older guidance listed a six minute numerical test for engineering.

Length and marking

Four days to complete from the invitation (Rolls-Royce's September 2026 application guide), one sitting, one attempt per cycle, and under 45 minutes in all. The situational exercise and questionnaire are untimed; the logical puzzle is six minutes, with an untimed practice task before it. A downloadable report afterwards describes your speed and accuracy against other candidates.

What we cannot confirm

Which six minute puzzle it is: prep sites and most candidate accounts say Aon's switchChallenge, and an applicant in the 2026/27 cycle reported Aon's gapChallenge, the grid with the question mark. Rolls-Royce names neither. Practise both and read the explanation screen before the timer starts, which several candidates say is the thing to do.

Closest practice here

Grid Gap is built on the gapChallenge and Switch on the switchChallenge.

One HireVue sitting after the application form, which BAE's adverts describe as an online assessment with interactive activities and an on-demand video interview, and its apprenticeship pages as gamified challenges and video questions. Candidates describe three or four recorded questions on motivation, teamwork and problem solving, with each question shown before you record, then three short games at the end that they call fairly simple: a maths one, a puzzle one, and one where you pick which of two pictures you lean towards. A live competency interview, virtual or face to face, follows for those who pass, and some schemes add an assessment centre.

Length and marking

Reported at 30 to 45 minutes in all. The video questions carry about 30 seconds to prepare and two to three minutes to answer, one applicant reports three attempts at each, and the games run a few minutes apiece. BAE's graduate FAQ has allowed one application per academic year, and the same sitting serves graduate schemes, summer internships and the apprenticeship levels, with the process varying by role. Offers depend on security clearance (Baseline Personnel Security Standard at least), which is checked after the interview rather than here.

What we cannot confirm

What the three games are: candidates name them by content rather than title, and the only source naming them (Shapedance for pattern recognition, Digitspan for working memory, Portrait for reading faces, all from HireVue's game based assessment) is a prep site, so treat those as likely rather than confirmed. Whether a separate SHL numerical and verbal test still runs for graduate roles: prep guides list one, older forum posts describe one, BAE's current adverts do not mention one and recent candidates report only the HireVue sitting. BAE publishes no question counts, no clock and no pass mark, and several candidates say it gives no feedback after the test. BAE's AI policy allows generating practice questions and forbids using AI to complete the assessment or answer during interviews.

Closest practice here

Firm Prep rehearses recorded questions with marking, which is the part that carries the weight. Memory Digits is the digit span game, Battery trains arriving cold at short games whose rules change each time and opens with a numbers to target round, and Mental Maths covers the arithmetic. Nothing here scores the picture preference game, which reads personality rather than ability, so answer it as yourself.

One untimed online sitting after the short application form (AtkinsRéalis takes no CV or cover letter for early careers roles), the same for graduates, placements and apprentices. The selection pages describe interactive exercises, including cognitive tasks on how you approach problems and process information; the firm's 2024 preparation blog describes a day in the life: situational judgement questions, numerical and verbal reasoning questions, and three short video answers at the end. Candidates describe scenarios answered as a most and least pick of four, maths read off emails and graphs with a calculator allowed, true, false or cannot say verbal items, and in accounts to 2025 three recorded video questions. A live interview on Microsoft Teams with two interviewers follows, usually with a short business presentation to prepare; summer placements get a matching call with the business instead of an interview.

Length and marking

About 45 minutes by AtkinsRéalis's own figure and up to an hour by its blog, in one sitting, with no clock on any step; prep sites add that speed still counts towards the cognitive score. Prep sites put it at about 25 items, around 12 situational, around 10 numerical and verbal, and up to 3 written or video questions, with two minutes to record each video answer and one take. One application per recruitment campaign, a practice version on the Candidate Preparation Zone, extra time and a change of format on request, and an answer within two weeks of the interviews closing.

What we cannot confirm

Whether there are video questions: the current selection and FAQ pages say there are no video interview questions in this assessment, the firm's own 2024 blog and candidate accounts to 2025 describe three at the end, and a 2026/27 prep guide calls the video a separate part two. Prepare for them and do not be surprised when they are absent. The provider: AtkinsRéalis names none, the practice zone runs on a Preparation Plus domain, and prep sites name Amberjack, whose blended Future Potential assessment is the same shape; neither names the other. No counts, marking or pass mark are published. Applicants in the 2026 cycle reported a business review stage after passing, with waits of up to three months before an interview, which the official pages do not mention.

Closest practice here

The Capp-Style Assessment is the nearest shape: untimed scenarios as a most and least pick, then chart questions without a clock. Situational Strengths drills the ranking, Numerical Reasoning the tables and charts, Verbal Statements the true, false or cannot say items, and Firm Prep the Teams interview and any recorded questions.