The quietest filter in Bain recruiting. No countdown on screen, no score at the end, and most candidates finish with no idea whether they passed. What the five sections contain and how the scoring actually works.
Updated September 202612 min readChecked against 2025 and 2026 candidate reports
5 sectionsNot every sitting includes all of them
~15 itemsPer cognitive section
60-75 minTypical full sitting
70th+Reported shortlisting percentile
Overview
What the Bain SOVA assessment is
SOVA is an assessment platform, not a Bain invention. Bain licenses it and configures which modules candidates see, which is why two people applying in the same cycle can describe slightly different tests and both be telling the truth. What stays constant is the shape: a set of multiple choice reasoning blocks wrapped around a behavioral questionnaire, delivered in the browser, usually within a few days of submitting your application.
The structural difference from McKinsey Solve or BCG's Casey chatbot matters more than it first appears. Solve builds a game around ecosystem data. Casey runs a case in a chat window. SOVA runs a traditional psychometric battery and asks whether your raw reasoning and your working style clear the bar.
What this means for prep: case practice will not move your SOVA score. This is a different muscle from structuring and driving a case, and the candidates who struggle are usually the ones who assumed their case prep covered it.
Before you prepare anything, confirm which test you are actually taking. Bain assigns assessments by office and by role, and some processes now use TestGorilla or a one way HireVue video instead. The invitation email names the platform, and preparing for the wrong one is the most expensive mistake available at this stage. The Bain deadlines page shows where in the cycle the assessment falls.
Format
The five sections of the SOVA test
A full sitting runs about an hour to an hour and fifteen, with the work style block taking a disproportionate share of that because of its item count. Cognitive sections typically run around 15 items each. Nothing counts down on screen, but every section records how long you take.
sova-sections.fig
SECTION 1Situational judgmentWorkplace scenarios. Pick the most effective and the least effective response, so each item is two decisions.
SECTION 2Verbal reasoningTrue, False or Cannot Say against a dense business passage. Missing from some 2026 sittings.
SECTION 3Numerical reasoningCharts and tables with multiple choice arithmetic. Simple math, unforgiving distractors.
SECTION 4Logical reasoningAbstract sequences and matrices with a hidden rule. The least practiced section.
SECTION 5Work styleForced choice statements, most and least like you. No right answers, but a scoring model underneath.
Section presence varies by sitting. Several 2026 candidates have reported a verbal block missing entirely, running numerical, logical and personality only.
Untimed is not the same as unhurried
The absence of a visible timer reads as permission to take your time. It is not. SOVA records how long you spend per item, and response time contributes to your result alongside accuracy. A candidate who is 90 percent accurate at a normal working pace can outrank one who is 95 percent accurate but took twice as long.
Interactive
Try a question from each section
Below is each section with what it measures and a live sample item in the real answer format. These are original examples written in the style of the test, not real SOVA content. Work through one before reading the explanation: the gap between what you thought the question was testing and what it was actually testing is the most useful thing on this page.
SOVA section explorerpractice mode · untimed
Situational judgment
Short workplace scenarios with four or five possible responses. You pick the most effective and the least effective, which means each item is really two decisions.
Scenario basedClient and team dilemmas, no consulting knowledge needed
Most and leastTwo marks per item, so half credit is common
Judged on fitScored against Bain's values, not general politeness
Scenario
You are two days from a client steering committee. Reviewing the model overnight, you find an input error in a section your manager built. Correcting it moves the headline savings number down by about 15 percent. Your manager is on a flight and unreachable until tomorrow afternoon.
Which response is most effective?
Most effective: A. Least effective: C.
A does the two things SJT scoring rewards together: it protects the accuracy of the client deliverable and it keeps the person who owns the work informed. Doing one without the other is what separates a full mark from a half mark.
B is not unethical, it is just slow. It burns a day of the two you have and leaves a known error in a client model. D skips your own team entirely and puts an unverified concern in front of the client, which is why it usually scores below B. C is least effective because it hides a material change from the person accountable for it, and concealment is the behavior these items are built to catch.
Verbal reasoning
A dense business passage followed by statements you mark True, False, or Cannot Say. Some 2026 sittings have dropped this section, so treat it as likely rather than certain.
Closed worldJudge only against the passage, never outside knowledge
Cannot SayThe option candidates under-use and lose most marks on
Sometimes absentReported missing from several 2026 sittings
Passage
A regional grocery chain opened 40 convenience format stores last year, taking its total to 310. Management reported that convenience stores reached profitability in an average of 11 months, compared with 19 months for its full size supermarkets. The company has stated it will prioritize convenience openings in the coming year.
Statement: convenience stores are more profitable than the company's supermarkets.
Correct: C, Cannot Say.
The passage tells you convenience stores reach profitability faster. Speed to profitability and level of profitability are different quantities, and nothing here compares margin or absolute profit. A store that breaks even in 11 months can still earn less than one that takes 19.
This is the highest yield habit in the section: check whether the statement swaps one metric for a neighboring one. Faster is not larger, prioritized is not committed, and reported is not verified.
Numerical reasoning
Charts and tables with multiple choice arithmetic on top. The math is straightforward. The difficulty is reading the right cell and choosing an estimate route fast enough.
Data firstPercentages, ratios, growth rates, weighted averages
CalculatorUsually permitted, but options are spaced for estimation
Trap optionsWrong answers match common misreads of the table
Exhibit
Revenue by division, in millions of dollars
Division
2024
2025
Industrial
420
462
Consumer
310
372
Services
170
166
Which division grew fastest in percentage terms, and by roughly how much?
Correct: B.
Consumer moves from 310 to 372, a rise of 62. Sixty two over 310 is exactly 20 percent. Industrial rises 42 on 420, which is 10 percent. Services falls.
Notice how the wrong options are built. A is correct arithmetic on the wrong division. D reports the absolute change as if it were a percentage. C is what you get if you divide the change by the new figure instead of the old one, the most common slip under time pressure. Anchoring on the base year is the whole discipline here, and our math drills run this exact format.
Logical reasoning
Abstract sequences and matrices governed by a hidden rule. This is the section most candidates have never practiced, which makes it the biggest score differentiator in the test.
Rule findingRotation, shading, count, position, or side number
EscalatingItems typically get harder as the section runs
One rule at a timeIsolate a single attribute rather than reading the whole tile
Sequence
?
Which figure comes next in the sequence?
Correct: A.
Two independent rules run at once. Side count increases by one each step: three, four, five, so six. Fill alternates: solid, outline, solid, so outline. Answer A satisfies both.
The method that works under pressure is to test one attribute across the whole row before looking at any other. Candidates who read each tile as a whole picture stall on items where two simple rules are interleaved. If no rule surfaces within about 30 seconds, mark your best guess and move, because completing the section matters more than winning any single item.
Work style and personality
The longest block by item count. Forced choice statements where you pick which is most and least like you. There is no correct answer, but there is a scoring model.
Forced choiceYou cannot rate everything highly, so trade-offs are unavoidable
Consistency checkedRelated statements repeat in different wordings
Values mappedScored against Bain traits, not a generic profile
Which statement is most like you?
There is no key here, but the mapping is not random.
A maps to bias toward action and the answer first habit Bain trains into consultants. C maps to team orientation, the trait behind the internal line that a Bainie never lets another Bainie fail. B maps to independent working, which is not penalized but reads weaker against a collaborative norm group. D maps to structure preference, useful in delivery roles and neutral to slightly negative for ambiguous problem solving.
Two things follow. Answer honestly, because forced choice designs are built to surface inconsistency across repeated items and a constructed persona falls apart over 80 or more statements. And where two statements are genuinely equal for you, the tiebreak that favors action and team reads better against this norm group.
In a real sitting these would arrive back to back with no feedback between them, and the clock you cannot see would already be running.
Scoring
How the SOVA assessment is scored
Bain publishes nothing about SOVA scoring, so what follows is assembled from how the platform works, how comparable psychometric batteries are scored, and consistent candidate reporting. Treat it as a well supported model rather than an official document.
Cognitive sections are percentile ranked
Your raw accuracy on verbal, numerical, and logical is converted into a percentile against a norm group of other applicants. That is why a section that felt hard may still score well: everyone else found it hard too. It also means chasing a perfect raw score is the wrong target. Consistent accuracy across all three sections beats a spike in one and a collapse in another, because a weak section drags the composite down more than a strong one lifts it.
Speed counts even though nothing counts down
SOVA records how long you spend per item, and response time contributes to your result alongside accuracy. Practically, that means working at a deliberate crawl is a scoring decision rather than a safe one.
What this changes about your prep: do not practice untimed just because the real test looks untimed. Set a target pace per item, around 45 to 60 seconds for numerical and 30 to 40 for logical, and treat overrunning it as a mistake in its own right.
Work style is scored on alignment, not correctness
The personality block does not produce right and wrong answers. Each statement is keyed toward a trait, and your responses are compared against the profile the norm group looks like. An internal consistency check runs underneath, which is what makes trying to fake a profile risky rather than clever.
Benchmark
What score do you need to pass?
There is no published cutoff, and the practical bar moves by office, by role, and by how strong the applicant pool is that cycle. The most consistent reporting puts shortlisting somewhere in the top 20 to 30 percent of candidates, which is the 70th to 80th percentile or better. Some offices also use SOVA alongside the first round rather than as a hard gate before it, in which case a middling result is survivable if the rest of your application is strong.
percentile-bands.fig
85th plusClear pass at almost any office
70th to 84thLikely through, the working target
50th to 69thBorderline, decided by the rest of your file
Below 50thBelow typical screening range
Bands reflect reported outcomes rather than an official Bain rubric. Treat the 70th percentile as a planning target, not a guarantee.
Strategy
Section by section strategy
Section
Where marks are lost
What to do instead
Situational judgment
Picking the most agreeable option rather than the most effective one, and treating most and least as one decision
Ask which option protects both the client outcome and the team relationship. Choose least effective separately, looking for concealment or escalation over someone's head
Verbal
Marking True on statements that are plausible but not stated, and avoiding Cannot Say
Judge strictly against the passage. If a statement swaps one metric for a neighboring one, it is Cannot Say
Numerical
Dividing by the wrong base, reading the wrong row, and over-calculating when options are far apart
Identify the base year first, then estimate. Check option spacing before reaching for precision
Logical
Trying to read the whole tile at once, and burning three minutes on one hard item
Test one attribute across the row at a time. Cap yourself at 30 to 40 seconds, then guess and move
Work style
Constructing an ideal consultant persona, which fails the consistency check
Answer honestly. Where two statements feel equal, favor the one showing bias to action or team orientation
Pitfalls
The five most expensive SOVA mistakes
Preparing for the wrong test
Bain runs SOVA, TestGorilla, and HireVue video across different offices and roles. Read the invitation email before you spend a weekend on the wrong format.
Treating untimed as unhurried
Response time is recorded and feeds the score. Working at a deliberate crawl is a scoring decision, not a safe one.
Skipping logical reasoning in prep
It is the section fewest candidates have ever practiced and therefore the one where a few hours of work moves your percentile most.
Gaming the personality section
Forced choice items with repeated themes are designed to expose an invented profile, and inconsistency reads worse than an ordinary honest one.
Sitting it on a phone
Data tables and shape sequences need a real screen. Use a laptop, a quiet room, and a browser with nothing else open.
Preparation
How to prepare for the SOVA assessment
SOVA rewards two separate things: accurate reasoning at a working pace, and familiarity with question formats most candidates have never seen. Two weeks of light, consistent work covers both.
Days 1 to 2. Confirm the format
Check the invitation, note the deadline, and ask the recruiter whether a video component is attached. Sit the free situational judgment section to see the interface cold.
Days 3 to 6. Numerical, at pace
Twenty minutes a day of chart and table work with a stopwatch running. Percentage change, ratios and weighted averages cover most of what appears. The math drills are built for this.
Days 7 to 9. Logical
The unfamiliar section, so give it dedicated blocks rather than fitting it around the edges. Practice naming the rule out loud before choosing an answer.
Days 10 to 11. Verbal
Read dense business writing and drill the True, False, Cannot Say discipline until Cannot Say stops feeling like a cop out.
Days 12 to 13. Full mock
One complete sitting, all five sections, one go. Review every item you got wrong and label which of the five mistakes above it was.
Day 14. Light
No new material. Confirm your setup, then leave it alone.
If you clear the assessment, the next stage is candidate led cases and the Experience Interview, which is a different discipline entirely. The Bain interview guide covers that round, the case library has the practice cases, and Bain salary data has what the offer is worth by level and region.
FAQ
Bain SOVA: frequently asked questions
Is the Bain SOVA test timed?
There is no countdown on screen, but the platform records how long you spend on each item and that time contributes to your result. Functionally it is timed, it just does not show you the clock. Work at a brisk, sustainable pace rather than a deliberate one.
How long does the Bain SOVA assessment take?
Most candidates report between one hour and one hour fifteen minutes for a full sitting, with the personality block taking a large share because of its item count. Sittings that omit verbal reasoning run shorter.
Can I use a calculator on the numerical section?
A calculator is generally permitted and candidates commonly use their own. Answer options are often spaced widely enough that estimation is faster than exact calculation, so reaching for the calculator on every item costs time you are being scored on.
Does every Bain office use SOVA?
No. Assessment choice varies by office and by role, and several processes now use TestGorilla or a one way HireVue video instead. Your invitation email names the platform, and your recruiter can confirm if it is unclear.
Does the verbal reasoning section still appear?
Not always. Several 2026 sittings have been reported without a verbal block, running numerical, logical and personality only. Prepare for it lightly rather than skipping it, since configuration can differ between offices in the same cycle.
Can you retake the Bain SOVA assessment?
Not within the same application. If your application is unsuccessful you would normally reapply in a later cycle, and firms typically ask candidates to wait a defined period before applying again. Confirm the waiting period with the office you applied to, since it is set locally.
Where can I practice the SOVA format?
Our SOVA full mock recreates all five sections, the hidden clock and the no-feedback-until-the-end rule, with the situational judgment section free. The add-on also opens the TestGorilla sprint.
Both Bain assessments, rebuilt faithfully: the SOVA full mock and the TestGorilla sprint. The first section of each is free, the full suite comes with the add-on.
