Skip to main content
Back to Blogs
Provider Guide10 min read

BCG and Bain Online Assessments: Casey, Sova and What Each One Rewards

BCG is reported to run a chatbot case called Casey. Bain a Sova battery with no clock. What each rewards, and how both differ from McKinsey Solve.

The short version
The three MBB assessments have almost nothing in common. There is no such thing as preparing for "the MBB online test".
BCG is widely reported to run a chatbot case called Casey: roughly half an hour, under ten questions, ending in a one minute video recommendation.
Bain is widely reported to run a Sova battery with no countdown on screen, and coaches say your response times are part of the score.
Both firms vary the stage by office and role. Bain reportedly skips it for some roles and uses TestGorilla instead of Sova in others.
None of the three publishes specifications, so everything below is attributed. Your recruiter can settle it in one email.

McKinsey, BCG and Bain get grouped into one acronym and then, by candidates applying to all three, into one revision plan. That does not survive contact with the assessments. McKinsey runs a single bespoke simulation more or less everywhere. BCG runs a conversational case in most offices and something shorter elsewhere. Bain runs a standard psychometric battery in some offices, a different vendor in others, and in some roles nothing at all.

First, the honest position on sourcing

None of the three firms publishes a specification for its online stage. BCG's own case interview preparation page on careers.bcg.com describes the live case interview and links its OneDay@BCG simulation, and mentions no online case, chatbot or test stage at all. Sova publishes candidate guidance naming no clients. Bain publishes nothing.

So every number below comes from coaching firms and preparation sellers. We name them inline, because the useful question is not “is this true” but “how many independent people saw the same thing”. Where they disagree, we say so rather than picking the tidier number. A reply from your recruiter outranks all of it, and asking what the online stage contains is an ordinary question that is almost always answered.

BCG: the Casey chatbot case

Casey is not an aptitude test. It is a case interview delivered through a chat window, with a bot asking the questions instead of a consultant. You work a single business problem from framing through to recommendation, and the interface looks closer to a messaging app than to a test platform. The strongest reporting on it comes from StrategyCase, MConsultingPrep, CaseBasix and MyConsultingCoach, and they broadly agree on the shape while differing on the exact numbers.

Length
MConsultingPrep and CaseBasix say 25 to 30 minutes. StrategyCase says about 30, MyConsultingCoach about 35. Plan for half an hour of continuous work.
Question count
MConsultingPrep and CaseBasix say 6 to 8. StrategyCase and MyConsultingCoach say 8 to 10. Either way it is a small number of high-weight questions, not a sweep.
Answer formats
Single and multi-select multiple choice, a numeric entry box, and short free text. MConsultingPrep caps the free text at four to six lines and puts multi-select at 40 to 60 per cent of questions.
The video at the end
A one minute spoken recommendation. CaseBasix and MConsultingPrep both describe 60 seconds to prepare and 60 seconds to deliver. It sits outside the main clock.
Tools
A calculator is permitted. StrategyCase says Excel is too, and that scratch paper is fine. Confirm this in your invitation rather than assuming it.
Revisiting answers
Reported as final on submission, with no way back. CaseBasix advises assuming pausing is unavailable as well.
Wrong answers
No penalty reported by StrategyCase or MConsultingPrep. Nothing should be left blank.

What Casey rewards shows up in the first question rather than the last. Candidates report that the case often opens by presenting a set of possible data sources and asking which ones matter before any analysis happens. That is a structuring question wearing a multiple-choice costume, and it is a signal BCG could not extract from a numerical reasoning item.

The arithmetic then has to be right first time, because you cannot go back and repair it. A chained calculation starting from a mis-keyed figure produces a confident, unrecoverable wrong answer. Working on paper and only then typing is slower per question and faster overall.

The minute nobody rehearses
The closing video is the part candidates prepare least and the part that is most obviously trainable. Sixty seconds fits a recommendation and two or three supporting reasons, nothing more. Record yourself doing it three times before the day. If you have never spoken to a camera under a clock, our guide to what recorded interviews actually score is the faster read than anything about the case itself.

BCG's other online screens, and why your friend sat a different test

Casey is not the only thing BCG has sent. MyConsultingCoach describes a separate, shorter screen it calls the CCA, around 30 to 35 minutes of reasoning and behavioural questions rather than a case, and says it is SHL-based. The same page describes office-specific variations, including a 30 minute cognitive test in Switzerland and a short numerical assessment in the UK, and states plainly that BCG does not run one identical online assessment everywhere.

Older formats circulate too. StrategyCase describes Casey's predecessor, the Potential Test, as closer to 45 minutes with about 23 questions and no video, and dates the Casey rollout to around 2020, starting in Asian offices. Game-based screening has also been associated with BCG at various points. Our BCG assessment page holds the SHL association at medium confidence for exactly this reason: the sources describe real things, but possibly real things from another office in another year. A 2023 guide can be accurate and still wrong for you.

Bain: the test with no clock, when there is a test at all

Bain's online stage is reported across many independent preparation sources to run on Sova, and the qualifications come from those same sources. PrepLounge states that some offices and roles require no digital assessment at all and that others use TestGorilla instead. CaseBasix dates a TestGorilla introduction to 2024. GraduatesFirst lists Sova, TestGorilla or HireVue depending on office and role, and describes a short video-recorded component in some processes.

If the invitation arrives from sovaonline.com, it is Sova.

Reported section
Situational judgement
Workplace scenarios. Sova asks for the single best and single worst response rather than a rating for every option, a different task from the format most people practise.
Reported section
Numerical reasoning
Business data in tables and charts. Multiple sources put the Bain deployment at around 15 questions. Sova tells candidates to have a calculator and paper ready.
Reported section
Logical reasoning
Abstract sequences: identify the rule running through a set of frames and continue it. No business content.
Reported section
Personality
A self-report questionnaire. Not scored right or wrong, but it is scored, and inconsistency across similar items is visible.
Disputed
Verbal reasoning
Some 2026 sources say Bain has dropped it. CaseBasix still covers it as part of the aptitude section. Prepare it lightly rather than ignoring it.
Disputed
Total length
PrepLounge reports Bain mentioning 30 to 40 minutes. CaseBasix says most candidates take 45 to 60. Both agree there is no hard limit.

The defining feature is the missing countdown. Nothing expires, and every source that discusses scoring says your pace is recorded and forms part of the result. PrepLounge describes a blended speed and accuracy score reported as a percentile against a norm group, with no published pass mark, and says consistent accuracy is what puts candidates in the top 20 to 30 per cent.

Sova publishes no Bain-specific spec, but its candidate preparation hub tells every candidate to “Complete all tasks in one sitting, if possible”, to have “a pen, paper and calculator for the ability questions”, and that “time is not recorded on the video recording sections, so this will not affect your results”. That last sentence is the closest thing to first-party confirmation that time is recorded on the sections that are not video.

The format mistake that costs most on the Bain SJT
SHL, Cubiks, Cappfinity and Kenexa situational judgement tests ask you to rate every response option. Sova asks you to pick one best and one worst and leave the rest alone. Practising the rating format does not teach the selection format, because one punishes hedging and the other rewards it. Our Sova guide covers the pacing problem in more depth.

What the three assessments actually reward

This is where grouping them together does real damage, because preparation that helps at one firm is close to wasted at another.

McKinsey
Solve
A bespoke simulation with no business content. It rewards reading an objective precisely, gathering only what it needs, and keeping decisions consistent with what the game shows you. Case frameworks do not help.
BCG
Casey
A case interview in a chat window. It rewards business judgement expressed as structure, arithmetic that is right first time, and one minute of clear spoken recommendation. Case practice transfers directly.
Bain
Sova
A conventional psychometric battery. It rewards speed and accuracy on ordinary reasoning items and internal consistency on self-report. Case practice does almost nothing for it.
Case preparation is the highest-return revision for BCG's online stage and close to irrelevant for Bain's. Timed numerical drilling is the highest-return revision for Bain's and close to irrelevant for McKinsey's.

The one thing all three share is that a calculator does not save you. Solve hands you a basic one and a great deal of data to filter, Casey hands you one and no route back to fix a bad figure, and Sova assumes one is on the desk while timing how long you take to use it. In each case the bottleneck is knowing which number you need.

If you are preparing for McKinsey too, the invitation email carries more information than anything else you have: our guide to what the Solve invitation length tells you about the test explains why 65 minutes and 85 minutes are two different assessments.

What to do when you do not know which test you are getting

  • Email the recruiter. One reply gives you the vendor, the length and usually the sections. Nothing else on this page is as reliable.
  • Check the sender domain when the invitation lands. sovaonline.com means Sova. A BCG-hosted chat interface means Casey. Ten seconds, no guesswork.
  • Drill numerical speed regardless. Every version of every MBB online stage rewards handling quantitative information quickly and correctly, and it transfers to the case interviews too. Timed numerical practice is the highest return hour available.
  • Start casing before you pass the screen. The gap between an online stage and a first round is often days, and casing takes weeks. At all three firms the interviews, not the screen, decide the outcome.

What is genuinely not known

Nobody outside these firms knows the cut score for any of these assessments, or whether a cut score is even the mechanism rather than a ranking. Nobody can tell you how heavily BCG weights the video against the written answers, or how Sova converts response time into a score. And nobody can tell you which test your office is running this cycle, because the firms change it without announcing it.

Which is the argument for the boring advice. Practise the Sova sections against your own benchmark so you know what your pace feels like, work a case out loud so the Casey video is not the first time you have delivered a recommendation in sixty seconds, and ask the recruiter. Our Bain and BCG pages track what we can evidence as it changes.

Ready to Start Practicing?

Apply these strategies with our comprehensive practice platform

Start Practising Free