In all the years I have interviewed, nobody has ever handed me a candidate's test score. I get a name, a resume, and a 45-minute slot. The test did its job before you reached me, which tells you what it is: a gate, not a grade.
A case interview is scored by a person against a rubric, out loud, in a room. An aptitude test is scored by software against a comparison group, alone, against a clock. The two share arithmetic and almost nothing else. Most candidates who fail a consulting screen can do the math. They fail because they prepared for the wrong sport, or treated a 90-second item like a 30-minute case.
What follows: what each screen measures, what the firms publish versus what candidates repeat, a seconds-per-item budget, and how I would spend ten hours if a test landed in your inbox.
Two scoring systems, one candidate
A case interview is criterion-referenced. You pass or fail against a fixed sheet: structure, math, judgment, communication. An unusually strong afternoon can produce four passes out of four; I have signed off on days like that.
An ability test is norm-referenced. SHL, the vendor behind many professional-services screens, says so in its own sample report: a percentile of 50 means the candidate "performed better than 50% of the people in the comparison group" (SHL Verify Ability Report). The employer picks the comparison group and the cut line. Your raw score is not the number anyone looks at. Your rank is.
Two consequences. No firm publishes a pass mark, so anyone quoting one is guessing. And the line can move with the size of the pile; that is my inference from how norm-referenced screening works, not something any firm states.
Why lean on a weaker instrument than your own interviewers? Because it is cheap and it scales. The most careful recent meta-analysis of hiring methods, Sackett and colleagues in the Journal of Applied Psychology, revised the validity of cognitive ability tests from .51 down to .31 and left structured interviews on top at .42. A test predicts reasonably well and costs almost nothing per candidate; an interview predicts better and costs two consultants an hour each. So the test shrinks the pile and the interview picks from it.
| Selection method | Schmidt and Hunter (1998) | Sackett et al. (2022) |
|---|---|---|
| Structured interview | .51 | .42 |
| Work sample | .54 | .33 |
| Cognitive ability test | .51 | .31 |
| Unstructured interview | .38 | .19 |
Operational validity correlations with job performance; higher predicts better. A case interview is a structured interview wrapped around a work sample; a screen is a cognitive ability test, sometimes with a personality or situational-judgment module bolted on.
What the firms publish, and what candidates repeat
The middle column is what you can verify on the firm's own site today; the right column is what circulates among candidates. Treat it as weather, not law.
| Screen | The firm states | Candidates report | Source |
|---|---|---|---|
| McKinsey Solve | Gamified; "no need to prepare"; length is in your invitation email; untimed tutorials; PC or Mac only; no AI tools or notes; weighed with the rest of your application | Two or three tasks (Redrock Study, Sea Wolf, and since early 2026 the Sustainable Futures Lab); 65 or 85 minutes | Solve page, Solve FAQ |
| BCG Online Case ("Casey") | Nothing quantitative; the careers site describes only the live case | Chatbot case, 8 to 10 sequential prompts, 25 to 35 minutes, no going back, short recorded recommendation | BCG case prep page |
| BCG Consulting Career Assessment (CCA) | Nothing quantitative | On SHL; two personality blocks plus about 8 numerical and logic items in about 12 minutes; introduced 2024, mostly US offices | none published |
| Bain digital assessment | Mandatory where required; Sova or TestGorilla, named in your email; 30 to 40 minutes; pen, paper, and calculator allowed; TestGorilla in one sitting, Sova may be split | Sova mostly in EMEA, TestGorilla mostly in the US; numerical, verbal, logical, and situational items | Bain assessment hub |
| SHL-style screens (Big Four and others) | Verify Interactive Numerical: 18 minutes, up to 10 items, adaptive, unsupervised, aimed at roles including professional services | Module choice shifts by country and intake year | SHL fact sheet |
Notice the pattern. McKinsey and Bain tell you the tool rules and that the format is in the email. BCG tells you nothing about either screen. Nobody publishes a question count, a pass mark, or a weighting; a guide that hands you those as facts has promoted folklore. The most useful line in all of it is McKinsey's: your assessment's length is in your invitation email. Read the email twice. It is the only specification you will get.
Same arithmetic, different product
One table handled two ways. Marlowe Foods is invented; the numbers are round on purpose.
| Region | Revenue 2024 ($m) | Revenue 2025 ($m) | Operating margin 2025 |
|---|---|---|---|
| North | 240 | 264 | 12% |
| South | 180 | 171 | 8% |
| West | 120 | 150 | 15% |
Test item: "What was Marlowe's total operating profit in 2025, to the nearest $1m?"
The test wants 68. The route: 264 at 12% is 26.4 plus 5.28, call it 31.7. 171 at 8% is 13.7. 150 at 15% is 22.5. Sum 67.9, so 68. Type it, move on. Elapsed: perhaps 50 seconds with a calculator, 75 without. Nothing else you know about Marlowe earns a point.
Case version: I slide the same table across and say, "What do you see?" Now 68 is the least interesting number on the page. I am listening for this: North still carries almost half the profit; South shrank 5 percent on the thinnest margin; West grew 25 percent at the best margin. So the questions are whether West's growth repeats and whether South is worth fixing or exiting. That takes 40 seconds to say and earns most of the marks on my sheet. The arithmetic earns almost none.
Same data, same math, opposite product. On the test the "so what" is wasted seconds; in the interview the number without it is a wasted question. Drift between the two and you leak points in both.
The time budget nobody writes down
Convert every format into seconds per item before you open it. SHL's fact sheet gives Verify Interactive Numerical up to 10 items in 18 minutes: 108 seconds each. Candidates report the CCA's timed block at about 8 items in 12 minutes, so 90 seconds, and Casey at 8 to 10 prompts in 25 to 35 minutes, about three minutes each. Where the firm publishes only a total, as Bain does, divide by the item count on the first screen and hold yourself to it.
At 90 to 108 seconds per item, three rules follow.
Never spend more than double budget on one item. Three minutes on a stubborn chart in an 18-minute test is one-sixth of your time on one-tenth of the marks. Take your best estimate and go.
Read the qualifier before the data. "Net", "per month", "excluding North". Most wrong answers I see in practice runs are correct calculations of the wrong quantity.
Round early. On a multiple-choice or tolerance-scored item, 264 at 12% is "about 32" and that places it. Save exact arithmetic for items that want a typed number.
Where case habits help, and where they cost you
I have watched candidates with excellent case technique go quiet after a screen, not for lack of ability. The habits that impress me in the room punish them on a test.
| Case habit | On a test |
|---|---|
| Asking a clarifying question before you start | No one is there, and the clock runs |
| Laying out a framework before touching the numbers | Zero points; costs 20 to 40 seconds per item |
| Talking through your reasoning | Silence is fine; nobody is listening |
| Checking units, magnitudes, and chart titles first | Transfers fully; the best defense against the wrong-quantity error |
| Staying level when a number looks wrong | Transfers fully |
The exception is BCG's chatbot case, a case in test clothing. It wants structure and a short recorded recommendation, but on a fixed clock with no hints and no second attempt: a candidate-led case with nobody to lead. Practice it as a written case at speed, not as a psychometric test or a conversation.
Ned's rule. The invitation names the sport. If it says test, assessment, or game, nothing you do with a case book this week moves your score. If it says interview, nothing you do on a practice test moves mine. Prepare for whichever is scored next, and only that.
Read the email like a contract
Every firm that says anything says the same thing: the details are in the email. Extract five lines before you practice anything.
- Total length. Tells you which modules you face; Solve's length differs by invite.
- Platform name (Sova, TestGorilla, SHL, HireQuotient). Practice on that vendor's own sample items first, as Bain itself tells Sova candidates.
- Calculator, pen, and paper rules. Bain allows all three on the assessment and none in the interviews that follow. Practice with exactly what is allowed.
- One sitting or several. Plan the sittings rather than discovering the rule at minute 20.
- Deadline. Days, not weeks. Bain does not consider applications without a completed assessment.
Read the integrity clause. McKinsey and Bain both prohibit AI tools, screenshots, and recording; McKinsey also bans pre-prepared notes, and its FAQ says Solve varies its parameters so no two candidates see the same data. Do not test that claim.
Ten hours, ten days
The email arrives on day 0, the test is due on day 10, and you can find ten hours. Here is how I would spend them; adjust the ratio if you must, but keep the order.
| Day | Hours | Do |
|---|---|---|
| 1 | 1 | Read the invitation twice. Extract the five lines. Open the vendor's own sample items once, untimed |
| 2 | 1.5 | One full timed run of the exact format. Log three things only: items unfinished, items where you read the wrong quantity, seconds lost to arithmetic |
| 3 to 7 | 1 per day | Drill the biggest leak from day 2. Arithmetic: speed math. Reading: qualifier-first practice on charts. Unfinished items: a visible clock and the double-budget rule |
| 8 | 1.5 | Second full timed run, same hour of day as your slot, same tools the email allows |
| 9 | 1 | Review day 8 errors only. No new topics |
| 10 | 0 | Sleep, then sit the test |
Note what is missing: live cases. If the same email also books an interview the following week, add case practice from day 6 on separate days, never in the same evening as a timed run. Mixing them makes both sloppy; I have watched it happen to strong candidates.
Practice this today
One assignment. Find the invitation, or the firm's assessment page if you have not applied yet, and write the five lines on a card. Then do a single timed rep of the matching format and record the three leaks. Nothing else on day one.
Where CoachNed helps: the CCA simulator runs timed numerical blocks at about 90 seconds an item; Bain Sova practice and the Red Rock, Sea Wolf, and Sustainable Futures Lab rooms cover the other formats; the UrbanBrew chatbot case is Casey-style and forward-only. If your leak is arithmetic, quick math is free for as long as you want it. Everything else is open for seven days with no card; after that it is $120 for a recruiting season or $49 a month. Save the live voice case for the week an interview is actually scheduled.
CoachNed is independent and not affiliated with McKinsey, BCG, Bain, SHL, Sova, TestGorilla, or any other firm named here; its simulators are original practice material, not official items.
Frequently asked questions
Is a consulting aptitude test harder than a case interview?
Different, not harder. The test is arithmetic and careful reading under a clock with no help; the case is judgment and communication with a person who can steer you. Whichever one you did not practice will feel harder.
Can a strong case interview make up for a failed online test?
Usually not: the test comes first, and the interview does not happen if you miss the cut. Bain calls the assessment mandatory where required; McKinsey says Solve is weighed with the rest of your application, the softest language any firm uses, and still not a promise.
Do I need business knowledge for McKinsey Solve or Bain's assessment?
No. McKinsey's FAQ says Solve needs no business knowledge or gaming experience; Bain says TestGorilla needs no preparation and points Sova candidates to Sova's sample questions. Format familiarity moves your speed; content knowledge does not.
How long do I have to complete the test after the invitation?
Days, not weeks, and it varies by firm and office. Bain puts the deadline in the email; candidates report windows as short as 24 hours for some BCG screens. Treat anything under 72 hours as "start today".
Sources
- Solve, McKinsey's assessment game — "no need to prepare", result weighed with the rest of the application, ban on AI tools, notes, and screenshots. Checked 2026-09-24.
- Solve FAQ (McKinsey, PDF) — length in the invitation email, untimed tutorials, no business knowledge needed, parameters vary per candidate. Checked 2026-09-24.
- Digital assessment hub (Bain & Company) — Sova or TestGorilla, 30 to 40 minutes, calculator and sitting rules, deadline, AI-tool ban. Checked 2026-09-24.
- Case interview preparation (BCG Careers) — BCG's public description of the case interview; no Online Case or CCA specification is published. Checked 2026-09-24.
- SHL Verify Interactive Numerical Reasoning fact sheet (PDF) — 18 minutes, up to 10 questions, adaptive, unsupervised, professional-services roles. Checked 2026-09-24.
- SHL Verify Ability Report, sample (PDF) — percentile scoring against a comparison group. Checked 2026-09-24.
- Sackett, Zhang, Berry, and Lievens (2022), Journal of Applied Psychology — structured interviews .42, cognitive ability tests .31, unstructured interviews .19. Checked 2026-09-24.