Between the screen and the offer, a strategy firm will typically watch you work through four to six live cases, and the debrief does not average them. One flat case is usually enough to end the process; one brilliant case has never, in my experience, been enough to start an offer. So the entire job of case interview prep is to raise your worst performance, and most candidates spend their time polishing their best.
That changes what you practice and with whom. Polishing the ceiling means another profitability case with a friend who feeds you hints. Raising the floor means the case type you avoid, the industry you know nothing about, the interviewer who pushes back, and math done out loud when you would rather do it in silence.
What follows: how many times the process samples you, the five moments each sample is scored on, a case scored two ways, an eight-week plan built as variance reduction rather than a reading list, and one assignment for today.
How many times the process samples you
Firms publish less than prep sites pretend, so separate what they state from what candidates report.
McKinsey's interviewing page says that for most client-facing roles you "participate in a personal experience interview, followed by a problem-solving interview," and asks for two examples for each of four PEI themes (McKinsey, Interviewing). It does not publish round or interviewer counts; two rounds of two to three interviews is the commonly reported pattern.
BCG publishes by office, which is how you learn the process is not uniform. BCG Mexico states two rounds of "two/three case interviews that will last 45 minutes each," about a week apart, after a chatbot online case and a Consulting Career Assessment of about 30 minutes each (BCG Mexico, application process). BCG Switzerland describes a cognitive test, one virtual first-round case, and three in-person final-round cases, and adds that "a decision may already be made" after the first round (BCG Switzerland, application and interviews). Bain London states two rounds, both case interviews, after an online assessment of numerical reasoning, logical reasoning, and personality (Bain London recruiting).
Add it up, allowing for office variation: one screen, then four to six live cases with behavioral questions around them, two to four hours of you talking.
McKinsey told Business Insider it receives more than one million applications a year and planned to hire about 6,000 people in 2024, roughly 0.6%; BCG's head of talent put its figure at more than one million applicants and about 1% hired (Business Insider, November 2024). Each stage of that funnel is designed to cut, and the cut happens at your weakest sample.
As experience rather than any firm's policy: after a round, the debrief starts with the interview that went worst. A false positive is a consultant who cannot hold a room in front of a client; a false negative is invisible. A strong "yes" from me rarely rescues a clear "no" from a colleague. You are hired on your floor.
The arithmetic of a floor
Take a deliberately crude model: each case is a pass-or-fail sample with the same pass probability, you must pass all of them, and your odds are that probability multiplied by itself once per case.
| Pass rate per case | 4 cases | 5 cases | 6 cases |
|---|---|---|---|
| 70% | 24% | 17% | 12% |
| 80% | 41% | 33% | 26% |
| 90% | 66% | 59% | 53% |
| 95% | 81% | 77% | 74% |
Real interviews are not independent coin flips: a strong first case buys goodwill, partners weigh more, and a borderline candidate sometimes gets an extra interview. But the direction holds, with a sharper corollary once your pass rate differs by case type.
Suppose you are at 95% on five case types and 70% on the one you avoid. Over six cases that is 0.95 to the fifth power times 0.70, about 54%. Raise the weak type by ten points, to 80%, and you are at 62%. Take a strong type up by the only five points it has left, to 100%, and you reach 57%. The weak type is where the points are, and it is the one you skip because practicing it is unpleasant.
The five moments a case is scored on
The sheets I have filled in had five to seven boxes, and the boxes get filled at specific moments, not continuously. In a 25-minute case:
| Moment | Minutes | What the sheet rewards | Where the points leak |
|---|---|---|---|
| Opening | 0 to 4 | A restatement the client would sign, clarifying questions that change the answer, and a structure built for this question, inside two minutes | Five boxes recited from a book; three minutes of silence; clarifying questions that are really stalling |
| Direction | 4 to 7 | "I would start with X, because Y" | "I would look at all of these" |
| Math | 7 to 17 | The setup said aloud before any digit, a sane order of magnitude, a sanity check, an interpretation | Silent arithmetic; a wrong answer defended; a right answer with no so-what |
| Exhibit | Whenever it appears | Title, units, the one line that matters, and what it means for the client | Reading the chart back to me |
| Close | 22 to 25 | Answer first, two or three supports, one risk, one next step | A summary of what we did instead of an answer |
The opening is scored before you know it. By minute four I have a working hypothesis about you, and the rest of the case mostly confirms it. Overturning it is possible but expensive; you have to be visibly better than my expectation.
And candidates over-invest in the math box while under-investing in the judgment box. I have marked candidates down for getting the right number and saying nothing about it far more often than for a slip they caught. The sheet has no box for "computed to the dollar." It has a box for "knew what the number meant."
The behavioral half runs on a separate sheet with the same logic, scored on your weakest story, which is why CoachNed's behavioral simulator runs four follow-ups on every story.
One case, two candidates, one sheet
Prompt. Brightwater runs 150 self-service laundromats in the US Southeast, with same-store revenue flat for three years. The CEO asks whether to add wash-and-fold with pickup and delivery.
Data given on request. Each store draws on about 15,000 households. A survey says 2% would use the service monthly, at $35 an order (25 pounds at $1.40). Variable cost per order is $13. Fixed cost is about $60,000 per store per year for staff, a van share, and marketing.
The arithmetic.
- Households: 150 × 15,000 = 2,250,000
- Users at 2%: 45,000, or 300 per store
- Orders a year: 45,000 × 12 = 540,000
- Revenue: 540,000 × $35 = $18.9 million
- Contribution: 540,000 × ($35 − $13) = $11.9 million
- Fixed cost: 150 × $60,000 = $9.0 million
- Profit: about $2.9 million, or roughly $19,000 per store
The number that matters. Breakeven adoption is fixed cost divided by contribution per user-year: $9.0 million over $264 ($22 × 12) is about 34,000 users, which is 1.5% of households, or about 227 per store. The survey said 2%. The whole decision lives in that half-point gap, and stated intent in a survey runs above actual behavior.
| Moment | Candidate A | Candidate B |
|---|---|---|
| Setup aloud | Ninety seconds of silence, then a number | "Households, times adoption, times orders a year, times price" before touching a digit |
| Arithmetic | Correct to the dollar: $2,880,000 | Rounded to $11.9 million and $2.9 million, said aloud, correct |
| So-what | "It is profitable, so I would launch" | "About $19,000 a store, which is thin, and it depends on whether adoption is 1.5% or 2%. Pilot ten stores for a quarter and stop below 230 active households per store" |
| My marks | Math box full, judgment box empty | Both boxes full |
Candidate A did nothing wrong and will not be hired on that case. Candidate B rounded, thought out loud, and found the breakeven, the only number the CEO would remember. If Candidate A does that on one case out of five, it is the one the debrief talks about.
Ned's rule. Nobody gets hired for their best case and everybody gets rejected for their worst one. Prepare for the case you would least like to be given, delivered by the interviewer you would least like to be given it by.
A plan built to reduce variance
Most prep calendars are reading lists with dates. This one is organized around the floor, and every phase has a test.
| Phase | Weeks before the first interview | What you do | How you know it worked |
|---|---|---|---|
| Find the floor | 8 to 7 | Six cases in six types (profitability, market entry, pricing, M&A, operations, market sizing), out loud, scored on the same five moments | You can name your weakest type and moment, with a number on each |
| Raise it | 6 to 4 | A 15-minute daily drill on the weak moment; two cases a week in the weak type; one behavioral session a week | The weak type scores within a point of your average on three consecutive cases |
| Stress it | 3 to 2 | Cases in industries you know nothing about, timed, with a partner told to push back on every number | No case falls below your average for two weeks |
| Hold it | 1 | One case a day in mixed types, no new material, sleep | You are maintaining a level, not building one |
The fixed sheet is the point of phase one: without one you remember the cases that went well, which is the bias this guide exists to remove.
Phase two is where prep stops being fun; the weak type is weak because you avoid it. If market sizing is the leak, the market-sizing drill is your morning; if you go quiet at "so what," it is the synthesis drill.
Phase three is the one candidates skip and the one that raises the floor most. An interviewer who says "I don't believe that number" in minute twelve is a normal Tuesday, and you want to have heard it thirty times before it costs you anything.
The screen runs on its own, earlier clock. BCG's US page says the CCA invitation arrives 30 to 90 minutes after you apply, with 48 hours to complete it (BCG United States Associate opportunities). So one or two sittings of a simulator for Solve-style games, the Sova assessment, the CCA, or a Casey-style chatbot case belong in the week you apply.
Practice this today
Sit two cases today, out loud, in the type you would least like to be given tomorrow. Score each on the five moments with 0, 1, or 2. The lower total is your floor as of tonight; the lowest moment is your drill for the week.
If you have never done a case out loud, the five-minute first rep is three typed turns with instant scores and no account. If you have, the live voice case with Ned runs about eighteen minutes with a seven-score debrief, and the lowest of the seven is your floor. After that, the case library has 64 original cases you can filter by type, which is how you run phase one in a weekend.
Everything on CoachNed is open for seven days with no card; after that it is $120 for a recruiting season or $49 a month (pricing). CoachNed is independent and not affiliated with McKinsey, BCG, Bain, or any other firm named in this guide.
Frequently asked questions
How many case interviews do you need to pass to get an offer?
Typically four to six across two rounds, plus an online screen. BCG Mexico states two or three 45-minute cases per round and Bain London two rounds of case interviews; McKinsey publishes no count, and two rounds of two to three interviews is the commonly reported pattern.
How long should you prepare for a case interview?
Eight weeks of the plan above is comfortable; with two, run phases one and three only, two days to find the floor and ten to stress it. Spread beats volume: twenty cases across six types beat forty in one.
Do you need to memorize frameworks for case interviews?
No, and reciting one is a reliable way to lose the opening. McKinsey describes its case as a discussion of a "typical McKinsey client scenario," and Bain says it will "never ask you to memorize and recite material" (Bain, what to expect). The sheet rewards a structure built for the question asked.
Can you fail one case interview and still get an offer?
Sometimes, if it was a slip the other interviewers can explain away and the rest of the round is strong. In my experience a clear "no" from one interviewer is rarely overturned, which is why this guide is about your floor, not your average.
Sources
- Interviewing at McKinsey — PEI followed by a problem-solving interview for most client-facing roles; two examples per PEI theme. Checked 2026-09-24.
- BCG Mexico: application process — two or three 45-minute cases per round, rounds a week apart; Casey and CCA about 30 minutes each. Checked 2026-09-24.
- BCG Switzerland: application and interviews — cognitive test, one virtual first-round case, three in-person final cases, decision possible after round one. Checked 2026-09-24.
- BCG United States Associate opportunities — CCA invitation 30 to 90 minutes after applying; 30 to 35 minutes long; 48 hours to complete. Checked 2026-09-24.
- Bain London Associate Consultant recruiting — two rounds of case interviews after an online assessment of numerical reasoning, logical reasoning, and personality. Checked 2026-09-24.
- Bain: what to expect — Bain will "never ask you to memorize and recite material." Checked 2026-09-24.
- Business Insider: MBB explained — McKinsey's reported one-million-plus applications and about 6,000 planned 2024 hires; BCG's reported one-million-plus applicants and about 1% hired. Checked 2026-09-24.