Now the crown skill. The same case can run in two modes. The behaviour that earns a pass differs sharply between them, even though the underlying rigour doesn't.
What is an interviewer-led case?
In an interviewer-led room, the classic McKinsey style, the interviewer owns the map. They hand you the problem in pieces. First comes "how would you structure this?", then "here's an exhibit, what does it tell you?". Then a specific worry, "the client is worried about X, what would you say?", and finally "give me your recommendation." Your job is to answer the question you were asked crisply and completely, then stop and let them steer to the next piece. The cardinal virtues are responsiveness and precision. You don't run ahead to build the whole case unprompted, you don't resist when the interviewer redirects you, and when you're handed a math exhibit you work it cleanly and state the "so what." The cardinal sins are wandering off the question, trying to grab the wheel, and failing to synthesise when you're finally asked. The whole case is a series of well-answered questions that you tie together only at the synthesis. Worked Example 5 is a full interviewer-led dialogue.
Advertisement
What is a candidate-led case?
In a candidate-led room, you own the map. That's the classic Bain/BCG style, and it's how most real client rooms run. The interviewer says "here is the situation, take it away," and then mostly reacts. You lay out your structure unprompted. You state a hypothesis about where the answer probably lies and why you'll test that first, and you ask for the specific data you need ("do we have the cost-to-serve split by channel?"). You drive from branch to branch, flag when a result updates your hypothesis, and offer the recommendation without waiting to be asked. The cardinal virtues are initiative and prioritisation, plus synthesis you drive yourself. Prioritising here means going hypothesis-first with minimum killing analysis, and no ocean-boiling. The cardinal sins are waiting passively to be told the next step, asking for data with no hypothesis behind the request ("can I see everything?"), and structuring exhaustively without prioritising. Worked Example 6 is a full candidate-led dialogue on a comparable problem, so you can feel the contrast for yourself.
What changes between the modes, and how do you read the room?
What changes between modes is who drives structure and pace, whether you volunteer the hypothesis and next step or wait to be asked, and how much of the map you build unprompted. What never changes is the rigour of the tree, the correctness of the math, the honesty of the stress test, and the answer-first close. So how do you read the room in the first thirty seconds? Watch what the interviewer does after giving the prompt. If they keep talking and ask a specific question, it's interviewer-led. If they go quiet and wait, it's candidate-led. When in doubt, open with a light bid for the wheel: "Let me lay out how I'd approach this, and you can redirect me." It works in both modes. A candidate-led interviewer lets you run, an interviewer-led one takes the wheel back, and either way their response has told you which room you're in. Reading the mode is the junior cousin of leading the client, which you'll meet later, where the stakes are a real steering committee. At most firms one room comes before either mode: an online screening test ahead of the first case round. At McKinsey that's Solve. Elsewhere it's online cases or aptitude tests, varying by office (as of October 2026, check on each firm's careers page).