Case interviews are a crucial part of getting an offer at a consulting firm. They put you in the consultant's seat: the interviewer gives you a real business problem, such as why a company's profits are falling or whether it should enter a new market, and you work through it out loud. You structure the problem, ask for the data you need, run the numbers, and land on a recommendation.
It is not a quiz about business. It is a 30 minute working session, and the interviewer is judging how you think and approach a problem. Everything about how we built The Case Room's interviewer follows from that.
I know this format from both sides of the table. I went through it as a Kellogg MBA student and came out with offers from McKinsey, BCG, Bain, and PwC Strategy&. Then I spent five years at McKinsey, where I was an interviewer, and left as an Engagement Manager.
What happens in the room
Consulting firms use cases because the job is a case. A client hands you a messy problem, and you have to break it down, find the numbers that matter and tell a senior person what to do. The interview is a small, timed version of that.
The interviewer gives you a short business problem. Then you lead. The case usually moves through seven parts, and each one tests something specific:
- The prompt. The interviewer sets up the case: who the client is, what is happening, and the question they need answered. Do you listen for the one decision the client has to make, and can you play it back in a sentence before you start?
- Clarifying questions. Do you pin down what the client is actually trying to decide before you start solving?
- Structure. Can you break the decision into the few areas that will settle it, and name the most important details you would check under each?
- Exhibits. Handed a chart or table, can you say what it shows, pick out the insights that bear on the client's problem, and name the most important one and how it changes where you take the case next?
- Math. Do you know what to calculate and why, how to set up the calculation, what data you need and what assumptions to make, and can you get it roughly right and say what the result tells you?
- Brainstorm. Can you generate ideas that are specific to the client's problem but also creative, and group them so the interviewer can follow?
- Recommendation. Can you give a clear answer first, then your reasons and the risks?
A small example
Say the prompt is: "A client wants to leave a $120,000 job to open a coffee shop. Should she do it?"
A weak first answer sounds like this: "I'd add mobile ordering, run a loyalty discount and host events on weekends." Those are ideas for running a coffee shop. They don't help anyone decide whether to open one.
A strong first answer sounds like this: "She should open it if it can pay her more than her current salary at a risk she can live with. So I'd look at three things. First, the market: how many potential customers are nearby, how often they'd come and what competitors charge. Second, the economics of one shop: cups a day times price, minus rent, staff and ingredients, and how what's left compares with $120,000. Third, the risks: the upfront cost, how long until it breaks even and what happens if she's wrong. I'd start with the economics, because that's where the salary question gets decided."
The strong answer doesn't know the answer yet. It tells you how you would *know* the answer. That is what the interviewer is listening for.
Why you do the math by hand, out loud
People who have never done a case sometimes ask why we don't just do the math for the candidate. Any calculator could do it.
That's true, and it's the point. The arithmetic is the easy part. What the math tests is everything around it: choosing the right equation before touching a number, noticing when a result is off by a factor of ten, and thinking clearly while you talk under pressure. A candidate who says "revenue is cups a day times price times days open, so roughly 300 times $5 times 360" has shown me more than a correct final figure ever could.
It is also the job. In client meetings, someone asks "so what is that worth?" and there is no spreadsheet on the table. You set it up out loud, round sensibly and give a number you can defend.
Real interviewers don't hand you the answer, and neither do we. (For a step by step method, see The Five Parts of a Strong Case Interview Math Answer.)
How we made our interviewer behave like a real one
Our interviewer is an AI, and it still makes mistakes; we review real sessions and fix what we find. Here is how it is designed to behave, and why.
It responds to what you actually said. Each question has a list of what still needs to be covered, but the next thing you hear comes from your answer. If your structure is a list of labels, it asks what specifically you'd look at under one of them. A real interviewer follows your thinking, not a script.
It won't tell you your answer is good. Praise like "exactly right" is filtered out before it is spoken. If a number looks off, it may ask you to double-check it, but it never gives you the right one. In a real interview, the person across the table keeps a neutral face; confidence has to come from your own sanity check.
You can ask for a moment. Say "can I take a minute?" and you'll hear "Sure, take your time," then quiet. It checks "Are you still with me?" only after about two minutes of silence. Asking for time to think is normal; in fact, interviewers expect it.
Hints come when you ask, and they point rather than answer. Press the hint button and you get one step from the case's own hint list: what to think about, then where to look, then the approach. None of them states the answer. That is how a good interviewer helps a stuck candidate without doing the work for them.
If you're stuck on a number, it may give you one and move on. After a long stall on a calculation, it may offer the number so the case can keep going, or say "For the rest of the case, let's assume" a figure a later question needs. Real interviewers do this too. One missed calculation shouldn't end your case, and your feedback notes it and scores the reasoning you did.
Each question has a time limit. Most questions get up to about 10 minutes, and a one-minute reminder appears on screen rather than being spoken, so it doesn't break your concentration. At the limit you'll hear "Do you have any other thoughts on this question before we move on?", and then it moves on. Real interviews cover several questions in under an hour, and interviewers keep you moving.
It pushes back. If your structure is thin, it asks what else you'd look at or what specifically you'd check; and it will ask you to walk through how you got a number. A real interviewer tests whether you can defend your thinking, not just produce it.
The case ends with your recommendation. If time is running short, it moves you straight there. Every real case ends with "so what should the client do?"
Feedback comes question by question, scored against the standard I used when I was interviewing candidates at McKinsey.
How to get the most out of practice
- Talk out loud the whole time. Silence tells an interviewer nothing.
- Structure first. Before any idea or number, say how you would decide.
- Say the formula before you compute. "Revenue is volume times price" first, then the numbers.
- Give the answer first in the recommendation. "She should open it, because..." and then your reasons and risks.
Your first full case is free: try one at The Case Room.