The Case Interview, End to End

What a case is simulating

A client arrives with a problem stated imprecisely. A team has limited time, incomplete data and a partner who will ask what the answer is. The case compresses that into twenty-five minutes, and everything in the format maps onto something real.

That mapping is the most useful thing to hold in your head. When you clarify the objective, you are doing what a manager does in a scoping call. When you structure, you are building the workplan. When you ask for data, you are deciding what analysis is worth a week of somebody time. When you recommend, you are in the room with the client. An interviewer is not marking whether you know a framework. They are marking whether they would put you on the project.

The sequence

The first three minutes, which decide most cases

Write the prompt down as it is said, including every number. Candidates who trust their memory reliably misremember a figure and build ten minutes of analysis on it.

Then clarify, and ask questions that would change what you do. The objective in measurable terms is almost always worth asking about: "increase profitability" could mean restoring a margin, a pound figure, or a return by a date, and those lead to different cases. The business model is worth confirming if it is at all unusual. The timeframe and any constraint on what the client is willing to do are worth knowing. Two or three questions is right. Six is stalling, and interviewers recognise it.

Then ask for a moment to structure. Sixty to ninety seconds of silence is expected, and using it is a sign of composure rather than a confession. Rushing into a framework you have not adapted is the single most common way strong candidates fail.

Presenting a structure that is not a list

A good structure has three or four branches, each relevant to this problem, distinct from each other and collectively covering the ground. It should be specific enough that it could not be read out for a different case.

Present it by walking through the branches and saying what you would look at in each, then close by naming where you want to begin and why. That last sentence is what separates a structure from a table of contents: it tells the interviewer you have a hypothesis about where the answer lives, which is how consultants actually work.

If the interviewer redirects you, take the redirection without visible disappointment. Being steered is normal and is sometimes a test of whether you can be steered.

Working through the analysis

The middle of a case is a loop: form a hypothesis, ask for the data that would test it, do the arithmetic, say what the number means, and decide where that takes you.

• Ask for specific data rather than asking what they have. "Do we have revenue split by segment for the last three years?" is a consultant question; "what information is available?" is not.
• Do the arithmetic out loud, so an error can be caught early rather than compounding.
• After every number, say what it means. A calculated figure with no interpretation is an unfinished thought, and this is the most common gap in otherwise strong candidates.
• Narrate your navigation. "That rules out volume, so I want to look at price next" keeps the interviewer with you and shows the logic that is being scored.
• When you hit a dead end, say so and move. A candidate who abandons a branch cleanly looks decisive; one who keeps digging looks stuck.

Reading an exhibit

Exhibits are given to see whether you can extract a message rather than describe a picture. Take fifteen to twenty seconds of silence to read it properly before speaking. Talking immediately produces a description, which scores nothing.

Then say, in order: what the exhibit shows, the one thing that stands out, and what it means for the case. Check the axes, the units and the time period before drawing a conclusion, because an exhibit with an unusual scale or a missing year is frequently the point of the exhibit.

Closing

Ask for twenty seconds if you need them, then give the recommendation in the answer-first shape: the recommendation, the two or three reasons, the main risk, and what you would do next.

Sixty to ninety seconds is the right length. Do not recap the case; the interviewer was there. Do not hedge into a recommendation that could be read either way, because the whole exercise is a test of whether you will commit under uncertainty. If the evidence genuinely does not support a clean answer, say what you would recommend and what would change your mind, which is a decision rather than a fudge.

How to practise so that it works

Reading about cases is worth very little after the first few hours. The skill is a performance skill and it develops through live repetition with feedback.

• Twenty to thirty live cases is the usual number to become comfortable. Beyond about fifty, returns fall sharply.
• Alternate giving and receiving. Giving a case teaches you what an interviewer can see, which is the fastest way to fix your own habits.
• Practise with strangers eventually. A familiar partner stops pushing back and you stop improving.
• Drill arithmetic and market sizing separately and daily, because they fail independently of structure and they are the easiest points to recover.
• Keep a log: the case, what went wrong, one thing to change. Reviewing that log before your next session is worth more than an extra case.