ChannelPulse
What a case interview actually tests
A case is seven phases with a rough clock on each, and almost every avoidable failure is a timing failure — four minutes of clarifying questions, a framework nobody could follow, or a recommendation that arrives as the interviewer is closing their notebook. This lesson is the structure, the minutes, and what each phase is collecting.
What you'll be able to do
- Name the seven phases of a case and the rough time each one gets
- Play back a prompt in one sentence that proves you heard the client, the metric, and the decision
- Ask clarifying questions that scope the problem instead of mining for data
- Tell an interviewer-led round from a candidate-led one within the first two minutes, and adapt
A case interview is not a business quiz. It is thirty to forty minutes in which somebody watches you take an under-specified problem apart, and the thing being recorded is not whether you arrived somewhere — most cases have no single correct destination — but whether a stranger could have followed you there.
That distinction has a practical consequence. Almost nothing in a case rewards knowing more about the industry. Almost everything rewards a visible process: you restated the problem, you scoped it, you built a structure, you said which part you were entering and why, you did arithmetic that survived a sanity check, and you finished with a decision rather than a summary. Candidates who lose cases usually lose them on process, and specifically on time spent in the wrong phase.
Watch this before reading the rest of the lesson. It runs the same six-to-seven phase arc this lesson describes, on a live prompt, at roughly real speed — which is the part written explanations cannot convey. The chapter at 3:35 is the one to sit with: the claim that the case exists to test structure rather than business knowledge is the premise everything below depends on.
Jump to the part you need
The seven phases, with a clock on each
Here is the arc, with the timings that practitioners who run these rounds tend to quote for a 30-minute case. The numbers matter more than the labels.
| Phase | Minutes | What it is collecting |
|---|---|---|
| 1. Take in the prompt and play it back | 2 | Listening. Did you hear the client, the metric and the decision? |
| 2. Clarifying questions | 1–2 | Scoping judgment. Which ambiguity actually blocks you? |
| 3. Build your structure, in silence | 2–3 | Structuring. This is the phase you are allowed to go quiet in. |
| 4. Present the structure | 1–1.5 | Communication. Can someone else use your framework? |
| 5. Quantitative work | 10–12 | Arithmetic under observation, and interpretation of it. |
| 6. Qualitative work | 8–10 | Business judgment where no number exists. |
| 7. Recommendation | 1–2 | Decisiveness. A call, its reasons, its risks, and next steps. |
For a 45-minute case, the shape is the same and the middle stretches: 2–3 minutes on the prompt, 2 on questions, 3–4 building, 1–2 presenting, 15–18 quantitative, 12–15 qualitative, and still only about 2 minutes for the recommendation.
Read the two ends of that table together. The recommendation gets the least time and carries the most weight, because it is the only phase where the interviewer hears a decision. And the structure gets two or three minutes of silence, which is the only silence you are granted — spend it anywhere else and it reads as being stuck.
Phase 1: play the prompt back
The prompt is read aloud once. You will not get it again in full. Write it down, then say it back in one sentence that contains three things: who the client is, what metric is moving, and what decision they are trying to make.
To make sure I understand correctly: our client is a national grocery chain with about $5 billion in annual revenue, their operating margin has fallen over the last two years, and they want to know whether to keep investing in their private-label line. Is that the question?
That sentence takes twelve seconds and does four things. It proves you were listening. It surfaces a misunderstanding now rather than at minute twenty. It gets you a nod, which is the cheapest confidence you will acquire all round. And it forces you to name the decision, which is the thing candidates most often fail to extract from a prompt — they hear "margins are down" and start diagnosing, having never noticed they were asked about a specific investment.
If your playback and the interviewer's prompt differ, they will correct you. That correction is worth more than anything you would have said in its place.
The note-taking layout
A small piece of physical craft that consistently separates people who look organised from people who are organised. Turn the page landscape. Draw one vertical line about two-thirds across. The left two-thirds holds your structure — the framework you will present and then work through. The right third holds everything the interviewer says: numbers they give you, constraints, side remarks.
The reason to split the page is that the two kinds of writing have different lifetimes. Your structure gets referred to for the whole case. Their data arrives once, in the middle of a sentence, and you will need it again ten minutes later. Mixed together on one page, the structure gets buried under scratch arithmetic and you end up rebuilding it from memory while talking.
Phase 2: questions that scope, not questions that mine
There are two kinds of question you can ask, and only one of them is scored well.
Scoping questions narrow the problem so a structure becomes possible. What counts as success — margin recovery, absolute profit, share? Which geography, which product line, what time horizon? Does "private label" include the frozen range? These are questions whose answers change the shape of your framework.
Data questions ask for numbers. What is their current margin? How many stores? What is competitor share? These are not clarifying questions; they are the quantitative phase arriving four minutes early. Ask them later, one at a time, at the point in your structure where you need them — that is where they read as analysis instead of as fishing.
The test is simple: if I get this answer, does my framework change? If yes, ask it now. If it just fills a box you have not drawn yet, wait.
Interviewer-led and candidate-led, and how to tell within two minutes
Cases come in two formats and the difference changes how you behave, not what you know.
| Interviewer-led | Candidate-led | |
|---|---|---|
| Typically | McKinsey | BCG, Bain, and most others (Bain often a hybrid) |
| Who chooses the next step | The interviewer asks specific questions in sequence | You decide what to explore and in what order |
| Your structure | Present it briefly, then follow their prompts | Present it fully, then lead yourself through it |
| Hypothesis | Less emphasis on stating one upfront | State one early and update it out loud |
| Length | Around 25–30 minutes | Around 30–35 minutes |
The tell arrives right after you present your framework. In an interviewer-led case the interviewer says something like "let's start with the cost side — here's an exhibit." In a candidate-led case they say "where would you like to begin?" That single sentence tells you which format you are in.
The adaptation is small but real. In a candidate-led case, you are responsible for transitions: "I've established that volume is flat, so I'm going to leave revenue and move to costs." Without those sentences the case drifts and the interviewer has to steer, which is a mark against you. In an interviewer-led case, volunteering the same transitions cuts across the interviewer's own sequence — answer what was asked, finish cleanly, and let them move you.
The four things being scored
Whatever the format, the scorecard behind a case round tends to reduce to four items, and they are not weighted the way candidates assume.
Structure. Could a colleague pick up your framework and use it? This is scored in phase 3 and 4, and it is the item that most often decides the round.
Quantitative comfort. Not speed. Accuracy, a stated approach before you start computing, and a sanity check afterwards. The next-but-one lesson is entirely about this.
Business judgment. Whether your qualitative reasoning is grounded — a real reason a customer would switch, a real reason a cost is fixed — rather than a plausible-sounding list.
Communication. Signposting, pace, and whether the interviewer ever has to ask "sorry, where are we?" This is the item candidates think is soft and interviewers treat as diagnostic, because consulting work is largely the act of making a stranger understand something quickly.
Notice what is absent: knowing the industry, and getting the "right" answer. Both help at the margin. Neither is on the list.
What to carry into the next lesson
Phase 3 — the two or three silent minutes where you build a structure — is the phase with the highest score-per-minute in the entire round, and it is the one most people improvise. The next lesson is about doing it deliberately: issue trees, what MECE actually means, and why the most reliable way to be exhaustive is to write an equation rather than a list of topics.