PartnerRound- 01ListenNote the goal and every number
- 02ClarifyTwo to four sharp questions
- 03StructureTake time, present buckets
- 04AnalyseHypothesis, maths, exhibits
- 05RecommendAnswer first, with numbers
By the end you can
- Run a case from prompt to recommendation without waiting to be told what comes next
- Ask clarifying questions that change your structure, and skip the ones that do not
- Present a structure, state a hypothesis and drive the analysis in interviewer-led and candidate-led cases
- Do case maths out loud so the interviewer can follow it, check it and give you credit for it
On this page
01How a case runs
A case is a business problem solved out loud, usually in 25 to 35 minutes inside a 45 to 60-minute interview that also covers your experience. The interviewer holds the facts and releases them when you ask, or when the case calls for them. You are scored on how you think, not on reaching a secret answer.
Whatever the industry, every case has the same five phases. Knowing them means you always know where you are and what comes next, which is half of looking structured.
Two habits separate a pass from a strong pass. You keep linking each step back to the client's objective. And every number you produce ends with a sentence on what it means.
02Listen, take notes, clarify
The prompt is usually 30 to 60 seconds of speech packed with detail, and it is rarely repeated. Write it down. Keep one sheet for the prompt and facts, one for your structure and one for maths, and label every number with its unit.
- The client: who they are, what they sell, how they make money.
- The objective: the exact question, plus any target number or deadline.
- Every number, with its unit and period: '$30m profit, last year'.
- Anything the interviewer stresses or repeats. It is rarely there by accident.
Then play the question back in one sentence. It confirms you heard it, buys a few seconds and catches a misunderstanding before it costs you ten minutes.
Clarifying questions
Ask two to four questions, and only ones whose answers would change your structure or your maths. Good ones cover the objective, the business model, the scope and any ambiguous definition.
Weak
What trends are affecting coffee shops? Who are their main competitors? What's their market share? How many employees do they have?
Strong
Two quick questions. Is the goal to get back to $30m, and by when? And are all 200 stores company-owned, since that changes which costs the client controls?
Why: The weak version starts the analysis before there is a structure, and asks for data the interviewer would hand over later anyway. The strong version asks only what changes the plan, and says why.
03Take time, then present a structure
Always ask for time: 'Could I take a minute to structure my thoughts?' Then take 60 to 90 seconds in silence. Nobody marks you down for the pause, and a structure improvised while talking rarely holds together.
Build three or four buckets that are specific to this client, each with two or three questions or drivers underneath. Tailor the labels. 'Transactions per store and average ticket' tells the interviewer far more than 'Revenue'.
Turn the page so the interviewer can see it and present top-down: how many areas, their names, then the detail in each. Close by saying where you would start and why. That last sentence turns a list into a plan.
04Lead with a hypothesis and drive
A hypothesis is your best current guess at the answer, stated so it can be proved wrong. It tells the interviewer why you want each piece of data, and gives you a reason to pick one branch of your structure over another.
State it early, hold it loosely and update it out loud: 'That rules out costs as the main driver, so I'm now focusing on traffic.' Changing your hypothesis is not a weakness. Ignoring evidence is.
Driving means you choose the next step and say why, without waiting to be asked. Every time you finish a piece of analysis, close with three things: what you found, what it means, and what you want to look at next.
Weak
OK, so revenue went down by $22m. What would you like me to look at next?
Strong
So revenue fell $22m while the store count stayed at 200, which puts the problem inside the stores. Transactions per store fell 10% while the average ticket rose 5%. My hypothesis is that we're losing customers, to a competitor or to the price rise. I'd like to see whether the fall is spread across all stores or concentrated in some.
Why: The weak answer hands the case back to the interviewer. The strong one states the finding with numbers, says what it means, updates the hypothesis and picks the next step with a reason. That is what driving sounds like.
Interviewer-led and candidate-led cases
Firms run cases in two broad styles. Treat these as tendencies, not rules: formats vary by office and by interviewer, and plenty of cases mix the two.
Prepare for both. In an interviewer-led case you still drive inside each question: structure the answer, say what it means, link it back. In a candidate-led case you still listen for steers. If the interviewer asks 'What about costs?', that is a hint, not small talk.
05Maths, exhibits and brainstorms
Interviewers score how you reach a number as much as the number itself. Silent maths gives them nothing to credit and nothing to correct. Talk through four beats.
- 01Set upSay the approach before any arithmetic
- 02ComputeRound sensibly, narrate key steps
- 03CheckIs the answer the right size?
- 04So whatWhat it means for the client
Set-up is the beat candidates skip and the one that matters most. Once the interviewer has agreed your approach, an arithmetic slip is a small error. Without it, they cannot tell a slip from a misunderstanding.
- Keep units and zeros under control: 400k × $5 = $2.0m, not 2,000,000.
- Round only where it cannot change the answer, and say so: 'I'll call that roughly 30%.'
- If you lose your place, say 'Let me redo that step' and redo it. A calm recovery scores better than hiding the slip.
- Finish with the so-what. A number on its own is not an answer.
Exhibits
When a chart or table lands, say what it shows, ask for 20 seconds, then lead with the one insight that matters, with a number, and what it means for the client. The lesson Reading charts and exhibits covers the full routine.
Brainstorms
When asked for ideas, take a few seconds, name three or four buckets before any ideas, give two or three specific ideas in each, then say which you would do first. The lesson Structured brainstorming covers the splits that make this quick.
06What to say at each step
Most of the polish in a strong case comes from a dozen phrases used at the right moment. They signal structure, buy thinking time and keep the interviewer oriented.
The close is a recommendation, not a recap: the answer first, two or three reasons with numbers, the main risk and the next steps, in about a minute. The lesson The 60-second recommendation covers it in full.
07A full case, start to finish
Worked example
Coffee chain: profit down from $30m to $20m
1. Prompt and clarification
Prompt: 'Our client runs 200 coffee shops. Profit has fallen from $30m to $20m in two years. Why, and what should they do?'
The candidate plays it back and asks two questions. Answers: the goal is to get back to $30m within two years, and all 200 stores are company-owned and were open throughout.
2. Structure and hypothesis
The candidate presents revenue, costs and market, as in the structure above, and states a hypothesis: since the store count is flat, revenue per store has probably fallen. They ask for revenue and cost data for both years.
3. Maths, out loud
- Set up: 'Revenue is stores × transactions per store × average ticket. Variable cost is total transactions × $2. Profit is revenue minus variable and fixed costs.'
- Two years ago: 400k × $5.00 = $2.0m per store, × 200 stores = $400m. Last year: 360k × $5.25. 360 × 5 is 1,800 and 360 × 0.25 is 90, so $1.89m per store, × 200 = $378m. Revenue is down $22m.
- Variable costs: 80m transactions × $2 = $160m, falling to 72m × $2 = $144m. Fixed costs rose $4m. Check: $400m - $160m - $210m = $30m, and $378m - $144m - $214m = $20m. It ties out.
- Split the $10m fall. Volume: 8m fewer transactions × $3.00 contribution each = -$24m. Price: 72m transactions × $0.25 more each = +$18m. Fixed costs: -$4m. Total: -$10m.
- So what: 'This is a traffic problem. Losing 10% of transactions cost $24m, and the 5% price rise won back only $18m. Costs are a minor factor.'
4. Exhibit
The candidate asks whether the traffic fall is even across stores. The interviewer shows a split: 50 stores now have a new competitor within a few minutes' walk and lost 25% of their transactions. The other 150 lost 5%.
Reading it: 50 × 100k = 5m lost transactions near the competitor, and 150 × 20k = 3m elsewhere, which ties to the 8m total. Over 60% of the lost traffic, 5m of 8m transactions, comes from a quarter of the stores. The 5% loss elsewhere is plausibly the price rise.
5. Brainstorm and close
Asked for ideas, the candidate groups them by where the problem sits: winning back traffic in the 50 contested stores, through a loyalty offer, faster morning service or a sharper breakfast range, and protecting the other 150 by testing how sensitive their customers are to price.
Close: 'Profit fell $10m because traffic fell 10%, which cost $24m and was only partly offset by $18m from higher prices. Over 60% of the lost traffic is in the 50 stores facing a new competitor, so I'd focus there first with a plan to win customers back. The main risk is a price war. As a next step, I'd test retention offers in ten of those stores before rolling out.'
08How cases are scored
Firms do not publish their scoresheets, and they revise them. But the areas below appear in some form at McKinsey, BCG and Bain. Interviewers typically note evidence on each, then make an overall call on whether you move forward.
One error rarely sinks a candidate; a pattern does. A maths slip you catch yourself costs little. The same slip uncaught, with a recommendation built on top of it, usually costs a lot.
Check yourself
The interviewer has just given you the prompt. Which clarifying question is most worth asking?
Check yourself
In a candidate-led case, you have just finished a calculation. What should you do next?
Cheat sheet
The lines worth rereading the night before.
- 1Five phases: listen, clarify, structure, analyse, recommend. Always know which one you are in.
- 2Play the prompt back in one sentence, then ask two to four questions that would change your structure.
- 3Ask for a minute, structure in silence, present top-down and end with where you would start and why.
- 4State a hypothesis early and update it out loud when the data disagrees.
- 5Maths in four beats: set up, compute, check, so what.
- 6After every finding: what it is, what it means, what comes next.
- 7Interviewer-led: answer the question, then link it back. Candidate-led: choose the next step and say why.
- 8Close with the answer first, reasons with numbers, the main risk and next steps.
Now drill it
Reading builds the model; drills make it automatic. Start with the first topic, easiest drills first.
- Clarifying questions7 drills · Ask the few questions that change your structure: objective, scope, metric, timeframe and constraints.
- Profitability7 drills · Find out why profit moved by splitting it cleanly, segmenting early and asking the question that localises the problem.
- Executive synthesis7 drills · Give the CEO a 60-second recommendation: answer first, three reasons with numbers, risks and next steps.
Read the full guide
The same ground on the blog, with more worked examples and practice on the page.
Questions candidates ask
How long should I take to structure a case?
About 60 to 90 seconds in silence, then a minute or two to present. Ask first: 'Could I take a minute to structure my thoughts?' Much more than two minutes starts to look slow, so practise until 90 seconds is enough.
How many clarifying questions should I ask in a case interview?
Two to four. Ask about the objective, the business model and the scope, and only where the answer would change your structure. Data questions such as market share belong later, inside your analysis.
What is the difference between interviewer-led and candidate-led cases?
In an interviewer-led case, the usual McKinsey style, the interviewer takes you through a set sequence of questions. In a candidate-led case, common at BCG and Bain, you decide what to analyse next and say why. Formats vary by office and interviewer, so prepare for both.
Can I use a calculator in a case interview?
Not in live case interviews at McKinsey, BCG or Bain: the maths is done on paper. Online assessments can have different rules, so follow your invitation.
What happens if I make a maths mistake in a case?
Say so and fix it: 'Let me redo that step.' A slip you catch yourself usually costs little. An uncaught error that flows into your recommendation costs far more, which is why every answer gets a quick sense-check.