McKinsey case interview: how to solve the Problem Solving Interview, with a full example
How the McKinsey case works, the question types it uses (structure, exhibit, math, brainstorming and synthesis) and an example case solved step by step, with the math done by hand.
Practice case math
Every McKinsey interview has two parts: the Personal Experience Interview (PEI) and a case, which McKinsey calls the Problem Solving Interview. This guide covers the case: how it works, the question types it uses and what a good answer looks like, with an example case solved from start to finish.
This guide combines what McKinsey publishes on its careers site with what prep guides and candidates describe. The example case is made up. Your recruiter gives you the instructions for your own process: if anything differs, go with what they tell you.
What it is and how it works
You get a client’s problem and solve it out loud with your interviewer. What sets McKinsey apart is that the case is interviewer-led: the interviewer decides what gets analyzed and in what order, and asks you questions one at a time.
- Length: according to the guides, 25 to 40 minutes of an interview that lasts about an hour. The rest is the PEI.
- Questions: usually three to six. Each one is assessed on its own, so one weak answer doesn’t sink the whole case, but one brilliant answer doesn’t make up for it either.
- No calculator: you do the math by hand, on paper.
- Rounds: the guides describe a first round of two interviews and a final round of two or three, with more senior interviewers. In the final round the case can be more open and conversational.
At BCG or Bain, by contrast, the candidate usually leads the case: you decide what to analyze next. At McKinsey you don’t have to guess the path, but you do have to answer each question well and tie it back to the client’s goal.
What they assess
The guides agree on five things:
- Structure: whether you break a problem into parts that don’t overlap and that together cover everything.
- Quantitative analysis: whether you get the math right and, above all, whether you know what the result means.
- Business judgment and creativity: whether your ideas make sense for this client and don’t come from a memorized list.
- Reading data: whether you find what matters in a chart or a table without describing all of it.
- Communication: whether you get to the point, say where you’re going before you start and close with a conclusion.
The five question types
| Type | What they ask | What a good answer does |
|---|---|---|
| Opening structure | “What would you look at to solve this?” | Ask for a minute, build 3 or 4 branches designed for this case and explain which one you’d check first and why. |
| Chart or table | They hand you some data and ask what you see. | Start with what stands out and what it means for the client, not with a description of the axes. |
| Math | A specific calculation: profit, breakeven, growth, capacity. | Say how you’ll calculate before you start, watch the units and zeros, and finish with what the number means. |
| Brainstorming | “What other options does the client have?” | Ideas sorted into groups, not a loose list. Categories first, then examples. |
| Synthesis | “The CEO walks into the room. What do you tell her?” | Recommendation first, two or three reasons with numbers, one risk and the next steps. |
Not every case has all of them, and sometimes the case ends without a formal synthesis. Even so, it’s worth closing every answer with a sentence about what it means for the client: it’s a small synthesis.
An example case, solved
Kora is a chain of 40 gyms. Last year its revenue grew, but its profit fell by more than half. The CEO wants to know why and what to do.
Question 1: the structure
“What would you look at to understand why profit fell?”
Ask for a minute to organize your thoughts, then present something like this:
- Revenue: how many members there are on average, how much each one pays and other income, such as paid classes or the shop.
- Costs of running the gyms: rent, staff, energy and maintenance. These are mostly fixed.
- Costs of getting members: advertising, promotions and commissions. They depend on how many new members you need to sign up, and that depends on how many leave.
- The market: whether cheaper competitors arrived or what customers want has changed.
And close with a hypothesis: “Since revenue went up, I’d start with costs. In particular I’d look at acquisition, because if more members are leaving, you have to spend more to replace them.”
Notice that this isn’t a generic “revenue minus costs” outline. Separating the cost of getting members from the fixed costs already says something about how a gym works, and that’s what is being assessed.
Question 2: the table
“The team pulled this data together. What stands out to you?”
| 2024 | 2025 | |
|---|---|---|
| Average members | 60,000 | 66,000 |
| Average monthly fee | US$30 | US$28 |
| New members during the year | 24,000 | 42,000 |
| Cost to acquire one new member | US$100 | US$120 |
| Fixed costs for the year | US$15.0 million | US$15.4 million |
A good answer goes straight to the odd part: “42,000 new members joined, but the average only went up by 6,000. So about 36,000 people left during the year, more than half the base. If the base was stable in 2024, about 24,000 left then, 40%. Churn got worse, and on top of that each new member costs more and pays a lower fee.”
It’s a rough calculation, because the table shows averages, and it’s fine to say so. What matters is that a five-row table turned into a single idea.
Question 3: the math
“Work out the profit for each year. How much of the drop comes from acquisition?”
First say how you’ll calculate it: “Profit equals fee revenue minus fixed costs minus acquisition costs. I’ll work out both years and compare each part.”
Fee revenue (members × fee × 12 months):
- 2024: 60,000 × 30 = 1,800,000 a month, × 12 = US$21.6 million.
- 2025: 66,000 × 28 = 1,848,000 a month, × 12 = US$22.2 million (22.176, to be exact).
A shortcut for 1,848,000 × 12: it’s 1,848,000 × 10 plus 1,848,000 × 2, so 18,480,000 + 3,696,000 = 22,176,000.
Acquisition cost (new members × cost per member):
- 2024: 24,000 × 100 = US$2.4 million.
- 2025: 42,000 × 120 = US$5.04 million.
Profit:
- 2024: 21.6 − 15.0 − 2.4 = US$4.2 million.
- 2025: 22.176 − 15.4 − 5.04 = US$1.736 million, about 1.7.
Profit fell by about US$2.46 million, close to 59%. Now the part that matters: revenue went up by 0.58 million and fixed costs by 0.4. Acquisition went up by 2.64 million, more than the whole drop. Without that increase, profit would have grown.
Close with what it means: “The problem isn’t the gyms, it’s member churn. Kora is paying more and more to fill a bucket with a bigger hole in it.”
Question 4: brainstorming
“How could Kora get fewer members to leave?”
Instead of throwing out loose ideas, sort them by the stage of the member’s life:
- Before joining: aim promotions at people more likely to stay (for example, near their work or home), and swap discounts for a first month with extra benefits instead of a lower fee forever.
- The first months, when most people drop out: an initial assessment, a plan with a trainer and a follow-up if someone stops coming for two weeks.
- Along the way: group classes, which build commitment to other people, and schedules that fit those who go before or after work.
- When leaving: offer to freeze the membership for a month before cancelling, and ask why they’re leaving, to know what to fix.
Grouping them like this shows you thought through the whole problem, and makes it easy to follow with “I’d start with the first months, because that’s where most of the churn happens.”
Question 5: a scenario and the synthesis
“Kora is considering going back to a US$30 fee with no discounts. It estimates that 30,000 new members would join each year, the average base would settle at 62,000 and everything else would stay the same. Is it worth it?”
- Revenue: 62,000 × 30 = 1,860,000 a month, × 12 = US$22.32 million.
- Acquisition: 30,000 × 120 = US$3.6 million.
- Profit: 22.32 − 15.4 − 3.6 = US$3.32 million, about 1.58 million more than today, almost double.
And the synthesis for the CEO:
“I recommend dropping the discounts and tackling member churn. Profit fell from 4.2 to 1.7 million dollars because acquisition costs doubled: more than half the members now leave each year, and replacing them costs more. Going back to the 30-dollar fee, even with fewer members, would lift profit to about 3.3 million. The risk is that the estimate of 62,000 members is optimistic, so I’d test it first in a few gyms and, at the same time, work on the member’s first months, which is where most are lost.”
Recommendation, reasons with numbers, a risk and a next step, in about 30 seconds.
How to do the math without a calculator
- Write it down before calculating. One line with the formula prevents half the mistakes and lets your interviewer follow your reasoning.
- Set the zeros aside. 42,000 × 120 is 42 × 12 = 504, plus four zeros: 5,040,000.
- Split hard multiplications. × 12 is × 10 plus × 2. × 25 is × 100 divided by 4.
- Don’t over-round. The guides say McKinsey expects precise answers. Round only if you say so and it doesn’t change the conclusion.
- Check that the number makes sense. If a gym turns out to earn more than a mining company, something went wrong with the zeros.
- End with the “so what”. A number without an interpretation is only halfway there.
To train all of this, the mental math trainer gives you new questions every time: big numbers, percentages, growth and margins.
How to handle an interviewer-led case
- Don’t just wait. The interviewer leading doesn’t make you a passenger. End each answer by saying what you’d look at next: it’s often exactly the next question.
- Listen carefully to the question. “What stands out to you” is not the same as “describe the table”.
- Ask for a moment when you need it. Thirty seconds of silence to organize your thoughts looks much better than a messy answer.
- Ask if something isn’t clear, at the start and whenever new data appears.
- Go back to the goal. Every answer should connect to what the client wants to decide.
Common mistakes
- Using a memorized outline that doesn’t fit the case. McKinsey says on its careers site that there is not one logical solution.
- Describing the whole chart instead of saying what matters.
- Starting to calculate without saying how. If you get the math wrong, nobody can see that the reasoning was right.
- Getting the zeros wrong and not noticing.
- Brainstorming with no order, which ends up looking like a shopping list.
- Forgetting the PEI. It counts as much as the case in every interview.
How to prepare
| Stage | What to do |
|---|---|
| Week 1 | Read the sample cases McKinsey publishes on its careers site and solve them on your own, timing each question. |
| Weeks 1 to 4 | Ten minutes of mental math every day. |
| Weeks 2 to 4 | Cases with a partner who plays the interviewer, and about 15 minutes of feedback after each one. The guides talk about 15 to 25 cases in total. |
| Throughout | Keep a list of your mistakes and work on one at a time: structure, math, reading charts or synthesis. |
| Last week | Few cases, done well. Review your PEI stories and rest. |
Practicing with a partner matters more than reading: the case is judged by how you say it out loud, not by what you know.
Frequently asked questions
Can I use a calculator? No. You do the math by hand, and having paper to write it down helps.
Do I need to know the case’s industry? No. The case gives you the data you need. Business common sense does help: how a company makes money and what makes it spend.
Do I have to use a well-known framework? No. Knowing the classic structures is a useful starting point, but what’s assessed is building one designed for the case.
What if I make a math mistake? If you notice, say so and fix it. Making a mistake once and correcting it is much better than carrying on with an absurd number.
Where can I find real McKinsey cases? McKinsey publishes sample cases on its careers site and asks candidates not to share the content of their interviews. The case in this guide is our own.