The case interview is how McKinsey, BCG, Bain, and every tier-2 firm decide who gets offered. You walk into 25 to 40 minutes with a deliberately ambiguous business problem — "Our client is a regional airline losing market share. What would you do?" — and you have to structure it, do the math out loud, and land a recommendation. This guide covers the four pillars the round actually tests, the case patterns that show up in 80% of interviews, and how to drill them against an AI interviewer that probes the way partners do.
What a case interview looks like
Most cases follow a consistent rhythm, though firms vary in how much they interrupt:
- Prompt and clarifying questions (1–3 min). The interviewer gives you a short prompt. You confirm the objective, time horizon, and what success looks like. Ask two or three sharp clarifying questions before you structure.
- Structure (2–4 min). You take a minute of silence to organize your thinking on paper, then lay out a framework — usually three to four buckets — that covers the problem space without double-counting. You check with the interviewer before proceeding.
- Analysis (10–20 min). You drive through the framework, pulling on the thread the interviewer guides you toward. Expect quantitative exhibits (charts, tables), mental math, and follow-up questions.
- Synthesis and recommendation (2–4 min). You summarize what you learned and make a clear recommendation — a specific action — supported by the two or three most important findings. You flag risks and next steps.
McKinsey runs more interviewer-led cases, where the partner drives the flow. BCG and Bain lean candidate-led, where you are expected to direct the analysis. Tier-2 firms vary. The underlying skill is the same.
The four pillars interviewers score
- Structure. Is your framework MECE — mutually exclusive, collectively exhaustive? Does it fit this specific problem rather than being a memorized template forced onto it?
- Quantitative fluency. Can you do mental math accurately under pressure? Can you read an exhibit and extract the insight without being hand-held?
- Business judgment. Do your recommendations make sense for a real executive? Do you anticipate second-order effects, competitive response, and implementation risks?
- Communication. Is your structure signposted clearly? Do you summarize before moving on? When you synthesize, is it crisp — "Based on X, Y, and Z, I recommend A, with the main risk being B"?
Polish matters more in consulting than in most other tracks. A candidate who mumbles the right answer loses to a candidate who confidently delivers the second-right answer. This is not because consulting values confidence over correctness; it is because the job is presenting findings to executives, and the interview simulates that.
Common case types
About 80% of cases fit one of these patterns. Drilling one per pattern fluently is a better investment than reading 40 different cases.
- Profitability. Profit is falling — is it revenue or costs? Which product line, which customer segment, which cost bucket? The classic "decompose profit and diagnose" structure.
- Market entry. Should our client enter this market? Market sizing, attractiveness, competitive dynamics, our ability to win, financial viability.
- M&A. Should our client acquire this target? Strategic fit, valuation, integration risk, regulatory path.
- Pricing. Is the client pricing correctly? Cost-plus, competitor-based, willingness-to-pay, price elasticity.
- Growth strategy. How should our client grow revenue 20% over three years? Organic vs. inorganic, current markets vs. new, product vs. channel.
- Operations. Plant is inefficient, supply chain is slow, customer service is expensive — diagnose the bottleneck and size the fix.
- Market sizing. How many golf balls fit in a Boeing 747? How many tires are sold in the US each year? Pure Fermi estimation, usually as a warm-up or standalone mini-case.
Preparation roadmap
- Weeks 1–2: Frameworks and mental math. Learn the canonical frameworks — profitability tree, market entry, M&A — and practice using them as starting points, not endpoints. Drill mental math until you can multiply three-digit numbers, compute percentages, and work in 10^6 and 10^9 without slowing down.
- Weeks 3–4: Cased by peer or book. Work through 15–20 cases with structured feedback. Focus on pattern recognition — what type is this, what are the two or three most likely drivers, what numbers will I need?
- Week 5: Exhibit fluency. Drill exhibit-heavy cases. Given a chart, extract the insight in 30 seconds. Given a table, identify the anomaly in under a minute. This is the most under-practiced skill and the most differentiating one.
- Week 6: Full mocks under time pressure. Four to six full cases per day with voice and follow-ups. Against a timer. With synthesis at the end of each.
Every candidate underinvests in synthesis. The last two minutes of a case — where you deliver the recommendation — carry disproportionate weight because partners remember the ending.
How to practice with InterviewDen
The Case Interview track on InterviewDen runs a full case with a voice-driven AI interviewer that reads your structure, pushes back on weak branches, presents exhibits at the right moments, and asks for your synthesis at the end. The interviewer does mental math checks — it catches arithmetic errors, asks you to verify your units, and plays the role of a partner who is deliberately hard to impress.
The debrief scores structure, analysis, business judgment, and communication, with specific moments in the transcript flagged. You see exactly where your framework was not MECE, where your math went wrong, and where your synthesis was strong or vague.
Common mistakes
- Memorized frameworks forced onto the wrong case. Using the profitability tree on a market-entry case is a red flag. Frameworks are starting points; tailoring them to the specific prompt is the skill being tested.
- Silent math. Doing arithmetic in your head without narrating means interviewers cannot follow your reasoning and cannot tell whether you made a logic error or an arithmetic error.
- Burying the recommendation. If the interviewer has to ask "so what would you recommend?" your synthesis has failed. Lead with the recommendation, then support it.
- Ignoring the client context. A recommendation that works in a McKinsey report but is impossible for this specific mid-sized client to implement is a weak recommendation.
- Not pausing to take stock. Strong candidates pause, summarize, and check in with the interviewer every five to seven minutes: "So far I've learned X and Y. I want to look at Z next — does that make sense?"
- Treating exhibits as decoration. An exhibit is always the key to the next insight. If you are skimming past it without extracting a specific number, you are missing the point.
FAQ
How many cases should I do before my first interview?
Around 30–50 practice cases is the standard range, with at least 15 of those as live mocks rather than self-casing. Quality matters more than quantity past 20 cases.
Do I need to go to a target school to get into MBB?
No, but it is harder from a non-target. Networking and undergraduate case competitions matter more for non-target candidates. The interview itself is the same.
How much mental math do I actually need?
Enough to compute 17% × $480M in your head within 15 seconds without slowing the conversation. You do not need to be a savant; you need to be fluent and reliable.
Is the PEI round (personal experience interview) at McKinsey scored separately?
Yes, and it is roughly 50% of the weight. Prepare three PEI stories — personal impact, entrepreneurial drive, leadership — with STAR structure, and drill them until they are under three minutes each.
Are cases still relevant now that consulting is using more AI?
Yes. Firms still hire new analysts and consultants against the same case format. What has shifted is the emphasis on business judgment and synthesis over pure analytical grind — exactly the things a case tests.
Candidate-led vs. interviewer-led — how should I adapt?
Ask the interviewer how they want to structure it. If they drive, follow their lead and answer crisply. If they sit back, take ownership — pose the questions, do the math, signpost your next move.
Can I use a calculator?
Usually no. Most firms expect mental math. A few allow paper arithmetic. None allow a calculator.