In a McKinsey first-round slate of eight, three or four candidates will treat the case like a quiz: wait for the next question, answer it cleanly, then go quiet. Their trees are tidy. Their math is fine. They still get dinged. The interviewer was not looking for a polite quiz-show contestant.
McKinsey's problem-solving interview is interviewer-led, which means the firm sequences discrete scored asks. What the sheet rewards changes at each hand-off: structure, exhibit, brainstorm, quant, synthesis. After this page you should know what a pass sounds like at each stage, run a fully worked mini-case in that shape, and stop practicing as if McKinsey were just a louder version of a candidate-led firm.
What McKinsey actually publishes
McKinsey's interviewing page is the primary source. For most client-facing roles it describes a personal experience interview (PEI), then a problem-solving interview (the case). Expertise interviews appear for some roles. Your recruiter confirms the logistics; offices differ.
On the problem-solving interview, the firm is explicit about the skill set it wants to see: how you structure tough, ambiguous challenges, identify important issues, deal with the implications of facts and data, formulate conclusions and recommendations, and articulate your thoughts in a fast-moving discussion. That sentence is scoring language, not marketing.
McKinsey also publishes free practice cases with suggested logic. Start with the ones on the interviewing hub — among them Beautify, Electro-Light, Diconsa, and Talbot Trucks. Read them as format samples: overall approach first, then a sequence of narrow questions, calculators not allowed, talk through the arithmetic so the interviewer can follow (and rescue) you.
Slot length is not a secret McKinsey leaks on Glassdoor. A representative public description from McKinsey's Middle East how-to-apply page puts a typical individual interview around 40 to 50 minutes with three parts: experience, case discussion, and your questions. Treat that as a planning range. Confirm your invite.
CoachNed is independent practice material and is not affiliated with McKinsey.
The five hand-offs and what each one scores
Interviewer-led does not mean passive. It means the interviewer chooses the path; you still own the quality of every answer. I score McKinsey-shaped cases as five separate marks that add up. A brilliant synthesis does not erase a template tree. A perfect number with no setup does not erase a mute exhibit read.
| Stage | What they ask | What a pass sounds like | What I stop listening for |
|---|---|---|---|
| Structure | "How would you approach this?" | Two or three client-specific branches, priority named, hypothesis you would kill | Profitability template on a public-sector or ops prompt |
| Exhibit | "What do you take from this?" | Describe, quantify, implicate for the hypothesis | Axis labels read aloud, then silence |
| Brainstorm | "What else could drive this?" | Two named buckets, four ideas each | Nine orphan ideas in random order |
| Quant | A calculation with given numbers | Equation out loud, units, then the number, then the so-what | Magic correct answer with invisible logic |
| Synthesis | "CEO walks in — what do you say?" | Answer first, one number, one risk, one next step | Recap of the analysis with a hedge |
The exhibit habit is the one candidate-led practice least trains. McKinsey typically hands charts one at a time with a narrow ask. If you practiced only "drive the case," the room feels like someone stole the steering wheel. The correction is mechanical: after every handoff, say what the number means for the hypothesis before they ask.
McKinsey's own practice cases reinforce the sequence. Electro-Light walks from launch factors to market data to a break-even share ask to "what would SuperSoda need to do to hit the target." That is not one continuous debate. It is four scored beats.
Worked mini-case: HarborCold, not a grocery chain
Prompt. HarborCold runs three refrigerated warehouses near Mid-Atlantic ports. Two years ago EBITDA margin was 18 percent on $84 million revenue. Today revenue is flat at $84 million and margin is 12 percent. Utilization: Warehouse A 92 percent, B 61 percent, C 74 percent. Variable cost is mostly energy and labor tied to cubic feet occupied. Fixed cost is rent, management, and cold-plant maintenance. The CEO wants a recommendation in six weeks: close B, raise rates system-wide, or invest $6 million in automation at A and C?
Stage 1 — Structure (keep it short; they will steer). Three branches: (1) mix / utilization — is the margin hit coming from underused capacity, especially B; (2) price — are rates lagging cost inflation; (3) cost intensity — energy or labor per occupied cubic foot up even where utilization is healthy. Hypothesis: production of "cold space sold" is not the story if revenue is flat; underused fixed cost at B plus possible rate lag is the first test. I would ask for contribution by warehouse before I talk about automation.
Stage 2 — Exhibit. Interviewer hands a table: contribution after variable cost is A $11.2 million, B $1.4 million, C $6.8 million. Fixed costs allocated by square footage: A $4.0 million, B $3.6 million, C $3.8 million. Night energy rates rose 22 percent; A and C run more overnight pulls than B.
Describe, quantify, implicate. A carries most of the profit. B's contribution barely covers half its allocated fixed cost ($1.4 vs $3.6). Closing B would remove $3.6 million of fixed cost if that rent and plant cost are avoidable, and lose only $1.4 million of contribution — a $2.2 million EBITDA lift if the fixed cost is real and exit costs are small. Rate increases system-wide would tax A and C customers who already fill the profitable sheds. Automation at A and C does not fix B's hole.
Stage 3 — Brainstorm (they ask why B is empty). Two buckets: demand-side (shippers shifted to inland dry ports; contract loss of one grocery chain; seasonal seafood down) and supply-side (B's dock doors cannot take 53-foot reefers; temperature zones mis-specced for pharma; sales team compensated on A first). Four ideas each beats a laundry list.
Stage 4 — Quant. They give: avoidable fixed cost at B if closed = $3.1 million (not the full $3.6; $0.5 million is shared corporate). One-time exit and customer move cost = $0.9 million. Of B's volume, 40 percent can relocate to C at 80 percent of prior contribution margin (C has space; A is full).
Math out loud:
- Lost contribution if B closes and nothing relocates: $1.4 million.
- Recovered contribution from relocation: 0.40 × $1.4 × 0.80 = $0.448 million.
- Net annual contribution lost: $1.4 − $0.448 = $0.952 million.
- Net annual fixed-cost save: $3.1 million.
- Steady-state EBITDA lift: $3.1 − $0.952 ≈ $2.15 million.
- Payback on the $0.9 million exit cost: under six months.
Compare automation: $6 million capex for a claimed 15 percent variable-cost cut on A+C. Variable cost on A+C contribution base is not given cleanly; even if you generously assume $10 million of variable cost on A+C, 15 percent is $1.5 million a year — worse near-term cash than closing B, and it leaves the empty shed.
Stage 5 — Synthesis (60–90 seconds). "Close B within the quarter, relocate the movable 40 percent of volume to C, and do not raise rates system-wide yet. Steady-state EBITDA lift is about $2.1 million a year after the avoidable fixed-cost save, with exit cost paid back in under six months. Hold the $6 million automation until A and C are full and you have a clean variable-cost baseline. The risk is that $3.1 million of 'avoidable' fixed cost includes a lease break fee the exhibit did not show — I would confirm the lease before the board memo."
That is the McKinsey shape: client-specific tree, one exhibit with an implication, arithmetic with units, a recommendation that names a number and a risk. Official cases on McKinsey's site train the same motion with different industries; this one is original practice, not a firm prompt.
Ned's rule. After every McKinsey hand-off, volunteer the implication for the decision in one sentence before you wait for the next question. Interviewer-led is not a gift of passivity; it is a gift of a narrower ask. Use the width you still have.
PEI sits beside the case, not under it
McKinsey's interviewing page says the PEI comes first for most client-facing roles, then the problem-solving interview. It asks you to come with personal examples in detail — your specific role and the actions that mattered — and suggests two examples for each of four themes listed on that page. Labels on themes have shifted in public materials over time; match whatever your recruiter names on the invite, not a blog's rename chart.
I treat PEI and case as separate fail conditions. A strong HarborCold close does not rescue a hollow story when the interviewer digs on your actual decisions under pressure. Prepare depth, not slogans. For story structure and probes, use the PEI guide and the behavioral practice path.
Where candidates leak points in this room
- Going mute because they are steering. After every exhibit, you still own the so-what.
- Template trees. A generic profitability frame on a warehouse, a ministry, or a network design reads as you did not hear the prompt.
- Math without a setup. A slightly wrong number with a clear equation scores better than a silent miracle.
- Synthesis that recaps. If sentence one is not the recommendation, you burned the ninety seconds.
- Ignoring PEI. The case does not average with a thin personal story.
- Skipping the official cases. Beautify and Electro-Light exist so you learn the hand-off rhythm before you invent frameworks.
Online assessments (Solve and its modules) are a separate screen in many pipelines. They are not a substitute for this case. Confirm the invite; do not let an assessment eat the week you needed for interviewer-led reps.
Drill this shape today
One concrete assignment: open Electro-Light and speak every answer out loud as five labeled stages — structure, exhibit, brainstorm or qualitative, quant, synthesis — timing the close at ninety seconds. Then run HarborCold again from memory without looking at the numbers until you have stated the equation.
For a scored first rep without an account, use /start (three typed turns, instant scores). For a full voice loop with a debrief, use /interview (CopperCup, about eighteen minutes). Isolate structure and chart habits on the structure and charts drills; close practice on synthesis. More interviewer-led reps live in the McKinsey practice cases bank and the case library. Everything is open for seven days, no card; then $120 for a recruiting season or $49 a month.
Frequently asked questions
How long is a McKinsey case interview?
Public McKinsey materials describe a typical individual interview in the 40-to-50-minute range, combining PEI (or experience questions), a problem-solving case, and time for your questions. Exact timing varies by office and round. Use your recruiter's invite as the schedule, not a blog's average.
Is the McKinsey case interviewer-led or candidate-led?
McKinsey's published practice cases are built as a sequence of interviewer questions with suggested logic — overall approach, then data and calculations, then recommendations. In the room that usually feels interviewer-led: they choose the next exhibit and ask. You still structure, hypothesize, quantify out loud, and synthesize. Do not sit and wait.
Do I need business knowledge for McKinsey cases?
McKinsey frames the problem-solving interview around structuring ambiguity, using facts, and communicating a recommendation — not around industry trivia. You need enough commercial sense to interpret an exhibit. You do not need prior consulting experience or a memorized framework library.
How should I use McKinsey's official practice cases?
Work them aloud, stage by stage, with a partner or a timer. Use the firm's suggested logic as a check after you answer, not as a script to memorize. Then switch to original prompts (like HarborCold or the practice bank) so you are not rehearsing known answers.
What is the difference between PEI and the case?
PEI tests how you behaved in real situations — your role, actions, and impact across the themes McKinsey lists. The case tests how you solve a hypothetical client problem under a clock. Both can fail you independently. Prepare both.
Sources
- Interviewing at McKinsey — PEI then problem-solving interview; what the firm says it evaluates; links to sample cases. Checked 2026-09-24.
- Electro-Light practice case — official interviewer-led sequence: launch factors, market data, break-even share, go-to-market ask. Checked 2026-09-24.
- Beautify practice case — official format sample with virtual-channel and AI exhibit logic. Checked 2026-09-24.
- Diconsa practice case — structure, survey exhibit, qualitative sales challenges. Checked 2026-09-24.
- Talbot Trucks practice case — TCO comparison and pricing math without a calculator. Checked 2026-09-24.
- How to apply — McKinsey Middle East — representative 40–50 minute interview structure (experience, case, candidate questions). Checked 2026-09-24.