McKinsey Case Interview Examples: 10 Interviewer-Led Practice Cases (2026)

McKinsey case interview examples are worked simulations of an interviewer-led problem solving interview: you open with a structure and a hypothesis, the interviewer hands you one exhibit at a time and controls the question order, and you close with a synthesis that leads with the answer. This page is the practice bank. Below are 10 fully worked McKinsey-style cases spanning profitability, market sizing, market entry, M&A, operations, pricing, and public sector prompts, each with the prompt, the opening structure, the exhibit the interviewer would hand you, the arithmetic, and the recommendation. Around them sit the parts of the round the cases alone do not cover.

Every McKinsey interview pairs the case with a Personal Experience Interview, which runs 10 to 20 minutes inside an interview slot of roughly 50 minutes and probes a single story from every angle rather than sampling several. Before the live rounds you sit Solve, which in 2026 most commonly arrives as two modules, Red Rock Study at around 35 minutes and Sea Wolf at around 30 minutes. The McKinsey case interview guide covers that process end to end. This page is the material you practise on.

McKinsey practice cases at a glance

How a McKinsey case works: the interviewer leads. You open with a structure and a hypothesis, then the interviewer sets the question order and hands you one exhibit at a time, over roughly 25 minutes of a 50-minute slot that also holds the PEI. Inside the case you meet five separate asks: structure, exhibit, brainstorm, calculation and synthesis, each scored on its own.

Start free now: EV Charging Hub Profitability is an interviewer-led profitability case on a Spanish charging network with cafes attached: 8 questions, 3 with math, 3 exhibits, and a median of 17 minutes across 45 completed voice runs. An AI interviewer runs it out loud and releases each exhibit when you reach it, the way a McKinsey interviewer would. Your first full case is free when you create an account, and the worked excerpt further down runs its numbers. A free account also gets one full run in every simulator inside the app: a complete McKinsey Solve simulation each of Red Rock, Sea Wolf and Sustainable Futures Lab, a full Casey-style case, a full CCA-style diagnostic and a complete Bain SOVA-style diagnostic.

These are original practice cases written in McKinsey's interviewer-led format, 5 of the 22 McKinsey cases in the library; none is McKinsey's own material. Every debrief scores seven categories: structure (25% of the score), quantitative analysis (20%), business judgment (15%), and problem understanding, exhibit reading, communication and synthesis (10% each), with written feedback on each question. Median times come from completed runs between June 25 and September 22, 2026, and count runs, not people.

McKinsey's six official cases are published free at mckinsey.com/careers/interviewing, each with a prompt, exhibits and a suggested approach. Work them first, then use the library cases above for interviewer-led reps with scored feedback.

What should a McKinsey practice case example show?

Most published "examples" stop at the prompt and a framework. That is the half of a case nobody fails. Candidates lose McKinsey rounds in the middle, on the exhibit they cannot quantify and the synthesis that never names a number, and an example that skips those teaches nothing you can be scored on.

Every one of the 10 cases below carries all seven. Read only the prompt, build your own structure for two minutes, then compare.

McKinsey vs BCG vs Bain: what actually changes in your practice

Practising for the wrong format is the most common self-inflicted wound in MBB prep. The three firms score the same underlying skills, but the room behaves differently, and a habit that earns points in one room costs them in another.

Get your McKinsey prep plan

Build my prep plan

1

Learn the format

17 steps

• Learn how to caseGuided learning · 35 min
• The Case Interview MethodLesson · 8 min
• Profitability & Turnaround: AfrihealSkill drill · 5 min
• Profitability & Turnaround: Berty BankSkill drill · 5 min

The trap runs in one direction more than the other. Candidates who drill candidate-led cases and then sit a McKinsey round often go quiet, waiting to be told what to do, because the interviewer is visibly steering. Interviewer-led does not mean passive. You still open the structure, you still state the hypothesis, and after every number you volunteer the implication before you are asked for it. What you give up is choosing the path, not choosing to think.

The fastest way to feel that difference is to run one interviewer-led profitability case where the exhibits arrive on someone else's schedule.

Practice a McKinsey profitability case end to end. A five brand US automaker is losing money and must decide which brands to keep. Answer this McKinsey profitability case step in the article. Read the worked Automotive Case Interview with exhibits

The five question types inside a McKinsey case

A McKinsey case is not one continuous conversation. It is a sequence of discrete asks, each scored on its own, and the interviewer moves you between them. Recognising which type you are in tells you what a good answer even looks like.

1. The structuring question. "How would you think about this?" You get roughly two minutes of silence to build a tree. What is scored is whether the branches are MECE, whether they are specific to this client rather than a template you recognised, and whether you name which branch you would test first and why. A tree with no stated priority is an unfinished answer.

2. The exhibit question. A single chart or table arrives, usually with a narrow ask attached: "What do you take from this?" The three-beat answer is describe, quantify, then implicate. Most candidates do the first, some do the second, and the ones who get callbacks always do the third. Technique for the read itself is in reading charts and exhibits.

3. The brainstorming question. "What could be driving this?" or "What else would you consider?" This is scored on whether your ideas have a spine. Six ideas grouped into two named categories beats nine loose ones, because the grouping is the evidence of structured thinking.

4. The quantitative question. One or two calculations per case, set up by you, done out loud, units attached. The setup carries more weight than the arithmetic. Say the equation before you compute it so the interviewer can correct a wrong path before you spend ninety seconds down it.

5. The synthesis question. "The CEO just walked in. What do you tell them?" Sixty to ninety seconds, answer first, one supporting number, one risk, one next step. It is graded separately from everything above it, which is why strong analysts still lose rounds here. The method is in the synthesis guide.

The one you cannot grade yourself is the first. You will always believe your own tree was MECE and hypothesis-led, because you wrote it. Get one scored by something that did not, such as the voice case at the top of this page.

The PEI: the other half of every McKinsey interview

Every McKinsey interview slot carries a case and a Personal Experience Interview, and the two are weighted as separate components. The PEI runs 10 to 20 minutes of a slot of roughly 50 minutes, with the case taking around 25 minutes and questions at the end. Candidates routinely spend eight weeks on cases and one evening on the PEI, then lose the round on the half they did not rehearse.

The shape is depth, not breadth. You get one opening question and then that single story is probed from every angle, with follow-ups running into double figures. McKinsey assesses four named dimensions, currently published as Leadership, Connection, Growth, and Drive, and the standard advice is two distinct stories per dimension so a repeat question does not corner you into reusing one. Notably, problem solving is not one of the dimensions. The firm tests analytical ability in the case, so a PEI story that is really a project summary answers a question nobody asked.

Airtime is where prepared candidates still lose points. The bulk of a five to eight minute story belongs to what you personally did, not to setting the scene. If your situation and complication run three minutes, you have spent the interviewer's attention on context and left ninety seconds for the actions being scored.

Full question banks, the dimension mapping, and worked story structures live in the McKinsey PEI guide and the PEI question list. When you are ready to rehearse out loud rather than read, run the stories through behavioral practice so the follow-up probing happens before the interviewer does it.

Rehearse your McKinsey PEI story and follow-ups

Your interviewer will ask:

“Tell me about a time you convinced someone who disagreed with you.”

“What did you personally do, and what changed because of your actions?”

Practice McKinsey PEI free

McKinsey Solve: the games you sit before any case

Solve is the digital assessment between the resume screen and the live rounds. It matters to your case prep for one practical reason: it consumes a week of your calendar at exactly the point you wanted to be running cases, and candidates who ignore it until the invitation arrives lose that week from case volume.

In 2026 the standard invitation contains two modules. Red Rock Study runs around 35 minutes and is a data interpretation case built on wildlife population data, structured in three phases: investigation, where you read source material and decide which numbers matter, analysis, where you run calculations on what you collected, and report, where you complete a short written summary. Sea Wolf runs around 30 minutes and is an ecosystem optimisation game where you assemble a set that satisfies hard constraints. An invitation of roughly 65 minutes usually means those two. An 85-minute invitation adds the 20-minute Sustainable Futures Lab, and a 95-minute one, reported since late August 2026, adds a newer 30-minute version of it.

The scoring is not what most candidates assume. On the data module, published guidance from assessment trackers puts the weighting at roughly 70 percent accuracy and 30 percent process, meaning how you got there is measured, not just the final value. Sea Wolf penalises constraint violations directly. McKinsey does not publish a passing score and you never see your own result.

Road to Offer offers one complete free run of each of four practice formats: Red Rock, Sea Wolf, SFL Classic and SFL Island, using original independent material. Red Rock includes the full Study and six Cases on a 35-minute practice clock; Sea Wolf includes all three sites on a 30-minute clock; Classic includes all 13 linked decisions without a countdown; Island includes onboarding and all three project days on a 30-minute clock. Start without an account, then create a free account to finish and keep your score and worked debrief. No payment is required for the first run or its review. The most efficient rehearsal is the constraint discipline in Sea Wolf, because it is the module where candidates lose points to rule-breaking rather than to reasoning. Run the free rounds and see whether you filter before you optimise.

The free debrief shows where points went and worked feedback on each submitted decision, with highlighted calculations and charts in Red Rock, constraint and treatment visuals in Sea Wolf, decision and dependency diagrams in Classic, and evidence chains, placement maps and team feedback in Island. Paid Learning mode adds step-by-step coaching and feedback as you go, untimed in the clocked games, with the matching game pack, Solve pack or Annual access. The same access includes the full step-by-step PDF guide for each covered game (SFL Classic, SFL Island, Sea Wolf or Red Rock); a free account unlocks only a 6-page preview of each.

The data module is the other half of most sittings, and it is the closest Solve gets to reading an exhibit in a case. Run it free here, and switch to Sea Wolf or the SFL island in the same block:

Sea Wolf rewards eliminating anything that breaks a hard rule first, then optimising what survives. For the data module and the third module, work the Red Rock study guide and the Sustainable Futures Lab guide separately (the new 30-minute format has its own free island run), because the switch between optimisation and data interpretation is itself the thing candidates fumble. The full assessment walkthrough is in the McKinsey Solve guide and the Sea Wolf guide.

The island is the newer 30-minute judgment format; its free run is linked in the cards above, and comparing its calls with Sea Wolf's rules shows where each one tests you.

Cases come after Solve, so plan both: build a free Solve plan from your invitation length and test date, or select unsure if you have no invitation yet. A free account lets you view and save dated practice steps for up to 28 days; the plan is not based on your score. Keep these cases for the days after Solve.

Worked excerpt: the EV charging case, exhibit by exhibit

This is the first half of EV Charging Hub Profitability from the table above, with the numbers from its own exhibits. Cover each answer and work the step before you read it.

The prompt. A client runs electric vehicle charging hubs across Spain, each with a cafe and retail space attached. Profit fell between 2022 and 2023. What is driving the decline, and what should the client do?

A structure that works. Profit equals revenue minus cost, split by the two businesses the client runs: charging, and cafe and retail. Two business lines can move in opposite directions, so the segment split comes first, and energy prices get checked as the obvious outside suspect.

Exhibit 1, electricity prices. Spain's price is €0.13 per kWh in 2022 and €0.13 in 2023, a 0% change. The benchmark price cannot explain a profit fall between those two years.

Exhibit 2, profit by segment (€ millions). Profit is revenue minus cost for each business line.

Exhibit 3, visits (thousands). Visits that can earn retail revenue are both-segment visits plus retail-only visits: 440 + 260 = 700 in 2022 and 250 + 180 = 430 in 2023. That is 270 fewer, a 38.6% fall, close to the 40% fall in retail revenue from €120M to €72M.

What a strong answer says. "Profit is down €29M, or 15.6%, and the whole decline sits in cafe and retail, which lost €48M while charging gained €19M. Energy prices were flat. Visits that include the cafe fell 38.6%, in line with the 40% retail revenue drop, so I would find out why charging customers stopped buying, then pilot charging-to-retail offers and measure incremental profit against a comparison group before scaling."

10 McKinsey case interview examples with worked solutions

Each case below is written the way a McKinsey interviewer would run it: prompt, your opening structure, the sub-question you would be handed, the exhibit value, the arithmetic, and the synthesis. Read the prompt only, cover the rest, and work it before you compare.

Case 1: National retail chain, margin decline

Prompt. A national retail chain with 400 stores has seen operating margin fall from 12 percent to 7 percent over three years on a $2B revenue base. Revenue is flat. The CEO wants to know why and what to do.

Your opening structure. Revenue is stated as flat, so it is not the driver, and saying so out loud is worth a point on its own. That leaves cost, split into cost of goods sold and operating expense, and operating expense split into labour, occupancy, logistics, and overhead. Hypothesis: the decline is operating-expense driven, most likely labour or occupancy, and the first test is whether cost of goods sold moved at all.

Interviewer question. "Here is the cost structure across the three years. What do you take from it?"

Exhibit. Cost of goods sold held flat at 55 percent of revenue. Selling, general and administrative expense rose from 33 percent to 38 percent of revenue.

The math. The SG&A move is 5 percentage points on a $2B base, so 0.05 x $2,000M = $100M of additional annual cost. The observed margin fall is 5 points, which on the same base is also $100M. The SG&A line accounts for the entire decline, with nothing left to explain.

Synthesis. "This is a $100M SG&A problem, not a pricing or sourcing problem. Cost of goods sold is flat, so the whole five point margin loss sits in operating expense, and labour and occupancy are the two candidates large enough to carry it. I would start with a labour scheduling audit and a lease renegotiation review. A 10 percent labour efficiency gain at $2M of labour per store recovers roughly $80M of the $100M. The main risk is that some of the increase is deliberate investment in service levels, so I would check whether the SG&A growth is concentrated in the stores with the strongest sales."

The structure method behind this decomposition is in the profitability framework guide.

The arithmetic here is easy and the framing is not, which is exactly why candidates lose it. Percentage points against a revenue base, held in your head, spoken out loud, while someone watches. That is a rehearsable skill on its own.

Quick Math

1. Revenue is $90M and cost is 70% of revenue. What is profit, in dollars?

   Build the percentage off the 10% anchor, then Profit = Revenue - Cost.

2. A market grows 9% a year. Using the Rule of 72, how many years does it take to double?

   Years to double = 72 / growth rate. It runs backwards too: divide 72 by the years a market took to double to recover its rate.

3. Segment A is 75% of revenue at a 16% margin. Segment B is 25% of revenue at a 40% margin. What is the blended margin, in percent?

   Blended margin = (Weight A x Margin A) + (Weight B x Margin B). Never average the two percentages: revenue mix moves this number.

Case 2: Public EV charging market size

Prompt. A private equity client is evaluating an investment in a US electric vehicle charging network. Before the strategy discussion, size the annual revenue market for public Level 2 charging in the United States today.

Your opening structure. Build bottom-up from vehicles rather than top-down from an industry figure, because the driver chain is defensible and the interviewer can challenge each link. Chain: electric vehicles on the road, the share that depends on public charging rather than home charging, sessions per vehicle per month, sessions a single station can serve, then revenue per station.

Interviewer question. "Walk me through your assumptions before you calculate anything."

The math. Take 4M electric vehicles on US roads. Roughly 40 percent lack reliable home charging and depend on public infrastructure, giving 1.6M vehicles. At 6 public sessions per vehicle per month, that is 9.6M sessions per month. A Level 2 station handles roughly 240 sessions per month at realistic utilisation, so 9.6M / 240 = 40,000 stations of demand. At $2,400 of annual revenue per station, the market is 40,000 x $2,400 = $96M, call it $100M.

Sanity check. State the range, not the point. The number is highly sensitive to the home-charging share: at 25 percent rather than 40 percent the market is $60M, and at 55 percent it is $132M. Naming that sensitivity unprompted is what separates a sized market from a guess.

Synthesis. "Public Level 2 charging is roughly a $100M annual revenue market, plus or minus about 35 percent depending on the home charging assumption, and it is currently supply constrained rather than demand constrained. That size does not support a returns case built on charging fees alone. The investment thesis has to rest on real estate positioning or on fast charging, which has different economics. My next test would be the utilisation curve, because station revenue is what the whole estimate hangs on."

The full driver-chain method is in market sizing step by step.

Sizing is the type where the interviewer challenges your assumptions in real time, so rehearse defending a number rather than producing one.

Case 3: Software company entering Southeast Asia

Prompt. A US mid-market software company with $500M of revenue wants to enter Southeast Asia. It has no regional presence. Should it enter, and if so how?

Your opening structure. Four branches: market attractiveness (size, growth, willingness to pay), competitive intensity, company fit (product localisation, regulatory exposure, capital), and entry mode (greenfield, acquisition, partnership). Hypothesis: entry is viable and the mode is the real question, so the analysis should spend its time there rather than relitigating whether the region is growing.

Interviewer question. "The regional SaaS spend is $8B today, growing 18 percent a year. Your client's product category is 12 percent of that spend and a realistic share for a new entrant is 5 percent. What is the opportunity?"

The math. Today: $8B x 12 percent = $960M addressable, x 5 percent capture = $48M. Three years out at 18 percent compound growth, $8B becomes roughly $13.2B, so $13.2B x 12 percent x 5 percent = about $79M. The opportunity runs $48M to $79M over the horizon, against a $500M revenue base, so this is a 10 to 15 percent revenue story, not a transformational one.

Synthesis. "Enter, but by partnership rather than greenfield. The three year opportunity is $48M to $79M, which is material at 10 to 15 percent of current revenue but not large enough to justify building a regional organisation from scratch. I would use Singapore as the hub for regulatory and talent reasons, budget $5M to $8M for product localisation, and treat greenfield expansion into Indonesia or Vietnam as a year three decision contingent on hitting share. The principal risk is that 5 percent share assumes a distribution partner we have not yet identified, so the partner search is the gating item, not the capital."

Branch selection for this case type is in the market entry framework.

Case 4: Analytics software acquisition

Prompt. A large consulting firm is considering acquiring a mid-size analytics software company for $180M. The target has $30M of revenue and $4.5M of EBITDA. Should the acquirer proceed?

Your opening structure. Three branches: strategic fit (does this close a capability gap the acquirer cannot build), financial assessment (entry multiple, credible synergies, integration cost), and risk (talent retention, technology integration, client conflict). Hypothesis: at $180M on $30M of revenue the price is full, so the deal only works if synergies are large and realisable, and that is the branch to test first.

Interviewer question. "The deal team models $6M of annual synergies, $3.5M from cross-sell into the acquirer's client base and $2.5M from cost. What does that do to the multiple?"

The math. Entry multiple on today's numbers: $180M / $4.5M = 40x EBITDA, and $180M / $30M = 6.0x revenue. Post-synergy EBITDA: $4.5M + $6M = $10.5M, so the effective multiple is $180M / $10.5M = 17.1x. Comparable analytics software transactions trade closer to 12x EBITDA, so even fully credited synergies leave the deal above market.

Quick Math

1. Cost synergies are $2.3M of G&A, $4.1M of procurement, $1.6M of facilities and $1.5M of shared sales operations. What is the annual cost synergy total?

   Cost synergy total = the sum of every annual line, each counted at full value because the acquirer controls them.

2. Gross revenue synergies are $15M a year and you apply a 40% execution haircut. What revenue synergy do you carry into the valuation? Give the exact figure, do not round.

   Haircut revenue synergies = Gross synergies x (1 - haircut), because revenue synergies depend on customers saying yes. Carry the exact figure.

3. Standalone value is $180M, the present value of synergies is $74M, and integration costs are $31M. What is the maximum price the acquirer should pay?

   Maximum price = Standalone value + PV of synergies - Integration costs. Above it, the seller keeps the whole synergy pool.

Synthesis. "The price is full even after synergies. Fully credited, the effective multiple is 17x against a comparable set closer to 12x, and cross-sell synergies are the softest half of that $6M because they depend on the acquirer's sales force selling a product it does not yet know. I would proceed only below roughly $130M, which brings the post-synergy multiple to about 12x. The alternative worth pricing is a commercial partnership, which plausibly captures the majority of the cross-sell benefit without the integration risk. My next test would be the retention terms for the target's engineering team, because at 40x trailing EBITDA the entire thesis is the people."

The branch structure and synergy treatment are in the M&A case framework.

Case 5: Bottling plant margin recovery

Prompt. A regional bottled water producer has seen gross margin fall 6 points in two years while volume grew 4 percent. Input costs are up but so is the industry's. The CEO wants the margin back within a year.

Your opening structure. Split the margin decline into price realisation, input cost, and conversion cost, then ask which moved most before proposing anything. Hypothesis: with volume growing and the industry facing the same input inflation, the client-specific loss is in price realisation or in a plant-level conversion problem, not in the raw material market.

Interviewer question. "Input costs rose 9 percent and the client passed through 4 percent of it in price. Conversion cost per litre rose 11 percent while industry conversion cost rose 3 percent. Where is the problem?"

The math. The pass-through gap is 5 points of input inflation absorbed rather than priced. But the sharper signal is conversion: 11 percent against an industry 3 percent means 8 points of client-specific cost growth, which is not an input market story at all. That is an operations problem sitting inside a case that was framed as a cost inflation problem.

Synthesis. "Two thirds of the margin loss is self-inflicted. Input inflation is real but industry-wide, and the client absorbed five points of it by under-pricing. The larger and more fixable problem is conversion cost, which grew 11 percent against an industry 3 percent, so eight points of that is specific to this plant network. I would run a line-level throughput and yield diagnostic before touching price, because a price increase into an uncompetitive cost base just moves volume to competitors. Recovering the eight point conversion gap alone gets most of the margin back inside the year."

Case 6: Airline profitability with flat ticket revenue

Prompt. A major US airline's ticket revenue is flat, but the CEO wants EBITDA margin up 4 points in two years. Where should the airline focus?

Your opening structure. Ticket revenue is stated as flat, so the tree runs on the non-ticket P&L: ancillary revenue (baggage, seat selection, cabin upgrades), the loyalty programme, cargo, and cost. Hypothesis: the margin points sit in ancillary yield and loyalty economics rather than in cost, because cost programmes at airlines are usually already running.

Interviewer question. "Baggage revenue has grown 8 percent a year, but the client's premium cabin baggage pricing sits below market. Separately, the co-brand credit card agreement has not been renegotiated in six years against a $2B loyalty programme. Which do you chase?"

The math. Loyalty first, on size. A 15 percent improvement in co-brand economics on a $2B programme is roughly $300M of incremental contribution, which against a large airline's EBITDA base is worth 1.5 to 2 margin points on its own. Ancillary pricing optimisation is smaller, roughly 1 to 1.5 points, but it lands in 12 to 18 months against 18 to 24 for a contract renegotiation.

Synthesis. "Do both, sequenced by time to impact rather than by size. Ancillary pricing gets you 1 to 1.5 points within 18 months and needs no counterparty, so start it immediately. The larger prize is the co-brand renegotiation, worth 1.5 to 2 points on a $2B programme, but it takes 18 to 24 months and depends on a partner. Together that is the 4 points, with the ancillary work de-risking the timeline. The main risk is that ancillary fee increases hit customer satisfaction scores that the loyalty partner also cares about, so the two levers need to be priced against each other, not run independently."

Case 7: Regional cinema chain, screen level economics

Prompt. A regional cinema chain with 60 locations is deciding whether to close its 12 weakest sites. Each of those sites loses money at the operating line. Should it close them?

Your opening structure. The naive answer is yes and it is wrong, which is what this case tests. Split each site's economics into contribution (revenue less variable cost) and allocated fixed cost, then ask which of the fixed costs actually disappear on closure. Hypothesis: some loss-making sites are contribution positive and are being sunk by allocated overhead and unavoidable lease liabilities, so closing them makes the chain worse.

Interviewer question. "Of the 12 sites, 7 have positive contribution margin averaging $180K a year but carry $260K each of allocated corporate overhead. The leases have an average four years remaining with no break clause. What do you recommend?"

The math. Closing a contribution-positive site removes $180K of contribution but does not remove the lease, and the $260K of allocated overhead simply redistributes across the remaining 53 sites rather than disappearing. So closing all 7 destroys 7 x $180K = $1.26M of annual contribution and saves nothing in the near term. Only the 5 contribution-negative sites are genuine closure candidates.

Synthesis. "Close 5, not 12. Seven of the loss-making sites are contribution positive and only look unprofitable because of a $260K corporate allocation that does not disappear when the site does, and with four years of unbreakable lease left, closing them destroys $1.26M of annual contribution for no cash saving. Close the five contribution-negative sites, sublet or negotiate exits on the other seven as leases approach expiry, and separately challenge the allocation methodology, because it is currently driving a decision that would have cost the company money."

This is the case type where candidates give a confident wrong answer fast. The tell is the word "allocated": if a cost is allocated rather than incurred, ask whether it leaves the business when the site does.

Case 8: Hospital system cost per patient

Prompt. A regional hospital system's operating cost per patient has risen 22 percent over four years while patient volume grew 8 percent. The CEO wants to know why and how to close the gap.

Your opening structure. Decompose cost per patient by category, then test two competing explanations before choosing a remedy: either the cost base grew, or the patients got sicker. Hypothesis: labour, which is typically 60 to 70 percent of hospital cost, is the driver, but case mix has to be ruled out first or the entire recommendation is wrong.

Interviewer question. "Labour cost per patient grew 28 percent, supplies grew 6 percent, and the case mix index is unchanged. Agency nursing has gone from 4 percent of nursing hours to 18 percent. What is happening?"

The math. With case mix flat, the patients did not get sicker, so the cost growth is real rather than explained. Agency nurses typically cost 2 to 3 times a permanent nurse per hour. Moving 14 points of nursing hours from permanent to agency at a 2.5x rate multiplies that slice of cost by 2.5, and on a labour base that dominates the P&L, that alone explains the bulk of the 28 percent labour increase.

Synthesis. "This is a nurse retention problem presenting as a cost problem. Case mix is flat, so the patients are not sicker, and supplies grew in line with inflation. Agency nursing rising from 4 percent to 18 percent of hours at roughly 2.5 times permanent cost accounts for most of the 28 percent labour increase. The fix is retention and scheduling, not a cost reduction programme: bringing agency utilisation back to 8 percent recovers most of the gap. The risk is that agency use is covering a genuine vacancy problem, so cutting it without fixing recruitment converts a cost problem into a capacity problem."

Case 9: Fast food chain losing share to a new entrant

Prompt. A fast food chain is losing share to a competitor that entered its markets 18 months ago. Traffic is down 12 percent in affected markets. What should it do?

Your opening structure. Diagnose before you respond. Branches: what the competitor is winning on (price, quality, speed, format), who is switching (which customer segments and dayparts), and what the client's response options cost. Hypothesis: the correct response depends entirely on the switching reason, so any recommendation before that diagnosis is a guess.

Interviewer question. "Survey data shows switchers cite ingredient quality and taste, not price, and 68 percent of them are aged 18 to 34. Price perception is unchanged. What now?"

The math. The wrong move here is a price cut, and the case rewards saying why quantitatively. On a chain running roughly 20 percent restaurant-level margin, a 10 percent price cut removes half the margin and, since price is not the switching reason, would not return the traffic. Product investment targeted at the items where the perception gap is widest costs a fraction of that and addresses the stated cause.

Synthesis. "Do not cut price. Switchers name quality and taste, price perception is unchanged, and a 10 percent price cut would take roughly half the restaurant-level margin while addressing a problem the customers did not report. The response is product investment concentrated on core protein items where the perception gap is largest, supported by targeted local marketing in affected markets while the reformulation lands. The risk is timing: product change takes two to three quarters to reach the guest, so I would pair it with a limited-time offer to hold traffic in the interim rather than letting the 12 percent decline compound."

Case 10: Public sector, raising voter turnout

Prompt. A city government wants to raise voter turnout in local elections from 23 percent to 35 percent within two cycles. The budget is $4M. What should it do?

Your opening structure. Public sector cases score judgment about constraints and stakeholders more heavily than commercial ones. Branches: who is not voting (which demographics and precincts), why (registration friction, access, awareness, or apathy), and which interventions have evidence of effect per dollar. Hypothesis: friction and awareness are addressable with $4M, apathy largely is not, so the plan should concentrate on the addressable share.

Interviewer question. "Of the non-voting population, 30 percent are unregistered, 25 percent are registered but cite polling access, and 45 percent cite lack of interest. What does that mean for the target?"

The math. The turnout gap is 12 points. The addressable share, unregistered plus access-constrained, is 55 percent of non-voters. Non-voters are 77 percent of the eligible population, so the addressable pool is 0.77 x 0.55 = about 42 percent of the electorate. Converting roughly 29 percent of that pool delivers the 12 points, which is aggressive but not absurd for registration and access interventions with a two-cycle horizon.

Synthesis. "The 12 point target is reachable only if you concentrate the $4M on the 55 percent of non-voters who are unregistered or access-constrained, which works out to about 42 percent of the electorate. That requires converting roughly 29 percent of that pool, which is at the optimistic end of what registration drives and polling access changes have achieved. I would spend the budget on automatic registration integration and polling location density in the lowest-turnout precincts, and explicitly not on general awareness advertising aimed at the 45 percent citing lack of interest, where the evidence of effect per dollar is weakest. The main risk is that access and interest are correlated, so I would run one precinct as a controlled pilot before committing the full budget."

The synthesis is the moment the entire case is scored, and it is the one skill in this list you can rehearse in isolation in ninety seconds a rep.

Self-diagnostic: are you practising like a McKinsey candidate?

Run this after every case rather than after every three, while the failure is still fresh enough to name.

1. Did you state a hypothesis before any data arrived? Not a restatement of the prompt, a directional claim you were willing to be wrong about. If your opening was "I would like to look at revenue and costs," you did not state one.
2. Did you name which branch you would test first, and why? A tree without a stated priority reads as a template. The priority is the thinking.
3. After each exhibit, did you volunteer the implication unprompted? Describe, quantify, implicate. If the interviewer had to ask "so what does that mean," you gave two thirds of an answer.
4. Did you say the equation out loud before computing it? In an interviewer-led case this is also your error-correction mechanism. It gives the interviewer a chance to redirect you before you spend ninety seconds on a wrong path.
5. Was your synthesis under ninety seconds, answer first, with one number in it? Time it. Most candidates who believe they are at ninety seconds are at two and a half minutes.
6. Did you name a risk to your own recommendation? McKinsey rewards the candidate who identifies the weakness in their own answer before the partner does.
7. Did you prepare a PEI story for this session too? Half the round is behavioural and it is the half that gets rehearsed last.

If you failed the same item three cases running, stop running full cases. One skill failing repeatedly is a drill problem, not a volume problem, and another 45 minute case will not fix it.

McKinsey's own official practice cases

Work the firm's published material before any student casebook, for one reason: these are the only cases where the suggested answer reflects McKinsey's own evaluation criteria rather than a student's reconstruction of it. Independent case catalogues consistently list six official McKinsey cases: Beautify, Diconsa, Electro-Light, GlobaPharm, National Education, and Talbot Trucks. They are published at mckinsey.com/careers/interviewing, each with a prompt, exhibits, and a written suggested approach.

How to sequence them:

• Start with Beautify. It is the cleanest demonstration of how a McKinsey interviewer drives a case, so it is the one to read for format rather than for content.
• Use Talbot Trucks for current themes. Sustainability and the electrification transition appear in McKinsey cases more often each year, and this is the official case that carries them.
• Use Electro-Light and GlobaPharm for the quantitative habit, because both carry the kind of setup-then-compute moment the live cases are built around.
• Use Diconsa and National Education for public and social sector reasoning, which behaves differently from a commercial P&L and shows up more than candidates expect.

Once the official set is exhausted, MBA consulting club casebooks are the sensible supplementary volume. Treat their suggested answers as one defensible path rather than as a scoring key, because they are student-written. When you run out of interviewer-led material entirely, the case library carries the same archetypes with the exhibits handed to you one at a time, which is the specific format habit the casebooks cannot rehearse.

How to use these McKinsey practice cases

Reading a worked case teaches you what good looks like. It does not teach you to produce it under time pressure with someone watching. The loop below is what converts one into the other.

The five-step loop for every case above

• ✓

  Read only the prompt, then cover everything below it. The value of a worked case is destroyed the moment you have read the answer. Treat the rest of the page as sealed.

• ✓

  Build your structure out loud in two minutes, on paper, timed. Silent structuring is easier than the real thing. Speaking it exposes the branches you cannot actually justify.

• ✓

  Answer the interviewer question before reading the exhibit values. Predicting what the data would need to show is the habit that makes you fast when it arrives.

• ✓

  Do the math on paper with units attached, then compare. Units are where most case arithmetic actually fails, and they are invisible when you do it in your head.

• ✓

  Deliver the synthesis in under 90 seconds, then compare against the written one. Synthesis is scored separately from the analysis, so it has to be rehearsed separately too.

Then wire each failure to a specific rep instead of running another full case:

• Structure came out slow or template-shaped: run free structure drills until the tree arrives hypothesis-led, and read the issue tree method.
• Arithmetic slipped or took too long: work case interview math practice and rep it on timed quick math.
• Exhibits took too long to read: work reading charts and exhibits, then rep graph reps until the describe-quantify-implicate order is automatic.
• Sizing assumptions collapsed under challenge: rebuild the driver chain with market sizing step by step.
• Synthesis rambled or led with the analysis: tighten it against the synthesis guide.
• The PEI half is unrehearsed: build the story bank from the PEI question list and rehearse the follow-ups in behavioral practice.
• Solve is on your calendar: work the Sea Wolf guide and the free module rounds before it eats a week of case time.
• Your application is still in flight: run the CV through the consulting resume grader and read the McKinsey resume guide before the round is scheduled.

Road to Offer exists for the part of this loop you cannot do alone. There are 550+ drills across structure, math, market sizing, graphs, synthesis and brainstorming, AI-graded full cases in voice or guided mode with feedback on structure, math and recommendation, free courses in Learning Mode, behavioral simulations for the PEI half, and assessment practice for the Solve modules. Signing up includes a case credit, so the first full graded case costs nothing.

For worked examples across other firm formats, see case interview examples, the BCG practice cases, and the supply chain case interview when your target firm runs operations-heavy cases.

Turn this guide into a McKinsey plan

Build my prep plan

1

Learn the format

17 steps

• Learn how to caseGuided learning · 35 min
• The Case Interview MethodLesson · 8 min
• Profitability & Turnaround: AfrihealSkill drill · 5 min
• Profitability & Turnaround: Berty BankSkill drill · 5 min

Sources and further reading (checked July 31, 2026)

• McKinsey Careers, interviewing and official practice cases: mckinsey.com/careers/interviewing
• StrategyCase, catalogue of 42 firm-provided cases including the six official McKinsey cases: strategycase.com/case-interview-examples-a-collection-from-mckinsey-and-others
• StrategyCase, McKinsey Personal Experience Interview dimensions, timing and story structure: strategycase.com/mckinsey-personal-experience-interview-the-only-post-you-need-to-read
• CaseStar, McKinsey Solve module durations, phases and scoring weights: casestar.io/guides/mckinsey-solve
• CraftingCases, reviewed collection of recorded case interview examples: craftingcases.com/case-interview-examples
• MyConsultingCoach, McKinsey Solve module rotation and invitation lengths: myconsultingcoach.com/mckinsey-problem-solving-game
• Wharton Consulting Club casebook, distributed by the Wharton Consulting Club