Most candidates treat McKinsey practice cases like a scavenger hunt: more prompts, more industries, more "examples" bookmarked in a private folder. That is the wrong scarcity. McKinsey already publishes free sample cases on its interviewing page, with logic, thought processes, and suggested answers. What fails in the room is not discovering another prompt. It is staying sharp through an interviewer-led middle — structure, one exhibit, math out loud, then a synthesis that leads with the answer — when someone else owns the question order.
This piece is a practice library post, not a process walkthrough. You will leave with a map of the firm's own cases (linked, not recopied), a way to rehearse them without memorizing the answers, one fully worked original mini-case you can run cold, and a weekly stack that matches how McKinsey actually sequences asks. For format and scoring in depth, use the McKinsey case interview guide.
What McKinsey publishes for free
On the careers site, McKinsey describes the problem-solving interview as a discussion of a typical client scenario and recommends reviewing its sample cases before you interview. Those cases are the primary source. They show how the firm wants you to organize thoughts, talk through calculations without a calculator, and move from structuring into quantitative and creativity questions. Do not treat third-party "McKinsey-style" banks as substitutes for this set.
As of September 2026, the careers interviewing hub links a small library of sample cases. Each has its own page. Use the table as a routing list — open the official page, cover the suggested answers, and work the questions yourself first.
| Official case | Rough subject (from McKinsey's own blurbs) | Page |
|---|---|---|
| Beautify | Whether training beauty consultants to use virtual channels could be profitable | Beautify |
| Diconsa | Using a rural Mexican store network for basic financial services (Gates Foundation context) | Diconsa |
| Electro-Light | Launch analysis for a lower-sugar electrolyte sports drink | Electro-Light |
| GlobaPharm | Whether a large pharmaco should acquire a biologicals start-up | GlobaPharm |
| National Education | Diagnosing and improving a fictional country's school system | National Education |
| Talbot Trucks | Whether to invest in producing and selling electric trucks | Talbot Trucks |
I am not reproducing their exhibits, numbers, or suggested answers here. That would train you to recognize McKinsey's own solutions instead of producing your own under pressure. Open the page, write your structure in two minutes, then compare.
Before live rounds, many candidates also sit Solve, McKinsey's gamified assessment. The firm says no preparation is required and provides an untimed tutorial inside the invitation. Treat Solve as a separate gate, not as a substitute for case volume. Older Problem Solving Test PDFs still float around the careers media library; they are historical selection materials, not the current Solve experience.
How interviewer-led practice actually works
McKinsey's public cases are written as sequenced questions: open with an overall approach, then handle a quantitative ask, then interpret data or brainstorm, then land a recommendation. In a live round that often feels interviewer-led — the interviewer chooses the next sub-question — even when the skills are the same ones BCG and Bain score in a candidate-led room.
The trap is going quiet because someone else is steering. Interviewer-led does not mean passive. You still own the opening tree, you still state a hypothesis, and after every number you volunteer the implication before you are asked. What you give up is choosing the path, not choosing to think.
What the sheet (in Ned's language, not McKinsey's) usually rewards across those asks:
| Ask type | What a strong answer looks like | Common leak |
|---|---|---|
| Structure | Two or three client-specific branches, priority named | A memorized template with no first test |
| Exhibit | Describe, quantify, implicate | Reading the chart and stopping |
| Math | Equation spoken first, units attached, sanity check | Silent arithmetic that hides a bad setup |
| Brainstorm | Ideas grouped under named categories | A laundry list with no spine |
| Synthesis | Answer first, one number, one risk, one next step | A summary of the analysis |
If your "practice" is reading prompts and nodding at frameworks, you are rehearsing the opening and skipping the scored middle.
Ned's rule. One official McKinsey case worked cold — structure written before you peek, math spoken aloud, synthesis timed to ninety seconds — beats five "examples" you only skim. Familiarity with the firm's answer key is not a skill the interviewer can score.
One original mini-case: HarborLine Cross-Dock
This case is invented. The numbers are internally consistent and not taken from any McKinsey sample. Cover everything below the prompt, work it for twelve to fifteen minutes, then compare.
Prompt. HarborLine is a regional less-than-truckload (LTL) carrier with $480M of annual revenue. Operating margin has fallen from 9 percent to 5 percent over two years while revenue stayed roughly flat. The CEO wants to know why margin fell and which lever is large enough to matter.
Opening structure (what I would want to hear). Revenue is flat, so say that out loud and park growth stories. Split the margin gap into price/mix versus cost. On cost, separate linehaul (fuel, driver wages, equipment), terminal/cross-dock labor, and overhead. Hypothesis: the five-point drop is concentrated in terminal labor and fuel, not in a pricing collapse — so the first ask should be a cost-stack comparison, not a market-share digression.
Interviewer question. "Here is HarborLine's cost structure as a share of revenue for the two years. What do you take from it, and how large is the dollar problem?"
Exhibit (hand this as a table).
| Cost line (% of revenue) | Year −2 | This year |
|---|---|---|
| Linehaul fuel and equipment | 28% | 29% |
| Linehaul driver wages | 22% | 22% |
| Terminal / cross-dock labor | 18% | 22% |
| Overhead and G&A | 13% | 12% |
| Operating margin | 9% | 5% |
| Implied other / residual | 10% | 10% |
Math (spoken setup first). Revenue base is $480M. Margin fell 4 percentage points (9% to 5%), so the dollar gap is 0.04 × $480M = $19.2M, call it $19M. Terminal labor rose 4 points (18% to 22%): 0.04 × $480M = $19.2M. Fuel and equipment rose 1 point: about $4.8M. Overhead improved by 1 point, saving about $4.8M. Net: terminal labor alone explains the entire margin decline; the fuel rise and overhead improvement roughly cancel.
Sanity check. If terminal labor is now 22% of $480M, that is about $106M of annual terminal cost. A 10 percent productivity gain on that base is roughly $10.6M — more than half the gap — which tells you the lever is operationally plausible, not a rounding error.
Synthesis (sixty to ninety seconds). "HarborLine has a $19M terminal-labor problem, not a pricing problem. Revenue is flat and driver wages did not move as a share of sales; the whole four-point margin loss sits in cross-dock labor. I would start with a shift-level productivity audit on the five largest terminals and a volume forecast that freezes overtime until utilization recovers. A 10 percent labor productivity gain on today's terminal base recovers roughly $11M of the $19M. Main risk: some of the labor growth is deliberate service investment for a large shipper — check whether the overtime spike is concentrated in a handful of accounts before you cut headcount."
If your structure ignored the flat-revenue clue, or your synthesis never named $19M, that is the drill target — not "find another logistics prompt."
A practice stack that matches the room
Volume without format is noise. Use three layers and stop when each layer is honest.
Layer 1 — Official cases (primary source). Work all six pages linked above at least once cold. McKinsey's interviewing overview is explicit that the samples include suggested answers so you can learn the types of questions they ask. Compare only after you have written your own structure and run the math out loud.
Layer 2 — One original interviewer-led rep per day. HarborLine above, then rotate industries. Prefer cases that force an exhibit read and a synthesis with a number. CoachNed's library has 64 original cases; the voice CopperCup interview is useful when you need the silence and follow-ups of a live round (~18 minutes, seven-score debrief).
Layer 3 — Skill drills for the leaks the cases expose. Structure trees that die on priority → /drills/structure. Sloppy setups → /drills/math or free /drills/quick-math. Flat exhibit reads → /drills/charts. Soft closes → /drills/synthesis.
Solve sits beside this stack, not inside it. When the invitation arrives, use McKinsey's own Solve page for rules and integrity expectations, and treat any simulator as unofficial. CoachNed has free-to-start modules at /mckinsey-red-rock, /mckinsey-sea-wolf, and /mckinsey-sfl if you want the interface muscle memory without confusing it for the firm's assessment.
The PEI is the other half of most McKinsey interview slots. Cases do not replace it. Build and pressure-test stories separately via the PEI guide and /behavioral.
What folklore gets wrong about "McKinsey cases"
Three claims circulate hard and deserve a clean label.
"You need dozens of unique McKinsey prompts." Folklore. The firm publishes a short set and asks you to practice the conversation those cases model. After you can run that pattern cold, more prompts have diminishing returns unless they expose a new leak (charts, synthesis, public-sector framing).
"Interviewer-led means wait for instructions." Folklore dressed as format advice. McKinsey's own case pages repeatedly tell you to organize thoughts before answering, develop an overall approach, and talk through calculations so the interviewer can follow. Passive candidates look unfinished even when the math is correct.
"Solve practice replaces case practice." Folklore. Solve is a digital gate with its own rules; McKinsey states results are considered with the rest of the application. Cases still decide the live rounds.
When someone on Reddit or Glassdoor claims a secret case bank or a fixed pass score for Solve, treat it as rumor until the firm or a durable primary source says otherwise.
Practice this week
Concrete assignment, not a vibe:
- Open Beautify and Electro-Light. Cover the suggested answers. Write a structure for each in two minutes, then complete every quantitative ask out loud.
- Run HarborLine cold with a timer. Record whether your synthesis named the $19M and a risk.
- Do one scored structure drill and one math drill on CoachNed: start at /start (three typed turns, no account required) or jump to /drills/structure and /drills/math. Everything on CoachNed is open for seven days with no card; then $120 for a recruiting season or $49 a month.
If you only have ninety minutes this week, skip layer 2 variety and do steps 1 and 2 twice. Repetition on the scored middle beats another industry bookmark.
Frequently asked questions
Where can I find official McKinsey practice cases?
On McKinsey's careers interviewing page, which links sample cases such as Beautify, Diconsa, Electro-Light, GlobaPharm, National Education, and Talbot Trucks. Those pages include suggested answers; use them after you attempt the questions yourself.
Are McKinsey case interviews always interviewer-led?
McKinsey's published samples are sequenced questions controlled by the case design, which feels interviewer-led in most first-round rooms. Partners sometimes run cases more conversationally. Prepare to open a structure and stay proactive either way; do not wait to be told what to think next.
How many McKinsey practice cases should I do?
Enough to make the question types automatic — typically the official set once cold, plus a small number of original cases each week until your exhibit reads and syntheses hold a number under time pressure. More prompts after that only help if they target a named weakness.
Does McKinsey Solve replace case practice?
No. Solve is a separate digital assessment. McKinsey says no preparation is required and that results are considered with the rest of your application. Live problem-solving interviews still require case reps.
Can I use competitor "McKinsey case" PDFs instead of the official ones?
You can practice skills on any well-built case, but the firm's own samples are the primary signal for question shape and expectations. Prefer those first, then add original cases — do not treat recaps of McKinsey's pages as the source of truth.
Sources
- Interviewing at McKinsey — official description of the personal experience and problem-solving interviews, plus links to sample cases. Checked 2026-09-24.
- Beautify practice case — sample case on virtual beauty-advisor economics. Checked 2026-09-24.
- Diconsa practice case — sample case on rural financial services via a store network. Checked 2026-09-24.
- Electro-Light practice case — sample product-launch case for an electrolyte drink. Checked 2026-09-24.
- GlobaPharm practice case — sample acquisition case in biologicals. Checked 2026-09-24.
- National Education practice case — sample public-sector education diagnostic. Checked 2026-09-24.
- Talbot Trucks practice case — sample case on electric truck investment. Checked 2026-09-24.
- Solve, McKinsey's assessment game — official Solve overview, integrity rules, and support contacts. Checked 2026-09-24.
CoachNed is not affiliated with McKinsey & Company.