Case Interview Examples: A Full Transcript, Scored Line by Line

Candidate: "Three quick questions. Is the margin decline broad-based across clubs, or concentrated in a region or a vintage of clubs? Has the business model changed in the period, new membership tiers or formats, or is it the same product? And how does the CEO define success: is 18% the target, and on what horizon?"

Interviewer: "Broad-based. I'll show you data on the model in a moment. And the CEO wants a credible path back to historical margin within two years."

Score: 4/5, judgment. Three questions, and each one changes the work. Concentration separates a sick company from a few sick clubs. The business-model question anticipates the real answer, tier mix, before any data appears. The success question converts a complaint into a target with a deadline. Interviewers read clarifiers as a preview of how you would run a first client meeting, and the average candidate wastes them, either asking for "any other information you have" or firing six questions to buy thinking time. What kept this from a 5: no hypothesis attached. "Margin down while revenue grows usually means price, mix, or cost inflation, and my questions aim there" would have shown the interviewer where the questions came from.

Candidate: "Give me ninety seconds to lay this out." [writes] "Margin fell seven points while revenue grew, so something inside the P&L deteriorated. I'd look in three places. First, revenue quality: member count, price per member, and the mix across membership tiers, because revenue can grow while its composition gets worse. Second, the cost base: rent, staff, and equipment at club level, and whether costs stepped up faster than membership. Third, the portfolio: how much of the chain is new clubs, because expansion can grow revenue and dilute margin at the same time. My working hypothesis is that the revenue growth is hiding the problem, not contradicting it. I'd start with revenue quality: it's one data pull, and a broad-based decline fits a company-wide mix or price issue better than a cost blowout in any one place."

Score: 4/5, structuring. Built for this client, not recited: tiers, club vintages, and ramp are gym economics, and a generic profitability tree contains none of them. The branches cover the ground without overlapping, there is a stated hypothesis, and the answer ends with a committed starting point and a reason. That last sentence is what separates a structure from a list. The gap that kept it at 4: nothing external. Competitors could produce this exact P&L shape, budget chains pulling price-sensitive members at one end and boutique studios pulling premium members at the other, and the structure has no branch where that would surface. The interviewer probes exactly there next.

Candidate: "Missing... I covered price and mix inside revenue quality, so, discounting?"

Interviewer: [waits]

Candidate: "Let me change the axis instead of renaming branches. Everything I listed is internal. Externally: competitors could be repricing the market underneath us, low-cost chains at one end and boutique studios at the other, pulling members from both sides. And one-off structural items: leases renewing at market rates, an equipment refresh cycle, anything that steps the cost base up without anyone deciding it this year."

Score: 3/5, structuring. The first response repackaged an existing branch. "Discounting" was already inside price and mix, and the interviewer's silence said so. The recovery is textbook: name the axis you searched, then search the one you skipped, and the two categories that followed were genuinely new. But the bar was met, not cleared, because the interviewer had to wait through a false start. The transferable rule: when probed for what's missing, never answer from inside your existing structure. Say the axis out loud, internal to external, ongoing to one-off, economics to strategy, and generate from the other side. This probe is standard in every format, and it is asked when your structure was fine, to test breadth under mild pressure.

Candidate: "So membership grew from 185,000 to 212,000, up about 15%. Base went from 92,000 to 138,000, and Peak declined from 93,000 to 74,000..."

Interviewer: [waits]

Candidate: "Here's the takeaway. All of the growth came from the discounted tier, and it isn't just growth, it's migration: Peak lost 19,000 members while the company added 27,000. So IronPeak isn't only attracting cheaper members, it's likely converting expensive ones into cheap ones. Revenue per member has to be falling, which is how revenue grows 3% while margin collapses. I'd want to know how much of the Base growth is downgrades versus new joiners, because those need different fixes."

Score: 3/5, communication. The content arrived, and the closing distinction, downgrades versus new joiners, is the sharpest sentence so far. But the first fifteen seconds narrated the chart to the person who drew it, and the interviewer's silence was the second rescue of the interview. The fix is mechanical, not intellectual. Take twenty silent seconds with any exhibit: title, axes, units, footnote. Then speak in this order: headline, evidence, implication. This candidate said the same content in discovery order, evidence first, headline late, and paid two points for it. Same material, reversed order, and this exchange is the best in the interview.

Candidate: "Setup first: average revenue per member, then and now, applied to today's base. Two years ago the split was roughly half and half, so the average sits at the midpoint of $30 and $60: $45 a month. Today Base is 138 of 212, call it 65%. So 0.65 times 30 is 19.50, plus 0.35 times 60 is 21, gives $40.50. That's $4.50 less per member per month, or $54 a year. Times 212,000 members: about $11.4 million of revenue a year, gone from mix alone. On $103 million of revenue, that's eleven points."

Score: 4/5, quantitative reasoning. The method was announced before any arithmetic moved, the rounding was deliberate and labeled, and the units stayed spoken the whole way: per member per month, then annualized, then scaled to the company. That order is the entire skill. When the setup is agreed first, an arithmetic slip costs little, because the interviewer can see the machine works. Notice also the final sentence: the number restated against company revenue, which is what makes it usable. What kept it at 4: the number arrived and stopped. Eleven points of revenue is context, not consequence, and the connection to the margin question came only when the interviewer asked for it. That question is next.

Candidate: "I'll build the bridge. Membership grew by 27,000; at the old $45 average, that's about $14.6 million of new revenue. Mix takes back $11.4 million. Net is roughly the $3 million of growth you gave me, so revenue reconciles. The missing piece is cost. Can I get operating cost, then and now?"

Interviewer: "Costs rose about $9.8 million. The chain went from 80 clubs to 88."

Candidate: "Then it closes. Eight new clubs at roughly a million of operating cost each is $8 million, plus wage and rent inflation on the existing 80, call it $1.8 million. So the bridge reads: volume added $14.6 million, mix gave back $11.4, cost grew $9.8. That nets to minus $6.6 million against the $6.7 you stated; the residual is rounding. And the read on the business: IronPeak bought its growth. It opened eight clubs and filled them with $30 members, adding cost at full price and revenue at half price. The growth is real. The unit economics underneath it flipped."

Score: 5/5, judgment. This is the exchange the interviewer retells in the debrief. The bridge is built unprompted, ties to the stated decline, and the residual is named as rounding instead of quietly absorbed, which signals honest arithmetic. The cost data is requested, not waited for. And the last three sentences turn a reconciliation into the thesis of the case: cost at full price, revenue at half price. Note what a 5 is not: a harder calculation. Every number here is arithmetic a tired person can do. A 5 is an ordinary calculation pushed one sentence further into meaning than anyone asked for.

Candidate: "Twenty seconds to organize." [pause] "Three buckets: fix the mix, fix the price architecture, fix the cost of growth. Mix: rebuild the upgrade path into Peak, gate the amenities members use most, classes, sauna, recovery, behind it; sell paid day-upgrades that sample it; push corporate accounts, which skew premium. Architecture: raise Base from $30, it's carrying the whole growth story; add a mid tier near $45 so downgraders have somewhere to land besides half price; trade perks, not discounts, for 12-month commitments. Cost of growth: pause openings in markets where new clubs cannibalize existing ones, and staff new clubs to their ramp curve instead of the mature-club template. If I pick one first: the Base price. 138,000 members at $30 is the biggest lever on the page, and it requires no construction."

Score: 4/5, structuring. Buckets announced before ideas, nine ideas across them, and each priced in gym reality: amenity gating, ramp-curve staffing, commitment perks. Generic brainstorms say "improve marketing" and "reduce costs"; this one could only belong to a gym chain, which is what the score rewards. The forced choice at the end, with a reason, is the habit interviewers remember, because most candidates present a buffet and wait. What kept it at 4: the list never left the membership P&L. Personal training attach, guest fees, and retail are a real margin lever in this business, and an entire bucket went unmentioned. Breadth probes exist to find that missing bucket, and this time the interviewer moved on instead.

Candidate: "Breakeven first, then a view. Base revenue today: 138,000 times $30 is $4.14 million a month. At $33, revenue holds if membership stays above 4.14 divided by 33, about 125,500. So they can lose roughly 12,500 members, 9% of the tier, before the increase loses money. And the real threshold is higher than that, because a lost member takes no cost out with them: the clubs stay open and staffed either way, so at the margin this is contribution, close to the full fee. My view: a 10% price move producing 9% attrition would be extreme for a $30 subscription; I'd expect low single digits. But that's an assumption to test, not a fact, so I'd stage it: two test markets, one renewal cycle, measure, then national."

Score: 4/5, quantitative reasoning. The breakeven is computed before any opinion is offered, which is the right order: the threshold turns a debate about elasticity into a comparison against a number. The fixed-cost logic is made explicit rather than assumed, and the elasticity view is labeled as an assumption with a test attached. That is exactly how a pricing recommendation should reach a partner. What kept it at 4: the prize was never sized. At the churn the candidate expects, call it 3%, the increase nets about $3.3 million a year, half the EBITDA decline recovered by repricing one tier. That sentence was available and unsaid, and sizing the upside is what turns analysis into a recommendation someone can fund.

Candidate: "Four, in order of concern. First, churn concentration. Everyone who arrived in Base these past two years, new joiner or downgrader, sorted themselves there on price. Blended churn assumptions will understate that cohort, so model the newest members separately. Second, the ladder effect: $33 Base against $60 Peak makes Peak look worse, not better. Raise Base alone and you accelerate the migration you're trying to reverse, so the mid tier and upgrade perks ship in the same move. Third, the competitive edges: a higher Base widens the gap to the budget chains below and narrows the discount to boutiques above, so both exits get more attractive. Fourth, mechanics: annual contracts reprice at renewal, so the P&L effect phases in over a year while the local press lands on day one, and notice terms need legal review state by state."

Score: 4/5, judgment. Ordered by concern, each risk carrying a mechanism rather than a label, and the second one is the kind of second-order effect most candidates never reach: repricing one tier reprices the whole ladder. The cohort point connects back to the exhibit, which is what judgment looks like, earlier findings constraining later recommendations. What kept it at 4: the risks are named, and none is owned. A 5 ends with which risk would change the recommendation and what evidence would settle it: "the only risk that kills the move is cohort churn above 9%, and the two test markets answer that in one renewal cycle." One sentence, and the risk list becomes a decision rule.

Candidate: "IronPeak doesn't have a growth problem or a cost problem. It has a mix problem, and mix is a choice. Three facts. All net growth came from the $30 tier while the $60 tier shrank by 19,000 members, so revenue per member fell 10%. That mix shift costs $11.4 million a year, more than the entire $6.7 million EBITDA decline; volume growth masked it. And the eight new clubs added cost at full rate while filling with discounted members, so expansion is currently diluting margin. My recommendation: reprice the ladder as one move, Base to $33, a mid tier near $45, real perks gated into Peak, staged through two test markets this quarter and national at the next renewal cycle, and hold new openings until club-level economics clear a margin bar. The risk that matters is churn in the newest, most price-sensitive cohorts, and the test markets answer it in one cycle. This path returns roughly half the lost margin within eighteen months. The rest requires Peak to grow again, and that's the next piece of work."

Score: 5/5, communication. Answer in the first sentence, three supports each carrying a number from the case, one risk with its test, one boundary. It lands in under ninety seconds because nothing in it is journey: no "we looked at," no chronology, no suspense. Two details are worth stealing. The numbers survived into the summary, which is what makes it sound like findings instead of opinion. And the last sentence tells the CEO what the plan does not fix, unprompted. Nothing builds credibility with a senior audience faster than volunteering the limit of your own recommendation before anyone probes for it.

Right after you leave the room, the interviewer writes per-question notes, grades a handful of dimensions, and brings both to a debrief where scores are calibrated against the other interviewers' cases. You never see the card. Here is this one:

• Structuring: 4. The opening structure and the brainstorm were tailored and prioritized (Q2, Q7). The probe recovery cost a point: one repackaged branch before the axis change (Q3).
• Quantitative reasoning: 4.5. Setup-first arithmetic with spoken units throughout (Q5, Q8), and an unprompted bridge that reconciled to the stated decline (Q6).
• Judgment: 4.5. Clarifiers that previewed the answer (Q1), the bought-growth thesis (Q6), cohort-level churn thinking (Q8, Q9).
• Communication: 3.5. Two silences, both waiting for a headline that was late (Q3, Q4). The synthesis shows the ceiling (Q10); the exhibit read shows the habit.
• Verdict: advance. Development note: leads with evidence, not insight, under pressure.

One pattern is worth noticing: the content never failed. Both weak moments were ordering problems, insight delivered late, and both fives were ordering excellence, insight delivered first. Most candidates prepare content and leave order to chance.

Not knowledge. Reread the two 3s and the two 5s and compare raw material: the exhibit read (a 3) contained the sharpest distinction in the interview, and the margin bridge (a 5) contained arithmetic a tired person can do. The scores diverge on packaging, and the pattern is consistent enough to state as rules.

Between 3 and 4, the question is: did the interviewer have to wait, prompt, or reorganize the answer for you? The repackaged branch needed a silence to fix. The chart narration needed a second one. Both answers got there; both needed help getting there, and per-question notes record exactly that.

Between 4 and 5, the question is: did you hand the interviewer a sentence they can repeat in the debrief? "Cost at full price, revenue at half price." "Mix is a choice." A 5 compresses the analysis into something portable. That is not a stylistic bonus; consultants are paid to make findings travel, and the interviewer is checking whether yours do.

Neither rule requires being smarter. Insight first, then evidence. Every number pushed one sentence into meaning. Announce the axis, announce the buckets, flag a choice. These are habits, and habits are trainable in a way brilliance is not.

The same case produces 2s with answers that are never wrong, only inadequate. Four moments, at a 2:

"I'd use the profitability framework. Profit is revenue minus costs, so I'd start by looking at revenue, and then I'd look at costs."

A structure with no client in it. Nothing about tiers, clubs, or members; swap in an airline and every word survives. Interviewers hear this several times a week.

"So Base went up a lot, and Peak went down... interesting... and the total is also up..."

An exhibit read with no destination. The narration continues until the interviewer rescues it, and the per-question note says so.

[ninety seconds of silence, then] "It's about $11 million."

The number is right and barely scores. The method was invisible, so it cannot be trusted or corrected, and the so-what never arrives.

"To summarize: first we looked at membership, then we quantified the mix impact, then we discussed pricing..."

A chronological recap. The CEO learns the agenda, not the answer.

None of these candidates lacked frameworks or arithmetic. Each failure is the packaging half of a skill they possessed, which is why prep that only accumulates knowledge plateaus at 3.

Directly. What changes across firms is who drives, not what is scored.

At McKinsey, the case is interviewer-led: these ten exchanges arrive as scripted questions, the interviewer controls the sequence, and each answer is graded on its own, so you cannot steer around a weak spot. The McKinsey case interview guide covers all six question types in that format. At BCG and Bain, the case is mostly candidate-led: after the structure, you propose each next step, so the transitions in this transcript, into the exhibit, into the math, into the recommendation, become moves you initiate. The BCG case interview guide shows what driving looks like, including the online case that now sits in front of the live rounds.

The four scoring dimensions do not change, and neither does the packaging bar from the previous section. One more constant: every one of these firms scores a behavioral interview separately, and it can end a candidacy on its own. McKinsey's version, the PEI, is the most structured; the PEI guide covers it at this same depth. A case-only prep plan loses interviews the case half won.