BCG Case Interview Questions and Answers: This Time You Drive

Resume screen first. Then, for most offices, the online case: Casey arrives by email, you complete it at home within a window, and it decides or heavily shapes whether you see a human. Some offices run it as a hard gate before round one; others send it in the same cycle as live interviews. A few also run a separate aptitude-style screen. The invitation email lists your exact sequence, and it beats any guide, including this one.

Then two live rounds, typically two to three interviews each, 45 to 60 minutes per interview. Every interview has the same anatomy: 10 to 15 minutes of fit and motivation questions, 25 to 35 minutes of case, a few minutes for your questions. Round one is usually consultants and project leaders. The final round is senior, principals and partners, and the cases there run more conversational and more opinionated: expect a senior partner to poke at your judgment, not your arithmetic. Some offices add a written case in the final round, a prepared-analysis-then-present format, and say so in advance.

Decisions come fast, often within days. The whole pipeline can run application to offer in three to six weeks, which means the preparation window is mostly before you apply, not after Casey lands.

Candidate-led means the case has no script you can ride. After the prompt, you produce the structure, choose where to dig, request the specific data you want, interpret what comes back, and steer to the next analysis until you judge the case answered. The interviewer plays two roles: a data source who responds to specific requests, and a skeptical client who pushes back on weak reasoning. What the interviewer will not do is rescue a drifting case. If you stop driving, the room goes quiet, and the quiet is the assessment.

The scoring covers the same fundamentals every top firm tests, structuring, quantitative accuracy, business judgment, communication, with one addition that changes your preparation: ownership. Did the case move because you moved it? Interviewers track who initiated each transition. A candidate who produces excellent analysis when prompted, but waits to be prompted, fails the exact test BCG built this format to run.

There is also a style difference worth knowing. BCG rewards hypothesis-led play: naming early what you suspect and testing it, rather than surveying every branch evenly. An opening like "my hypothesis is the share loss is price-driven, and I'd test that first" is not presumptuous here. It is the expected shape of ownership, provided you update it honestly when the data disagrees.

Casey, officially the Online Case Experience and delivered on the HireQuotient platform, is a chatbot that runs you through one business case in a messaging-style interface. Roughly 25 to 35 minutes of questions, 8 to 10 of them, some with sub-parts, then a 60-second video recommendation recorded on your webcam. You take it at home, alone, on a laptop, inside a completion window the email specifies.

Most BCG offices worldwide now use it, across the US, Europe, Asia, and Latin America, either as a hard screen before round one or in the same cycle as live interviews. BCG publishes no pass rate, and no retake rule; the working assumptions are that a meaningful share of applicants exit here and that you get one sitting per application. Prepare accordingly, and schedule the sitting inside the window after your preparation, never the night the email arrives.

One orientation fact changes how the whole thing feels: it is a single case, not a question bank. The chatbot's questions build on one client situation, so context you absorb early keeps paying through the sitting. Read the setup like you will be examined on it, because you will be, one question at a time, with no way back.

The interface is a chat thread. Casey presents the client situation, then questions arrive one at a time in a fixed sequence: multiple choice, numeric entry, and short free-text, with exhibits and datasets attached where relevant. Three rules define the experience. You cannot return to a previous question. You must answer to advance. And Casey gives nothing away: no hints, no reaction, no "interesting, tell me more." The feedback loop you train on with humans does not exist here.

Expect an early question about data: given the client's problem and a set of available data sources, which would you pull first? That is a structuring question wearing a data costume; it is asking whether you can name the analysis before the number, so answer it from a mental framework of the problem, not from which dataset looks richest.

The mix that follows leans quantitative: sizing, percentage and growth arithmetic, profitability math, chart reading. Free-text answers should be written the way you would speak a case answer: conclusion first, then the two reasons, in two or three sentences. Nobody scores elegance; the box rewards a structured answer delivered fast.

A calculator is usually permitted, and your invitation confirms it. Keep it, paper, and a pen within reach before you start, because the clock does not pause while you dig for one.

Live-case math, minus the partial credit. The recurring set: market sizing from stated assumptions, percentage change and growth rates, margin and profitability arithmetic, break-evens, and reading quantities off charts to combine them. Difficulty sits at confident-arithmetic level; the pressure comes from the one-way clock and from numeric-entry boxes that are simply right or wrong. In a live case, a slip narrated well costs little. In Casey, the box does not hear your reasoning.

So the bar is accuracy first, speed second. Set up each calculation before touching the calculator: write the formula fragment on paper, plug numbers, run it once. Sanity-check magnitude before submitting, a market size that implies every American buys nine units should die on your desk, not in the answer box. For multiple choice with numeric options, estimate before computing; options are usually spread wide enough that a rounded estimate identifies the answer and the full calculation just confirms it.

Sizing questions deserve one extra habit: state your assumption chain in the free-text space when the format allows it. Where Casey asks for the number alone, the number decides. Where it leaves room for reasoning, a visible chain of round assumptions is the difference between a wrong-but-rigorous answer and a wrong answer.

Ten minutes of daily drill for the week before, percentages, growth, margins, done on paper against a timer, covers the entire mathematical surface of the assessment.

The sitting ends with a webcam recording: your recommendation to the client, in 60 seconds, with a hard cutoff. It is the only moment a human at BCG will watch from your online case, which makes it disproportionately valuable, and it is the moment candidates prepare least. Treat it as one shot; retakes are not part of the design.

Budget the minute before you speak: recommendation in the first ten seconds, your strongest number from the case in the next twenty, a supporting point in the next twenty, one risk with the next step in the final ten. That shape survives the cutoff even if you run long, because the decision and the evidence are already delivered by second thirty. The inverted version, context first, recommendation last, gets guillotined mid-sentence with nothing on record.

Say the recommendation in client language: "I recommend entering the market through partnership rather than acquisition. Three things support that." Numbers from the case you just worked belong in the middle, one or two, spoken plainly. Do not recap the case, the viewer knows the case. Do not hedge into "it depends." Look at the camera, not the screen preview of yourself, and stop talking when you are done; dead air beats a trailing ramble.

Rehearse the mechanic itself: five one-take, 60-second recommendations into your phone this week, watched back. The discomfort of watching is the preparation working.

"Take it away" is the whole prompt. Nobody will ask you a first question; the first question is yours. Take your minute, build the structure, and open with a hypothesis: "My initial hypothesis is that we're losing share on price to low-cost carriers, and I want to test that before looking anywhere else."

The structure behind that opening still has to cover the ground: the competitive side, who is taking the share, on what routes, with what fares and schedule; the customer side, which segments are leaving, business or leisure, and what they say they are buying instead; the client side, cost position, network economics, product. Airline-specific texture is what separates a prepared candidate: load factors, route-level profitability, the schedule's value to business travelers, low-cost carriers' cost-per-seat advantage.

Then do the thing McKinsey never requires: pick your branch and request data in the same breath. "I'd start competitive. Can I see our fares against the main low-cost carrier on the top ten routes, alongside share change per route?" A structure at BCG is not a deliverable. It is a route map, and you prove it by driving onto it immediately, without waiting to be invited.

The drive sentence. Every time data lands, close the loop in one breath: what the finding is, what it means for your hypothesis, and what you want next. "Our fares are 30% above the low-cost carrier on eight of ten routes, so the price gap is real, which supports the hypothesis. Now I want to know if it explains the share loss: can I see share change against fare gap by route?" Finding, implication, next request. That sentence, repeated, is a candidate-led case.

Each element earns marks on its own. The finding shows you read the data. The implication shows judgment, and it must reference the hypothesis, which either strengthens, weakens, or dies in public. The next request shows ownership, and it is the piece candidates drop: they interpret well, then look up expectantly, and the case stalls in the silence they were supposed to fill.

Two disciplines keep the loop honest. Kill your hypothesis out loud when the data kills it: "the gap is only on two routes, so price isn't the story; I want to look at schedule next." Interviewers rate that update higher than a lucky first guess. And keep a visible thread: if an analysis is a detour, say why it earns the time. Drift is the candidate-led failure mode, and the interviewer will let you drift all the way to the end of the slot.

First, notice what is happening: BCG interviewers play skeptical clients, and the pushback is usually a test of your reasoning under challenge, not a correction. The response that fails fastest is instant surrender, "you're right, let me start over," because it reveals the hypothesis was a guess wearing a suit. The response that fails second-fastest is digging in on conviction alone.

The pass is evidence-weighted: restate what supports the hypothesis, hear what the interviewer is pointing at, and update by exactly the amount the evidence demands. "The fare gap on eight routes still says price matters. But you're right that share loss is worst on the two routes where fares match, so price can't be the whole story. Let me hold price as a partial driver and test schedule on those two." You have defended with data, conceded with data, and produced the next move, all in one turn.

Sometimes the pushback carries information: interviewers steer when a case has a scripted insight you are walking past. If the challenge names something specific, "have you thought about what corporate contracts do here?", that is a door, not a wall. Walk through it and say what you expect to find behind it. Treating hints as threats is how strong candidates lose winnable cases.

Set up aloud before computing: the comparison is charter revenue against the contribution of flying the route ourselves, so we need passengers, revenue, and variable cost per flight. Passengers: 150 seats at 70% is 105. Revenue: 105 times $80 is $8,400. Variable cost: 105 times $60 is $6,300. Contribution: $2,100 per flight. The charter pays $2,600, guaranteed, with most variable cost gone. Pure economics says take it, roughly $500 more per flight, about 24% better.

Then earn the question mark. This is a share-loss case, and parking capacity with a charter operator is how an airline retreats. So the answer with judgment in it: the charter wins per flight today, but accept it only where routes are structurally unprofitable, not merely under pressure, because seats you charter out are share you hand to the carrier already beating you, plus lost feed into the connecting network. Recommend the charter for the two routes where fares already match the low-cost carrier and losses look structural, and reinvest the freed aircraft where the fare gap, not the route, is the problem.

That second paragraph is the score. The interviewer gave clean numbers precisely so the arithmetic would not be the test. What is being tested is whether a profitable-looking number can still be a bad strategy, and whether you say so unprompted.

In an interviewer-led case, an exhibit is a question: read it and answer. In a candidate-led case, an exhibit is a resource: it appeared because you asked for something, or because the interviewer decided you need it, and either way what happens next is your call. That changes the standard. Reading it correctly is the floor; the score lives in what you do because of it.

The hygiene still comes first, thirty seconds of it: title, axes, units, footnotes, the place exhibits hide their traps. Then the insight, stated as a headline rather than a tour: "schedule frequency, not fare, tracks the share loss; the routes we lose worst are where the low-cost carrier flies four times daily to our two." Quantify the one comparison that makes it concrete.

Then the candidate-led move: fold the insight into the thread and drive. Update the hypothesis it touches, out loud. Name the analysis it makes necessary. Request the data. "That shifts my hypothesis from price to schedule on those routes, which changes the fix, frequency costs aircraft, not fare cuts. I want to see aircraft utilization next, to know if we can add frequency without new planes." An exhibit read brilliantly and then abandoned scores like a paragraph of good analysis in someone else's case.

BCG's creativity questions want structured breadth with a commercial spine: many ideas, organized, and each one priced against reality within a second of saying it. Announce the buckets first, twenty seconds of thought buys them: "I'll think about flying them for revenue, monetizing them without flying, and reducing the cost of owning them."

Fly for revenue: charter and wet-lease to other operators, off-peak leisure routes with fare stimulation, cargo conversions on overnight windows, ad-hoc sports and events flying. Monetize without flying: dry-lease surplus aircraft out, sale-and-leaseback on the oldest airframes, use the idle window for heavy maintenance so peak availability rises. Cut the cost of owning: renegotiate or return leases as they lapse, consolidate the fleet onto fewer types to cut spares and training.

Then do the thing the format is testing underneath: pick with a reason tied to the case. "Given the share fight, I'd rank wet-leasing to non-competing carriers first, it earns revenue without feeding the rival taking our share, and it keeps aircraft recoverable when we add frequency on the routes we're defending." A creative list is table stakes. A choice that remembers what problem the case is solving is what gets written in the interviewer's notes.

In a candidate-led case, the synthesis is yours to initiate, and initiating it is scored. When the analysis has answered the question, or the slot is visibly closing, take the floor: "Let me pull this together as a recommendation." That sentence, unprompted, signals ownership more cleanly than anything else in the interview, and its absence forces the interviewer to extract a summary from you, which is its own data point.

The shape is answer-first and numbered: the recommendation, two or three supports each carrying a number from the case, one risk, one next step. "We should defend the eight price-gap routes with a targeted fare response and fix frequency on the two schedule-loss routes. Three reasons: the fare gap of 30% explains share loss on eight of ten routes; the two exceptions track frequency, where they fly four daily to our two; and idle off-peak aircraft mean we can add frequency without capital. The risk is a price war we fund longer than they can; the next step is a route-level war-game of their response."

Deliver it standing on your evidence, committed. Senior interviewers push on it, "would you bet the client's money on that?", and the tested answer is a yes with the load-bearing assumption named, not a retreat into "we'd need more analysis." You drove the case; own where it landed.

Every BCG interview spends 10 to 15 minutes on fit, usually up front, and it is conversational rather than scripted: why consulting, why BCG, and one or two deep-dives into stories of leadership, impact, or steering through ambiguity. Because it is freestyle, any interviewer can take any story anywhere, and the follow-ups go deep enough that invented or borrowed stories fail on the third "and what did you say then?"

Prepare it like the interviews are connected, because they are: interviewers within a round compare notes. Build three or four stories with real depth, a persuasion under resistance, an initiative you started, a team you led through friction, and know each to the level of specific conversations, decisions you weighed, and what you would now do differently. The depth standard is the same one McKinsey's PEI enforces, and our McKinsey PEI guide covers story construction at that depth; the same stories, tuned, serve both firms.

The BCG-specific part is the why. "Why BCG" gets asked plainly and deserves a plain answer with evidence: a practice area you have a genuine link to, people you have spoken with, work of theirs you have read. Consulting-generic answers, prestige, learning, smart people, read as interchangeable because they are. One specific, checkable reason beats three impressive vague ones.