Consulting Case Interview Prep: McKinsey, BCG, Bain (2026 Guide)

What a 2026 MBB consulting interview actually looks like

The McKinsey, BCG, and Bain interview loops have converged enough that a candidate prepping for one is, with caveats, prepping for all three. The shape is roughly the same:

1. Networking and recruiter touchpoints. Coffee chats, on-campus or virtual events, a recruiter call. None of these are the interview, but all of them feed the screen.
2. Online assessment. McKinsey's Solve (the redesigned PST replacement, ecosystem game), BCG's Casey chatbot screen plus the Pymetrics behavioural game, Bain's SOVA situational assessment. These are pass/fail filters before the human round.
3. First round. Typically two cases plus a short behavioural in a single half-day. Two interviewers, each running one case.
4. Final round. Two to three more cases, this time with partners or principals, plus a longer behavioural conversation. McKinsey's PEI runs as a structured, scored module here.
5. Decision. Most firms turn around an offer within a week of the final round. A quiet rejection often comes faster.

The number of cases hasn't changed in a decade. What has changed is the screen at the top and the amount of behavioural weight at the bottom — both have increased, and both are now LLM-assisted on the firm's side. McKinsey's recruiters explicitly note that they read PEI transcripts looking for AI-generated patterns. The arms race is real and quietly shifting away from "memorised frameworks" toward "specific personal experience."

How the firms differ — McKinsey, BCG, and Bain

The cases look similar. The texture is different.

McKinsey runs the most interviewer-led cases. The interviewer hands you a structured prompt, asks you to lay out an issue tree, and then drives the case down specific branches in a specific order. Math is exact and the interviewer will correct you mid-calculation. The PEI round is structured: leadership, personal impact, entrepreneurial drive, and inclusive leadership are the four stories you should have ready, each with a measurable outcome. McKinsey wants precision and synthesis; the partner round especially weights the recommendation, not the framework.

BCG runs the most candidate-led cases. The interviewer gives you a prompt and lets you drive — you choose the structure, you choose which branch to explore, and the interviewer pushes back when you over-commit. BCG is also the most likely to throw a creativity or estimation curveball mid-case. The Casey chatbot screen is unique to BCG and tests structured-thinking-over-text, which is a different muscle than verbal casing.

Bain weights culture fit earliest and hardest. The case quality bar is the same as McKinsey and BCG, but Bain's interviewers are explicitly told to assess "would I want to be staffed with this person on a long European engagement?" The fit conversation in the first round is not a warm-up — it is the first half of the score. Bain's case style sits between McKinsey's and BCG's, and partners often hand you a Bain-internal slide to interpret rather than reading you a prompt.

The practical implication: prepping for one firm prepares you 80% for the other two. The remaining 20% is firm-specific texture, and it shows up under pressure.

The universal case structure: clarify, structure, analyse, recommend

Almost every MBB case can be reduced to four steps:

1. Clarify. Repeat the prompt back, ask one or two questions to confirm the objective and the constraints, name what success looks like. Ninety seconds maximum. Most candidates skip this and go straight to a framework, which costs them visibly.
2. Structure. Lay out an issue tree. The tree is not a memorised framework lifted from Case in Point — it is a tailored decomposition of the actual question. "Profitability of an airline" should not produce the same tree as "profitability of a SaaS company." Ten to fifteen seconds of silence here is fine and expected.
3. Analyse. Walk one branch of the tree. Do the math out loud. Sanity-check the numbers. When the interviewer hands you a chart or a data point, integrate it explicitly — name what it tells you, name what it doesn't tell you.
4. Recommend. A single sentence answering the prompt, followed by the two or three reasons, the one big risk, and the immediate next step. Aim for sixty seconds. Partners score the recommendation more than they score any single intermediate step.

The trap is the structure step. Case in Point and Case Interview Secrets (Victor Cheng) are reasonable starting points — they teach the rhythm of a case — but at MBB final round, an interviewer can spot a memorised "4 Ps" framework in five seconds and will mark you down for it. The framework is a starting point; the test is whether you tailor it to the actual problem in front of you.

Profitability cases: the canonical structure

Profitability cases are the most common case archetype across MBB. They almost always look the same: profit went up or down by some amount, find out why, recommend what to do.

The canonical decomposition is straightforward:

• Profit = Revenue − Cost.
• Revenue = Price × Volume.
• Cost = Fixed Cost + Variable Cost.

Past that, the question is which of those four levers — price, volume, fixed cost, variable cost — is moving. The interviewer always knows the answer; your job is to navigate to the right branch with as few wasted minutes as possible.

A worked example: "Our airline client's profit dropped 20% year over year. Why, and what should they do?"

A strong structure here is not "let me apply the 4 Ps." It is something like:

• Has revenue dropped, has cost increased, or both? (Always ask this first.)
• If revenue: is it ticket price or load factor? Is it a specific route or all routes? Is it a specific cabin class — economy is more price-sensitive, business is more cyclical?
• If cost: is it fuel (commodity), labour (often unionised, lagged), maintenance (lumpy), or airport fees (regulatory)?
• Cross-cut: is the issue industry-wide or firm-specific? If industry-wide, the recommendation is hedging or capacity discipline; if firm-specific, the recommendation is operational.

The math will be exact. Expect the interviewer to hand you a table — quarterly revenue by route, cost per available seat-kilometre — and ask you to compute a margin. Do it on paper. Read the number back. Do not trust mental math past three digits.

Market sizing: top-down vs. bottom-up

Market sizing questions test whether you can reason from first principles when you don't have the data. "How many electric scooters are sold in India per year?" is a canonical example. So is "How much revenue does a Starbucks store generate annually?"

Two ways to approach:

Top-down starts from a population and filters down. India's population is roughly 1.4 billion; urban population is roughly 35%, so 490 million; assume 40% of urban adults are between 18 and 45 and could plausibly use a scooter, so ~120 million; assume 5% of those buy a scooter in a given year, so ~6 million scooters per year; of which electric is ~10% in 2026, so ~600,000.

Bottom-up starts from production or distribution. India has roughly 25,000 scooter dealerships; the average dealership sells ~30 scooters a month; that's 9 million scooters a year; of which 10% electric, so ~900,000.

The two should land within the same order of magnitude. If they don't, name the discrepancy and reconcile — that's worth more than the number itself. Sanity-check ranges out loud: "Industry reports peg this at roughly 700,000, so my range of 600k to 900k is in the right neighbourhood."

The math will be approximate; the reasoning has to be exact. Round aggressively (1.4 billion → 1.5 billion is fine), but explain every assumption. The interviewer is scoring whether you can decompose, not whether you can multiply.

Market entry, growth, and M&A cases

Three other archetypes you should expect:

Market entry. "Should our retail client launch in Brazil?" The structure is roughly: market attractiveness (size, growth, profitability), competitive landscape (incumbents, switching costs, barriers), client capability fit (do we have the assets to win), and entry mode (organic, JV, acquisition). The recommendation is binary; the interesting part is naming the conditions under which it flips.

Growth strategy. "Our client's revenue is flat — how do they grow 15% next year?" Decompose growth into existing customers (penetration, ARPU), new customers (segments, geographies), new products (adjacent, transformational), and inorganic moves (M&A, partnerships). Senior interviewers want to see you cut by which lever moves the most for the least risk, not just list options.

M&A. "Should our client buy this competitor for $2B?" Three sub-questions: is the target attractive on its own, are there synergies, is the price right? Synergies have to be specified — revenue synergies (cross-sell), cost synergies (overhead, supply chain), tax synergies (NOLs, structure) — and risk-weighted, because most announced synergies don't materialise. A clean answer says yes if value created exceeds price paid by enough margin to absorb the integration risk.

Each of these has a textbook framework. None of the textbook frameworks survive contact with a partner round unless you tailor them.

The PEI / behavioural round

The behavioural side of an MBB loop is increasingly weighted, and McKinsey has formalised it as the Personal Experience Interview (PEI). Three or four structured stories on these themes:

• Leadership — a time you led a team or a project, with a measurable outcome.
• Personal impact — a time you persuaded someone to change their mind, ideally a senior stakeholder.
• Entrepreneurial drive — a time you started something from zero or rebuilt something that was failing.
• Inclusive leadership — a time you brought together people across difference and produced a better outcome than the dominant view would have.

The structure is the same as a STAR answer in any other behavioural round (Situation, Task, Action, Result), but with a McKinsey twist: interviewers are explicitly trained to dig for two layers of follow-up. "What did the other person say back?" "What did you say to that?" "How did the conversation actually end?" If your story is rehearsed too smoothly, the follow-ups will catch you.

A few rules that hold across all three firms:

1. Use first person aggressively. "I noticed", "I proposed", "I argued for". Every "we" is a point you can't claim.
2. Numbers in the result. "We hit 110% of the quarterly target", "the migration finished four weeks ahead of plan", "the new policy applied to roughly 6,000 employees". Generalities like "it went well" lose points.
3. Don't sand off the rough edges. A story where you were wrong for a week and then changed your mind is more impressive than a story where you saw the answer immediately. Interviewers know real work has friction.
4. Have one story per theme that you've never told a friend. PEI stories that have been polished for two years sound polished. Pick something specific enough that it has only ever existed in your head.

The honest amount of rehearsal is uncomfortable: most successful MBB candidates rehearse PEI stories thirty to fifty times each, often on video, until the rhythm is automatic but the words still feel unscripted. There is no shortcut.

Where a real-time AI assistant helps in a consulting interview — and where it doesn't

This is the section to read carefully. Case interviews are conversational and interactive in a way that coding rounds and most behavioural rounds are not. The interviewer is drilling on your thinking in real time, in both directions, with follow-ups designed specifically to expose memorised answers. A real-time AI assistant can be a thinking aid; it is dangerous as a script.

Feature	Profitability	Market sizing	Market entry	M&A	PEI / behavioural
AI help quality	Useful for issue tree	Useful for decomposition prompts	Useful for framework cues	Useful for synergy categories	Low — answers must be your own
Latency requirement	Sub-200 ms	Sub-200 ms	Sub-200 ms	Sub-200 ms	Not applicable
Stealth requirement	High	High	High	High	Critical — interviewer drills on follow-ups
Ethical comfort	Medium — same as a notepad	Medium — same as a notepad	Medium	Medium	Low — feels like fabrication
Recommended use mode	Thinking aid	Thinking aid	Thinking aid	Thinking aid	Not advised

The honest framing: a real-time copilot is at its best as a structuring prompt — a way to remind yourself to ask "is it a revenue problem or a cost problem?" before you commit to a branch. It is at its worst as a script, because case interviews are explicitly designed to reward improvisation and punish memorisation. A candidate who reads a fluent answer off a hidden screen will sound fluent for forty seconds and then collapse on the first follow-up question, because the follow-up is not in the assistant's first reply.

For PEI specifically, our recommendation is to not use the assistant during the round at all. The interviewer is drilling on a story you already lived; if the assistant is doing the storytelling, the story isn't yours, and the follow-up will reveal it. Use the assistant during preparation — to refine story structure, to surface inconsistencies, to coach STAR rhythm — and then close it for the actual round.

Acedly during a live consulting case

For the candidates who choose to use a real-time copilot during cases — typically the structured-thinking branches, not the PEI — Acedly is built around the constraints that matter:

• Eight verified meeting platforms. Most MBB interviews run on Zoom, Microsoft Teams, or — increasingly — Webex for finance-aligned consulting practices. Acedly verifies stealth on Zoom, Teams, Meet, Webex, Lark, Amazon Chime, Coderpad, and HackerRank, which covers more than 95% of consulting interview surfaces.
• ~98 ms median end-to-end latency. End-to-end means microphone-to-render, not just model latency. Anything over 250 ms produces a noticeable lag; case interviews don't tolerate lag because the interviewer is watching your face the entire time.
• Multi-model routing. Claude tends to be the strongest model for structured frameworks (issue trees, MECE decomposition); GPT tends to be stronger for PEI-style narrative reframing during preparation. The assistant routes between them based on the question type.
• 30+ spoken languages, single accuracy bar. Mandarin, Cantonese, Japanese, Korean, Spanish, Portuguese, French, German, Italian, Dutch, Hindi, and Vietnamese are among the most-asked in MBB rounds; the same accuracy bar applies across the full list. Cross-border MBB candidates routinely interview in two languages across rounds.
• Hidden from screen sharing at the OS level. Acedly is excluded from window-capture APIs on macOS and Windows. The interviewer sees the call tile and your shared materials; they do not see the assistant.

The framing for consulting candidates specifically: think of Acedly as a thinking aid for case structure, not a script. Use it to remind yourself of the issue tree, to keep the math straight, to surface a sanity-check range during sizing. Do not use it for PEI, and do not read its output verbatim during a case. The interviewer will hear the difference.

A 6-week MBB prep plan

Six weeks is the minimum amount of time most successful MBB candidates spend in active preparation, not counting the months of casual reading before that. The plan below is what we'd run if we had to start from scratch tomorrow:

Weeks 1–2: case fundamentals. Read Case in Point (Cosentino) cover-to-cover for vocabulary; read the first three chapters of Case Interview Secrets (Victor Cheng) for structure; do five untimed cases solo, walking through the four steps out loud. Use RocketBlocks or PrepLounge for the case banks — both publish actual MBB cases.

Weeks 3–4: live cases. Do thirty cases with a partner, half as interviewer and half as interviewee. CaseCoach and Management Consulted both offer matching services with peers and ex-MBB coaches. Aim for at least five cases with a former MBB interviewer; their feedback is meaningfully different from peer feedback and worth the cost. Record at least three of your sessions and watch them — every candidate has tics they don't know they have.

Week 5: PEI story banking. Pick three to four stories per theme (leadership, personal impact, entrepreneurial drive, inclusive leadership). Write each as a 90-second STAR answer. Rehearse out loud, on video, until the rhythm is automatic. Then have a friend ask you the obvious follow-ups: "What did the other person say?" "What did you do next?" "How did it actually end?" The follow-ups are where most candidates fail.

Week 6: company-specific. McKinsey: take two timed Solve practice tests and dedicate extra time to PEI structure. BCG: practice the Casey chatbot screen — the interaction modality is genuinely different from verbal cases, and reading-comprehension speed matters. Bain: practice slide-interpretation cases (Bain partners often hand you a Bain-internal slide) and rehearse the fit conversation with someone from the firm if you can.

The distribution between case practice and PEI is roughly 70/30 in early weeks and shifts to 50/50 by week six. Most candidates under-invest in PEI and over-invest in cases, and it shows in final-round outcomes.

Frequently asked questions

How long should I prep for an MBB consulting interview?

Are MBB still hiring in 2026 with AI changing consulting?

Do I need an MBA to get an MBB offer?

What's the difference between McKinsey, BCG, and Bain interviews?

Is using an AI assistant during a case interview a good idea?

What's the McKinsey PST and is it still used?