In a case interview, you’re handed a business problem you’ve never seen and get about 30 minutes to work through it out loud. BCG describes it as a realistic business challenge where “there isn’t always a single ‘right’ answer”; what counts is your approach and your reasoning.
Memorized frameworks won’t carry you through that. This guide covers the 2026 formats, the five stages of any case, how to use frameworks as building blocks, a fully worked example, a 4–6 week practice plan, and an interview-day checklist.
Cases are standard in consulting: MBB (McKinsey, BCG, and Bain), the strategy arms of the Big 4, and strategy boutiques. They also turn up in corporate strategy and in case-style interviews in tech. Investment banking is the exception; those interviews are technical.
Cases usually share a session with a behavioral interview, one of the three main interview types to prepare for.
Candidate-led cases. You drive: you propose the structure, choose which branch to explore, and ask for the data you need. Historically common at BCG and Bain.
Interviewer-led cases. The interviewer asks a set sequence of questions (structure this, read this exhibit, run this calculation, recommend), and may move you on before you feel finished. McKinsey’s Problem-Solving Interview, “sometimes referred to as a case interview” in the firm’s words, is widely described as interviewer-led, matching the question-by-question style of its published sample cases. Prepare for both: even here you can lead with a hypothesis and a suggested next question.
McKinsey Solve. A gamified problem-solving assessment for most consulting roles. McKinsey says there’s no need to prepare; its rules bar AI tools, notes, screenshots, and outside help.
BCG. BCG’s published process lists an application, a skills interview, a case interview, and a team interview. Many candidates also report an online, chatbot-style case (widely called Casey) ending with a recorded video recommendation. That isn’t on BCG’s official pages, so ask your recruiter.
Bain. Bain says you may “complete a screening test ahead of your interviews,” and some roles add a take-home challenge with about three days to finish. Many offices reportedly moved first-round cases to a standardized, interviewer-led format; senior rounds may still be candidate-led.
| Format | Who drives | What’s tested | How to prepare |
|---|---|---|---|
| Candidate-led live case | You | Structuring, prioritizing, asking for data, synthesis | Partners who withhold data until you ask |
| Interviewer-led live case | The interviewer | Structure, exhibits, math, crisp answers under pressure | Drill 3–5 minute answers to single questions |
| Online chatbot or written case | The platform | Charts, math, a recorded or written recommendation | Timed chart practice; record a 60-second close |
| Gamified assessment (Solve) | The game | Problem-solving in an unfamiliar system | No prep, per the firm; rest and a quiet setup |
If you’re an undergrad. Consulting timelines moved earlier for summer 2027: Bain’s first deadline was March 29, 2026, sophomore spring for most applicants, so prep may start in sophomore winter. Firms judge undergrads on structure and communication, not business knowledge.
If you’re an MBA student. Most programs run internship interviews in January; at MIT Sloan the 2026–27 interview period begins January 4, 2027. Interviewers expect more business judgment, especially in the recommendation.
Check when MBA consulting interviews happen at your school, then use the club’s fall case workshops, part of breaking into consulting through an MBA.
Memorized frameworks backfire: they sound generic, miss the real driver, and get heard all day. A case tests whether you can analyze a novel business problem in real time, so frameworks are half of prep; the other half is live cases with peers, where you structure a question you didn’t expect.
| Case type | Core question | Building blocks | Example prompt |
|---|---|---|---|
| Profitability | Why did profit change? | Revenue (volume × price) and costs (fixed, variable), by segment | Grocery chain’s margins fell two years running |
| Market entry | Should the client enter this market? | Market attractiveness, competition, client capabilities, entry economics | Pet insurer weighs a Canada launch |
| Growth | Where does growth come from? | Existing customers, new customers, new products, acquisitions | Ski resort wants 25% more winter revenue |
| Pricing | What price should the client set? | Cost-based, competitor-based, value-based | Software firm pricing an AI add-on |
| M&A or investment | Is the target worth buying? | Standalone value, synergies, price, risks, fit | PE fund weighs a car-wash chain |
| Market sizing (often inside another case) | How big is the market? | Top-down and bottom-up estimates, cross-checked | How many e-bikes sell in the US yearly? |
Private equity due-diligence cases are also common, a skill that carries over if you later consider moving from consulting into private equity.
The example below starts from the profitability building blocks, then splits the data in a way no generic framework would suggest.
Hypothetical case: RidgeLine Coffee is fictional, the figures invented, and nothing here comes from any firm’s case library. It runs candidate-led.
Interviewer: “Our client, RidgeLine Coffee, runs drive-through coffee kiosks in the Southwest. Revenue grew from 60millionin2024to71 million in 2026, but operating profit fell from 12millionto9.7 million. The CEO wants to know why, and whether to open 25 more kiosks in its newest metro.”
You: “So revenue is up 11millionwhileprofitisdown2.3 million. You want the cause and a call on the 25 kiosks. A few questions first?”
Flat pricing and costs rule out two branches before you draw them.
You: “Could I take a minute?” After 60 to 90 seconds:
You: “My hypothesis: revenue grew because RidgeLine added kiosks, so the new ones are probably less profitable than the mature ones, with added overhead on top.”
A generic tree stops at company totals, which look like a cost problem with no cause. Splitting by kiosk age opens the case.
Interviewer: “Here are the 2026 kiosk economics. Overhead was 3.0millionin2024and4.0 million in 2026, from a new regional office and launch marketing in the new metro. In 2024, all 60 kiosks were mature.”
| Metric (2026) | Mature kiosk (opened pre-2024) | New kiosk (opened 2025–26) |
|---|---|---|
| Kiosks | 60 | 20 |
| Revenue per kiosk | $1,000,000 | $550,000 |
| Ingredients, packaging (30%) | $300,000 | $165,000 |
| Labor | $350,000 | $350,000 |
| Rent | $100,000 | $100,000 |
| Kiosk contribution | $250,000 | –$65,000 |
Kiosk contribution is revenue minus that kiosk’s own costs, before overhead.
You: “Does this reconcile?”
You: “So the 2.3milliondeclineis1.3 million of new-kiosk losses plus 1.0millionofaddedoverhead;thematurekiosksareunchanged,andmarginfellfrom20450,000 of labor and rent, and 0.70 × revenue = 450,000gives∗∗about643,000**, roughly 17% above today’s 550,000.At360daysanda6.00 ticket, that’s a volume target.”
| Kiosk | Annual revenue | Per day | Transactions per day |
|---|---|---|---|
| New, today | $550,000 | ~$1,528 | ~255 |
| New, at breakeven | ~$643,000 | ~$1,786 | ~298 |
| Mature | $1,000,000 | ~$2,778 | ~463 |
You: “So a new kiosk needs about 43 more cars a day, three or four an hour over a 12-hour day. Is the shortfall even?”
You: “How do the 20 new kiosks compare with each other, and with the original kiosks’ ramp?”
Interviewer: “In the original metro, kiosks averaged 600,000inyearone,800,000 in year two, and 1,000,000fromyearthreeon.Thesehavebeenopenabout15months.The12commuter−corridorkiosksaverage700,000; the 8 retail-center kiosks average $325,000.”
You: “Check: 12 × 700,000+8×325,000 = 11.0million,or550,000 per kiosk. Consistent. Now contribution by type.”
You: “So it isn’t the new metro, it’s the 8 retail-center sites. The corridor kiosks beat the historical ramp, 700,000at15monthsagainsta600,000 year-one average; retail-center sites are far behind. And cost cuts don’t save them: at 250,000oflabor,contributionisstill0.70×325,000 – 350,000=∗∗–122,500**.”
About 60 seconds, answer first.
“RidgeLine should pause the 25-kiosk expansion until it changes how it picks sites, and exit or relocate the 8 retail-center kiosks as leases allow.
The whole decline comes from new kiosks, 1.3million,andnewoverhead,1.0 million, not the core business, and it sits in 8 retail-center sites down about 1.78millionayearcombined,whereevenasharplaborcutleaveseachatabout–122,500. The 12 corridor kiosks each contribute $40,000 and are ramping ahead of the historical curve, which supports expansion in corridor sites only.
Exiting the 8 sites would add up to about 1.78millionayearbeforeclosurecosts,liftingoperatingprofittoward11.5 million, with the corridor ramp closing the gap to 12million.Risks:lease−exitcosts,thinnerbrandpresence,and1.0 million of overhead across fewer kiosks.
Next steps: review lease terms and exit costs, rebuild the site-selection model around traffic counts, and set ramp milestones before approving new openings.”
An interviewer-led version asks directly: “What factors would you look at?”, then “What’s the breakeven revenue?”, then “What do you recommend?” Same skills, but each answer has to stand alone.
Case math is narrow and trainable. Practice these:
Calculators typically aren’t allowed in live cases, so practice on paper. McKinsey encourages practicing interview questions with AI, but generating answers in real time during an interview, or using AI in assessments where it isn’t permitted, breaks its rules.
Start before invitations arrive: the gap between invitation and interview can be a week or two, and structure and case math resist cramming.
If you’re an undergrad. Use your consulting club and career center’s mock-interview program for partners who withhold data until you ask. A common guideline is 15 to 25 live cases before first rounds.
If you’re an MBA student. Club case sessions cluster in November and December, and winter break works as a training camp before January interviews. A month-by-month internship prep calendar spreads cases, stories, and firm research across the fall.
Work one full case from prompt to recommendation this week, out loud and timed, to see which stage needs work.
