BCG case interview: practice cases with worked examples (2026)

A BCG case interview is commonly candidate-led: the interviewer sets out a client problem and you drive it, deciding what to analyse next, asking for data and saying why. Cases tend to be exhibit-heavy, with several charts or tables, some carrying more data than you need, and interviewers look for a hypothesis that you test and update as the evidence arrives. Candidates commonly report two or three interviews of about 45 minutes in a first round and more with senior staff in a final round, each with questions about your experience. Some offices add the Casey online case or a written case. Check your invitation.

Candidate-led changes what good looks like. Silence after an exhibit is a missed turn: the interviewer is waiting for you to say what it means and what you want next. BCG interviewers commonly probe hypothesis-driven problem solving: say what you think the answer is, how you would prove or kill it, and change your mind cleanly when the data says so.

Our library has four cases written in the BCG style: candidate-led, with exhibits you ask for and a hypothesis to keep updating. Each has a guided walkthrough and can be played live with the AI interviewer, who will wait for you to drive.

Start with Halcyon, the most familiar case type. Leave Helix and Northstar until you can drive a medium case without prompting.

All three firms test structure, maths, reading data and judgement. What differs is who drives, how much data you get and how fit is tested. Read this as a tendency, not a rule: interviewers vary, and every firm uses every case type.

If you apply to more than one firm, practise both rhythms: our McKinsey case interview guide covers the interviewer-led format.

A hypothesis is your current best answer, stated early so it can be tested, not a guess you defend. The loop repeats with every exhibit: say the hypothesis, ask for the data that would test it, read the exhibit for the one thing that matters, then say what changed. Here it is on a short illustration, separate from the library cases so nothing is spoiled.

In a candidate-led case you create many of the questions yourself, but the same types recur. Know which one you are in, and what the interviewer is marking.

Each example stops at the end of the first exhibit, with the hypothesis you would say out loud. The rest is yours to play.

Halcyon Mobile: growing into lower profit

The prompt. Halcyon is the third-largest mobile operator in its market. In two years it has grown subscribers 12.5% to 9 million, yet EBITDA has fallen by about a third. The CEO can't see how a growing business makes less money.

A structure that fits. Write profit as an equation and give the cost of acquiring growth its own branch; it is the one candidates forget. Open with a hypothesis: the growth may be coming from less valuable customers, or costing more to win.

The exhibit notes that net additions were 0.5m in both years. First calculations. Service revenue rose just $36m (1.25%) on 12.5% more subscribers, because ARPU fell 10%. Cost of service rose $9m and network costs $20m. The big move is acquisition: subsidies plus commissions went from $340m to $665m, up $325m, more than the whole $318m EBITDA fall. Check: +36 - 9 - 20 - 277 - 48 = -318. And gross additions rose 56%, from 1.7m to 2.66m, to deliver the same 0.5m of net growth.

BCGHalcyon Mobile grows its way to lower profitProfitability · Medium

Halcyon Mobile grows its way to lower profit

The best first BCG-style case: a profitability puzzle you drive from the P&L.

Interviewer

Halcyon Mobile is the third-largest mobile operator in its market. Over the past two years it has grown its subscriber base by 12.5% to 9 million, yet EBITDA has fallen by about a third. The CEO cannot understand how a growing business is making less money. What is going on, and what should Halcyon do?

Start the caseGuided, step by step

Larkfield: doubling online sales

The prompt. Larkfield is a mid-market fashion retailer with 220 stores. Store sales have been flat at about $1.2bn for four years, while online has grown to $300m. The new CEO has promised to double online revenue within three years. How would you help her get there?

The exhibit's note: online fashion in Larkfield's market grows about 8% a year. First calculations. Online grew from $238m to $300m in three years: 300 ÷ 238 = 1.26, or 8% a year, exactly the market's rate. Online's share of sales rose from 16% to 20% only because stores stood still. Doubling in three years needs the cube root of 2, about 1.26, so 26% a year: more than three times today's pace.

BCGLarkfield sets out to double online salesGrowth Strategy · Medium

Larkfield sets out to double online sales

A growth case built as a bridge from $300m to $600m. Test every lever against margin.

Interviewer

Larkfield is a mid-market fashion retailer with 220 stores and a growing online business. Store sales have been flat at about $1.2bn for four years, while online has grown to $300m. The new CEO has promised the board she will double online revenue within three years. How would you help her get there?

Start the caseGuided, step by step

Helix Diagnostics: pricing a new test

The prompt. Helix has a blood test that detects a post-surgical infection 40 hours earlier than the current standard of care. It is about to launch in US hospitals. What should it charge?

A structure that fits. Bracket the price before choosing it, then decide where to land:

• Floor: Helix's marginal cost per test.
• Reference: the price of the test it would replace.
• Ceiling: what earlier detection is worth to the hospital, per test.
• Price: the share of that value Helix captures, by hospital segment, given a likely competitor.

The exhibit's note: both tests are run on every patient after a target procedure, and hospitals are paid a fixed bundle per procedure, so complications come out of their own margin. First calculations. A typical hospital sees 1,200 × 6% = 72 infections a year, each caught 48 - 8 = 40 hours earlier with Helix. And the obvious anchors don't work: Helix's $62 cost is above the $45 incumbent price, so matching the current test would lose $17 on every test.

BCGHelix Diagnostics prices a new testPricing · Hard

Helix Diagnostics prices a new test

A hard pricing case. Be careful what you divide the value by.

Interviewer

Helix Diagnostics has developed a blood test that detects a particular post-surgical infection 40 hours earlier than the current standard of care. They are about to launch in US hospitals and need to set a price. What should they charge?

Start the caseGuided, step by step

Northstar Bank: losing deposits to a challenger

The prompt. Northstar is a mid-size UK retail bank. Over 18 months a digital-only challenger has taken £1.66bn of its retail deposits, about 9% of the £18.4bn it holds today. The CEO asks whether to match the challenger's savings rate.

A structure that fits. Weigh what defending costs against what losing costs, then segment: the answer is rarely the same for every customer.

The exhibit notes that almost all of the £1.66bn left from instant access, and that wholesale funding at 4.6% is the alternative. First calculations. Instant access is 62% of the book (11.4 ÷ 18.4), and it is where the money is leaving. Matching means a 1.9-point rise (4.1% - 2.2%), and a rate change reprices every pound in the product, not just the pounds at risk: 1.9% × £11.4bn = about £217m a year.

BCGNorthstar Bank loses its depositsCompetitive Response · Hard

Northstar Bank loses its deposits

A hard competitive-response case. Weigh the cost of defending against the cost of losing.

Interviewer

Northstar is a mid-size retail bank. Over eighteen months a digital-only challenger has been taking deposit share in their core market, and Northstar has lost about 9% of retail deposit balances. The CEO is asking whether to match the challenger's savings rate. What do you think?

Start the caseGuided, step by step

Two parts of the BCG process vary by office, so read your invitation first.

The Casey online case

Casey is a chatbot-led online case that some BCG offices use before interviews. Candidates commonly report about 25 to 35 minutes for around eight to ten questions, each on its own timer, ending with a short recorded video of your recommendation. You choose data, type short answers and do calculations as you go. Our Casey guide covers the format and lets you play a full practice case.

Written cases

Some BCG offices have used written cases, most often in later rounds. Formats vary, but a common pattern is a pack of documents and exhibits, a fixed time to analyse it alone, then a short presentation of your recommendation, sometimes on slides you prepare, followed by questions. It runs the same hypothesis loop as a live case, with more data to sift and nobody steering. If your invitation mentions one, practise reading a full case pack against the clock; the BCG hub covers the wider process.

Run this list after every practice case. Anything you can't tick is your next drill.

• I stated a hypothesis within the first few minutes and said how I would test it.
• I asked for specific data, and said why, rather than waiting to be handed it.
• On each exhibit I found the one insight, linked it to the last one and said what changed.
• I set up my own calculations and chose the ones that answered the question.
• When the data killed my hypothesis, I said so and stated a new one.
• I closed with the answer first, three reasons with numbers, a risk and next steps.

Play each case twice: guided first, with feedback on every answer, then live with the AI interviewer, who waits for you to drive. Write your opening hypothesis down before each live run, and afterwards compare how it evolved with the walkthrough.