BCG practice that works is candidate-led: you pick the next analysis, you choose the exhibit, you close without a partner feeding you questions. Reading ten fully worked McKinsey interviewer-led cases will teach you the wrong room. This page is the practice method plus one complete brewery case you can run cold.
For format and scoring, use the BCG case interview guide. For the chatbot screen, use Casey and the simulator. McKinsey's 10-case bank is a different product.
Last full pass: 24 August 2026. CoachNed is independent practice material, not affiliated with BCG.
The practice loop (steal this, not a casebook PDF)
| Step | Time | What you do | What you do not do |
|---|---|---|---|
| 1. Prompt only | 2 min | Cover the rest of the page. Objective + constraint out loud | Read the exhibit values |
| 2. Drive plan | 2 min | Tree + hypothesis + "I want exhibit X first" | Recite profitability / 3Cs / 4Ps |
| 3. Request data | 1 min | Say which chart and why | Flip through every table |
| 4. Quant | 4–6 min | Equation, units, number, implication | Silent calculator heroics |
| 5. Close | 90 sec | Answer, number, risk, next step | Recap of buckets |
Then wire the miss to a drill, not to another 40-minute case. Structure miss → structure drills. Chart miss → graph drills. Soft close → synthesis. Full scored rep: try a case.
Two Casey questions are free; ten original chatbot cases are on Sprint/Annual or $69. That pack trains Casey, not this live loop. Use both if your invite includes both.
Worked example: regional brewery, distributor squeeze
Prompt. Northbridge Brewing sells 1.4 million cases a year, 70 percent through a single distributor that just announced a 9 percent fee increase. Direct-to-retail (the other 30 percent) is operationally messy. EBITDA today $11.2 million. The founder wants a recommendation in one meeting: accept the increase, switch distributors, or pull more volume direct.
Cover the rest. Build your drive. Decision: three options, not "analyze the beer market." Branches: (1) what the 9 percent actually costs in dollars and whether price can move, (2) whether a second distributor exists at a fee that nets better, (3) whether direct contribution, after new opex, beats the distributor on the contestable volume. Hypothesis: the fee is a $1 million-plus problem; the answer is a mix, not a heroic in-house fleet.
Exhibit pack (open contribution by channel first).
Distributor channel: 980,000 cases. Revenue/case $18. Distributor fee today 18 percent of revenue; proposed 27 percent of revenue. Brewery COGS $7.20/case. Direct channel: 420,000 cases. Revenue/case $20 (no distributor). Direct opex (sales + logistics) $3.40/case today. A second distributor has quoted 20 percent of revenue but will only take brands that already have 15 percent share in its territory; Northbridge has 6 percent.
Math, spoken.
Distributor revenue = 980,000 × $18 = $17.64 million. Fee today 18% = $3.18 million. Fee at 27% = $4.76 million. Delta = $1.58 million, which is 14 percent of EBITDA — not a rounding error.
Contribution per distributor case today: $18 − $7.20 − 0.18×18 = $18 − $7.20 − $3.24 = $7.56. After the hike: $18 − $7.20 − $4.86 = $5.94. Direct contribution: $20 − $7.20 − $3.40 = $9.40. Direct is better per case, but only if opex stays at $3.40 when volume triples. A realistic step-up: add $1.1 million of people and trucks to move another 400k cases → $1.1 million / 400k = $2.75 extra opex per new direct case. New direct contribution on the shifted volume: $20 − $7.20 − ($3.40+$2.75) = $6.65, which is only $0.71 above the post-hike distributor case. Switching to the second distributor is not available at 6 percent share — the quote is a decoy until share moves.
Synthesis. "Do not accept the 9-point fee as a full swallow and do not 'go fully direct' this year. The hike is $1.58 million, about 14 percent of EBITDA. I would (1) pass through roughly half in price on the sticky draft accounts, targeting ~$0.8 million, (2) move 150–200k cases of the highest-margin SKUs direct where the route already exists, and (3) tell the distributor the rest of the fee is a two-year problem we will bid out once share in the adjacent territory is above 15 percent. Full switch is fantasy on the quote they gave. Risk: retailers refuse the pass-through and volume falls — I would test price on 10 percent of accounts before a chain-wide letter."
Run this prompt cold on a timer before you reread the numbers. Then do a different industry on CoachNed so you do not memorize beer.
What to practise besides this one case
BCG live cases draw from profitability, growth, ops, M&A, and public sector. You need reps in the drive, not a catalogue of ten solved stories on this URL. Official-style BCG materials and student casebooks are volume; treat their "answers" as one path.
Casey is a separate muscle: dataset selection, no going back, a short video rec at the end on some invites. Log Casey on different days from live mocks.
Fit: 6–8 PARADE stories, not 200 prompts. BCG fit guide. Do not duplicate McKinsey's PEI question list here.
Self-score after every practice case
| Check | Pass | Fail |
|---|---|---|
| Did I name the decision in sentence one? | "Accept / switch / mix" | "Let's look at the market" |
| Did I ask for one exhibit on purpose? | Named it | Read the pack in order |
| Did mix or units show up in the math? | Fee in dollars of EBITDA | Percent theater |
| Did I kill a decoy? | Second distributor quote | "We should switch" |
| Close under 90 seconds with a number? | $1.58 million in the first line | Bucket recap |
Three fails on the same row: stop full cases. Drill that row.
Failure modes this page is built to kill
- Using the McKinsey bank as BCG prep. Interviewer-led passivity is a BCG no.
- Reading the worked math first. You destroyed the only value of this brewery case.
- Casey-only prep. The chatbot never lets you choose exhibit 4. The live case does.
- Another casebook instead of a drill. If the tree is slow, structure is cheaper than case 14.
- No fit reps. BCG still scores whether a staffer trusts you. Ten cases and zero stories is how people surprise-fail.
When the loop is boring, you are doing it right. Variety comes from new prompts, not from a new framework name.