Casey is BCG's chatbot-guided online case. You get a business situation in slices, you answer 8 to 10 questions on a clock (commonly about 25–30 minutes; some BCG pages say ~30, some ~35), and you usually cannot go back. Some invites end with a short spoken or video recommendation. The email in your inbox beats every blog, including this one.
This page is format and method. Vendor comparison and the $69 ten-case pack: best Casey simulator. Live interviewer cases: BCG case interview guide. Two original questions run free at the chatbot.
Last full pass: 24 August 2026. CoachNed is independent practice material, not affiliated with BCG.
What Casey is not
It is not a live candidate-led case. There is no partner to rescue a vague structure. It is not McKinsey Solve (games). It is not Bain SOVA (psychometrics). It is not the Consulting Career Assessment (CCA) numerical test some offices use instead or in addition — if your invite says CCA, use the CCA materials, not this page.
It is also not a leaked BCG exam. Anyone selling "real Casey questions" is a problem. Use original practice.
Question types you should expect (shape, not a script)
| # in a typical 8–10 | Type | What a good answer does |
|---|---|---|
| 1–2 | Objective / structure or which data to open | Names the decision; picks the dataset that tests it |
| 2–5 | Exhibit + multiple choice or short numeric | Title, units, period, then the implication |
| 4–7 | Math (often with an on-screen calculator — follow invite rules) | Relationship first, then the number |
| 6–8 | Brainstorm / next analysis | Grouped ideas, one recommended next step |
| Last | Recommendation, sometimes 60-second video | Answer first, two reasons, one risk |
Pacing: if you average more than ~3 minutes early, the last items get rushed. Unanswered is worse than imperfect.
Regional BCG careers pages disagree on duration and calculator policy. Read yours.
Worked example: which dataset, then the math
Context the bot might give. A grocery chain wants to add automated pickup lockers in 400 existing stores. Question: which stores get the first 80 lockers this year?
Q1 — dataset selection (this is the Casey tell). You can open one of:
- A. Last year's total store revenue, ranked.
- B. Share of online orders that are currently picked in-store vs delivered, by store, plus parking-lot dwell time.
- C. Competitor locker locations within 5 km.
- D. Employee satisfaction scores.
Pick B. The decision is locker ROI via pickup density and friction, not "biggest store." A is a decoy for prestige stores. C is useful later for competitive response, not for the first 80. D is not the constraint.
Q2 — exhibit. Among stores with online mix ≥ 12 percent, average weekly pickup-eligible orders = 420. A locker module handles 70 pickups/day at 85 percent utilization before overflow to staff. Staff overflow costs $4.20 per extra order. Locker opex is $1,100/week per store. Contribution per pickup order vs delivery is $3.80 extra (last-mile save).
Math. Weekly locker capacity = 70 × 7 × 0.85 = 416.5, call it 417. Demand 420 ≈ at capacity. Extra contribution vs delivery: 420 × $3.80 = $1,596/week. Opex $1,100. Net ≈ $496/week before overflow. Overflow 3 orders × $4.20 is noise. Annualized ~$26k net per store before capex. If locker capex is $40k, simple payback ~1.5 years on this segment — only in the high online-mix stores. That is why Q1's dataset mattered.
Q3 — recommendation (text or video). "Pilot 80 lockers only in stores with ≥12 percent online mix and parking dwell below the chain median. Expected ~$500/week contribution after opex on the exhibit math. Do not rank by total revenue. Risk: 420 weekly orders is an average — weekend peaks will overflow; I would check Saturday demand before a 400-store rollout."
Practise that sequence — select data, compute, close — on Casey-style questions. CoachNed: two questions without an account, full timed case after signup, ten original cases on Sprint/Annual or $69.
Calculator, video, and integrity
Follow the invite. If they say no external calculator, do not open one. If they want camera on, treat it as a professional recording: answer first, then reasons, then stop. A 60-second rec that starts with "so um the case was about lockers…" has already lost the structure score.
Do not screenshot, do not paste into ChatGPT during the test, do not buy stolen items. BCG's process is an integrity screen as well as a case screen.
Failure modes that are Casey-specific
- Skipping dataset selection. You cannot recover if you opened total revenue and never saw pickup density.
- Going back. You cannot. Commit, live with it.
- Writing a live-case novel in the chat. The bot wants a crisp pick, a number, or a short rationale.
- Video that recaps. Recommendation, two numbers, one risk. Sit still.
- Only live casing. Candidate-led mocks teach you to ask for exhibits. Casey does not let you ask. You need chatbot reps.
- Confusing Casey with CCA. Different products. Invitation language.
A six-day plan
Day 1. Read your invite. Note time, question count, calculator, video.
Day 2. Two free chatbot questions. Debrief: did you pick the right data?
Day 3. Chart + percent drills on drills. Casey is exhibit math under a clock.
Day 4. One full timed Casey-style case.
Day 5. Record a 60-second rec of yesterday's case. Watch it. Cut the throat-clearing.
Day 6. Second timed case or the paid pack if you need volume. Then stop and sleep.
After Casey, you still have live BCG cases and fit. The chatbot is a gate, not the offer.