BCG Casey Video Recommendation: How to Answer in 60 Seconds

Part ofBCG Casey Online Case: The Complete Guide

The video recommendation is the last task in BCG's online case, and the only one where you speak. You spend about 25 to 30 minutes on Casey's chatbot questions. Then you get one minute to prepare and one minute to turn that work into a spoken answer to the client. This guide covers the format, what the video is judged on, a timed script, three sample answers for one practice case, and a rubric to score your own recordings. For the full test, from the dataset question to results, see our BCG online case guide.

to prepare, as candidates report it

60 s

Source: mconsultingprep.com

to record, with a hard cutoff

60 s

Source: hackingthecaseinterview.com

attempt, per the latest reports (earlier versions reportedly allowed 2)

1

Source: hackingthecaseinterview.com

chatbot questions before the video, in most reports (some say 6–8)

8–10

Source: hackingthecaseinterview.com

What happens in the Casey video question?

When the chatbot section ends, Casey asks you to record your final recommendation for the case on your webcam. BCG does not publish the specifications. The consensus below comes from prep sites that collect candidate reports:

What do BCG assessors look for?

BCG has not published a rubric for the video. Two things point to what matters. The first is BCG's own case-interview advice, which lists "think structurally", "communicate clearly" and "clarify assumptions" as the habits it wants to see (BCG Careers). The second is what prep sites infer from candidate reports. Prepmatter lists four criteria: a clear yes, no or conditional decision stated up front; two or three supporting reasons; confident, well-paced delivery; and a client-appropriate tone.

In practice you are judged on synthesis, not on recall. Summarizing what you did ("first we looked at the market, then we calculated...") is a common mistake. The assessor wants to know what the client should do and why. That is the pyramid principle: one governing point, with the supporting ideas grouped under it.

Nobody outside BCG knows whether a person, an AI or both assess the video. HTCI reports that BCG may use AI analysis alongside or instead of human review, but that is unconfirmed. A structure that works for both readers is a spoken answer that still makes sense as a plain transcript.

How do you structure a 60-second recommendation?

HTCI suggests the split used here: 10 seconds for the recommendation, 35 for three reasons and 15 for next steps or risks. Road to Offer and Prepmatter use close variants. What matters more than the exact split is how many words you can fit into it.

At the roughly 150 words a minute of ordinary conversation (National Center for Voice and Speech, via VirtualSpeech), 60 seconds holds about 150 words. A clarity-focused pace of about 140 words a minute (Baruch College, CUNY) gives you about 140. CaseBasix suggests 120 to 150 words. Nerves speed most people up, so draft about 115 words and let the extra seconds absorb pauses.

0–10 s: answer first

Say the decision and the one number that justifies it: "Northline should launch the airport shuttle. It pays back the $1.5 million investment in under three years." Do not open with "Hello, thank you" or background. If you want a greeting, keep it to three words ("Good morning, Northline").

10–45 s: three reasons with numbers

Signpost ("Three reasons"), then give each reason in one or two sentences with a figure you calculated or read in the case. The reasons should be MECE, meaning they don't overlap and together cover the case. Demand, profit and downside work because each answers a different question. Two reasons that are really the same point ("revenue is high" and "sales are strong") waste 12 seconds.

45–60 s: risk and next steps

Name the single biggest risk, then a concrete action that reduces it or tests it. Specific beats generic: "Sign a three-year contract and test two departure times" beats "we should monitor the market". If the cutoff hits here, you lose the least valuable part.

What does a strong Casey recommendation sound like?

Practice case (hypothetical). Northline Coaches, a regional bus operator, asks whether to launch an airport shuttle. The chatbot section established these figures: 1.2 million airport passengers a year from the region; an expected 8% share, or 96,000 trips; an $18 fare, giving $1.73 million in revenue; $1.2 million in annual operating cost; a $1.5 million upfront investment; and a planned rail link that opens in three years.

Strong (116 words, about 54 s at 130 wpm)

Northline should launch the airport shuttle. It pays back the $1.5 million investment in under three years. Three reasons. First, demand: an 8% share of the region's 1.2 million airport passengers is about 96,000 trips a year, or $1.7 million in revenue at $18 a ticket. Second, profit: operating costs are $1.2 million, so the route adds roughly half a million a year. Third, the downside is limited: we break even at a 5.6% share, well below our 8% base case. The main risk is the rail link planned to open in three years, which could take share. So I'd sign a three-year terminal contract, not five, and test two departure times in the first quarter.

The math checks out: $1.2M ÷ $18 = 66,700 trips, which is 5.6% of 1.2M. The payback is $1.5M ÷ about $0.53M, or 2.8 years.

Average (about 60 s, clipped)

I think Northline should probably go ahead with the shuttle. The market looks attractive and there's a lot of demand from airport passengers. The numbers are positive and the profit seems good. There are some risks, like competition from rail, so they should be careful. Next steps would be to do more analysis on pricing and look at the...

Why it's average: it leads with the decision, but hedges it twice ("I think", "probably"). It uses no case figures, so it could describe any case. The reasons overlap: "attractive market" and "a lot of demand" are the same point. The next step is generic, and the cutoff lands mid-sentence.

Weak (cut off before the answer)

So in this case we started by selecting the datasets on market size and costs. Then we calculated the demand, which was 96,000 trips, and the revenue, which came to about 1.7 million. After that we looked at the operating costs of 1.2 million and the break-even share. We also considered the rail link. So overall, based on all of this, I would say that...

Why it's weak: it narrates the process in order, so the recommendation never arrives. The numbers are right but support no claim. The answer-first 10/35/15 structure exists to prevent exactly this.

How do you score your own practice video?

This is DrillCase's practice rubric, not BCG's. Record, transcribe (your phone's dictation is enough) and score each line 0, 1 or 2.

A score of 8 or more is ready. With 6 or 7, fix the lowest line first. Below 6, drill the first sentence on its own until it comes out in under 10 seconds.

How should you deliver it on camera?

• Camera: at eye level, with your face lit from the front. Look at the lens, not the preview.
• Notes: three numbers and one keyword per reason on paper. Prep sites advise working on paper throughout the case, but reading a full script sounds flat on camera.
• Pace: slow down at the decision and the numbers, and pause briefly between reasons.
• Setup: do a test recording with the same webcam, browser and room before test day. Candidates report no pause option once the case starts.
• Integrity: use only what your invitation allows. BCG's guidance on AI in applications asks for "genuine human insight, not machine-generated output" (BCG Careers). Rehearse with scripts and timers beforehand, never during the live assessment.

How should you practice the Casey video?

The video only works if you have a case to summarize, so practice it at the end of a full chatbot case rather than on its own. Run a Casey-style case, then set a phone timer: 60 seconds to prepare, 60 to record. Transcribe the recording and score it against the rubric above. HTCI recommends at least five practice recordings before test day. Road to Offer's two-week plan goes further, with two days of five or more recordings a day.

DrillCase has 25+ Casey-style chatbot cases, one of them a free demo. Each gives you a finished case to turn into a recommendation. See them on the BCG Casey practice page or browse the case list. If your invitation names a different test, read Casey vs the BCG CCA first. For how chatbot cases differ from live ones, see the chatbot case format breakdown.