Went through the BCG data scientist interview process earlier this year for a role on their Advanced Analytics team. Sharing the full breakdown because I couldn't find anything specific when I was prepping.
The loop was four rounds total. First call was with the recruiter, 30 minutes, pretty standard screening. She was upfront that the DS role at BCG sits at the intersection of consulting and data work, so you need to be comfortable with both.
Second round was a technical screen with a data scientist on the team. About 45 minutes. SQL was the main focus: two questions, one was a straightforward multi-join aggregation (think cohort retention analysis style), the second involved window functions. They wanted me to explain my reasoning out loud as I went. No competitive programming tricks, just real analytical SQL.
Stats questions that came up: Explain the difference between p-values and confidence intervals to a non-technical stakeholder How would you decide if a new model is better than the current one in production? (They pushed on this a lot, wanted me to talk about A/B testing, statistical power, business trade-offs) Classic probability setup: given a skewed distribution of customer spend, which metric do you report and why?
Third round was a mini case interview. This surprised me. It was a structured business problem presented by a consultant, and I needed to use data to frame my approach. They weren't expecting me to crack the case like an MBA candidate but they did want to see logical structuring and hypothesis-driven thinking.
Final round was a longer stakeholder panel, two interviewers, 90 minutes. One focused on behavioral (STAR-format stuff: influence without authority, navigating ambiguity on a project, building trust with a skeptical client). The other went deeper on ML: how I've deployed models, monitoring strategies, when I'd use a simpler model vs. a complex one.
I got the offer. The technical bar for DS felt like a mid-tier tech company, not PhD-level. The case component was the wild card. If you're coming from pure tech with no consulting exposure, spend at least a week practicing structured problem framing before your loop.