Just cleared the coding portion of the McKinsey SWE loop last month. Here's the breakdown for anyone trying to figure out what to prep.
The process for the coding/online assessment piece had two parts:
Part 1: An async OA before the phone screen. Sent via HackerRank. 90 minutes, two problems. My problems were: A sliding window problem (medium on LeetCode equivalency) A graph connectivity problem (also medium, closer to hard if you went for the optimal solution)
The platform itself had a basic IDE. I used Python 3. The OA felt like a filter, not a deep signal, because the phone screen afterward covered similar ground live.
Part 2: A live coding round during the onsite. One interviewer, 60 minutes, CoderPad. I got one problem that escalated in three stages: naive solution first, then optimize, then 'what if the input is distributed across shards.' That third stage was really about thinking out loud more than implementing.
Difficulty overall: if you're comfortable with LeetCode mediums and have done a handful of tree/graph problems, you're fine. I didn't get any hard-difficulty problems. They're testing whether you can write clean, readable code under mild pressure, not whether you've memorized 150 patterns.
What they noted: I was asked to talk through my logic before coding, not after. The interviewer actively paused me when I jumped into code too fast and said 'can you walk me through the approach first?' Worth keeping in mind.
Languages accepted: Python, Java, JavaScript, C++. Recruiter confirmed this upfront.
One thing I'd flag: McKinsey's tech interview isn't identical across all groups. QuantumBlack runs the most structured coding process. Teams closer to the core consulting business might have lighter coding rounds or none at all. Worth asking early which group you're actually interviewing with.