Round 1: Measurement & Modeling Concepts
A case study: Google Meet has a big customer who complains that meetings keep dropping/disconnecting. They ask how you'd analyze and solve this problem, how you'd put together an analysis plan, and how you'd estimate the impact. If it's affecting retention, how do you build a model to predict retention (contract renewal)? How do you evaluate models, what attributes would you use, and when do you use logistic regression versus more complex models?
How do you decide whether to fix a bug or build something new?
Engineering shipped a new version that improves this feature, but you can't run an A/B test — how do you evaluate whether the feature is actually effective?
Round 2: Coding, Applied Analysis, & Experiments
No coding was actually asked. One of the questions that's come up before on the forum: if Google Maps added a feature recommending the best jogging route, how would you measure whether the idea is good?
How do you set up the experiment, what success metrics would you use, and getting into the details — what's the triggering logic, and how do you get the MDE?
They also asked a small follow-up: what kind of user segments could you build for Gmail, and how would you use that segment afterward?
The questions leaned more toward things you'd actually run into on the job. I felt like my answers were so-so — not great, not terrible — but there were some details I didn't handle well. During the interview, don't rush: make sure you nail down every concept and explain it clearly.
Got the rejection notice after 3 days 😮💨. Better luck next time.
Discussion
Loading comments…