Zoom · Primly Community

Zoom data scientist interview (SQL + case + stats), full loop notes

analyst_ana · 3 replies

went through the Zoom DS interview in Q1 2026 for a mid-senior role on their product analytics team. here's exactly what the process looked like and what they tested.

phone screen with HM: 30 min. they described the team scope (measuring product engagement, session quality, churn signals) and asked about my most complex analysis project. no SQL or stats in this round.

technical screen: this was two separate 45-minute sessions, back to back.

session 1 was SQL. they used a virtual whiteboard and gave me a schema with tables for users, meetings, participants, and events. the questions escalated in complexity: basic: count distinct users who hosted at least 3 meetings in April medium: compute 7-day rolling average of meeting minutes per user cohort hard: identify users whose session quality score dropped more than 20% week over week for 3+ consecutive weeks

the hard one took most of the time. I used CTEs and a lag window function. they were fine with me thinking out loud and adjusting.

session 2 was case/stats. they described a A/B test scenario: Zoom launched a new meeting UI to 5% of users and wants to know if it improved retention. walk me through your analysis.

I covered: randomization check, metric selection (7-day retention as primary), minimum detectable effect, sample size considerations, and then the actual statistical test (proportions z-test). they pushed on what I'd do if the treatment group had different device distributions (simulate confounding). good question.

behavioral round: one round, 45 min. mostly about cross-functional work with PMs and engineers. standard STAR.

what they care about: depth on SQL window functions, ability to explain stats decisions to non-DS stakeholders, and product intuition for a consumer-ish product. the 'how would you explain this to a PM' question came up twice.

3 replies

analyst_ana

the rolling average + week-over-week consecutive drop questions are really common at product analytics DS roles. I've seen almost identical SQL problems at Spotify and Airbnb too. good to know Zoom is in that same tier of rigor.

brand_ben

yeah the SQL feels like they want to see you're fluent in window functions specifically. frame, partition, lag/lead. if those feel shaky, drill those before anything else.

growth_gabe

the A/B test with confounding question is a great one. I've asked that exact type of question when I've been on hiring panels for DS candidates. it separates people who can run a t-test from people who actually think about validity.