There aren't many interview reports about this company on the forum, so I'm sharing mine.
It was a one-hour phone screen, split roughly half and half between SQL and experimentation.
SQL: there were two questions total. I don't remember the exact tables, but one of them asked me to compute the 7-day rolling average of some metric. I got stuck here — the interviewer's correct answer was that you need a window function with RANGE BETWEEN INTERVAL 6 DAY PRECEDING. I probably got dinged on this one.
Experimentation:
The interviewer shared a Google doc. The question was: the team wants to run two A/B experiments — Experiment 1 changes the order of items within each shelf, with a sample size of 50% of traffic; Experiment 2 changes the order of shelves within the Fetch app, with a sample size of 9% of traffic.
What do you need to watch out for when running these experiments? If you can't guarantee the traffic is fully isolated between them, does that mean the experiment conclusions are completely unusable?
If you run the experiments without isolating traffic between them, how would you measure the impact of Experiment 2 on Experiment 1?
The interviewer was very friendly throughout, kept nodding and saying "make sense," and didn't really push for more detail or give any hints during the case study.
Timeline: interviewed on Monday, got the rejection on Thursday.
Discussion
Loading comments…