I applied through a recruiter. The process took 4 weeks. I interviewed at Meta (Menlo Park, CA) in Jan 2018
Interview
I was contacted by the recruiter with whom I had a brief 10-15 minute phone call. Then we went ahead with a virtual interview with another data scientist. The online interview consisted of writing code in a text editor answering some data analysis questions such as how do you measure user engagement, how do you measure the effectiveness of a new feature. The interviewer was looking more for the thought process than actual execution of the code.
I got through the online interview and was scheduled for the onsite interview loop consisted of five interviews of 30 minutes each. Two required product analytics brainstorming, one was statistics and probability, and two more were performing ETL using SQL and also performing analysis using SQL. There was a lunch break of an hour in between.
I had a great experience throughout the interview process and the recruiter helps out a lot by emailing preparation materials and calling to discuss the material. Unfortunately I did not get through the last interview. I was not given a reason, just that there were many applicants.
Interview questions [1]
Question 1
How do you calculate monthly active users, churned users and resurrected users from a user activity log with userID and DateTime
Tough interview overall—definitely not what I expected. The technical rounds were intense, particularly when they had me design an A/B test for the News Feed ranking algorithm. I had to discuss metrics and sample sizes in detail. Lucky for me, the time I spent on PracHub right before the interview helped me nail that deep-dive question as it mirrored what I practiced. The behavioral questions felt standard but were still challenging. After a whirlwind process, they extended an offer, which I happily accepted.
Interview questions [1]
Question 1
Design an A/B test to evaluate a new ranking algorithm for the Facebook News Feed. Walk through metric selection (engagement, time-spent, MSI, well-being), unit of randomization given network effects between friends, sample size and power calculations, how you'd detect novelty effects vs. true lift, and how you'd handle a guardrail metric regressing while the primary metric is up.
Total 7 rounds: first round for resume screening, second for technical screening, then for on-site virtual with 4 interviews back to back, then hiring manager round after team matching and then salary negotiation with HR
Interview questions [1]
Question 1
Meta’s evaluation rubrics focus heavily on "Product Thinking over Fancy Math". Interviewers want to see if you can operate like a product owner with an analytical mindset, navigating messy scenarios affecting billions of users
The Interview Process is very structured -
First Tech Screening round - 45 mins (usually can extend a bit depending on the interviewer)
- 2 SQL Questions ( Medium to Hard ) - based on Joins
Full Loop - 4 rounds 45 mins each.
- SQL
- Behavioral
- Analytical Execution - stats & prob, A/B testing, case study
- Analytical Reasoning - Case study
Interview questions [1]
Question 1
Questions on Bayes Theorem, Probability distribution, etc.