Meta Data Analyst Interview Loop
Meta has three analyst-type roles, and people preparing for a Meta data analyst interview often mix them up. Data Analyst is the most business-facing of them. There's less experimentation and more turning existing data into something a partner can act on. The loop is a recruiter call, a screen that's mostly SQL, and an onsite of four 45-minute interviews: SQL, product and metrics, an analytical case, and behavioral. Check with your recruiter which one you're up for, because the focus differs.
What decides it
Speed with accuracy, and a clear point of view. Meta wants people who won't slow a team down, and hedging comes across as risk. Round 3 is where offers and rejections split.
Practice by round
Timed SQL on product tables. It's about five short questions, and they check fluency rather than depth. Meta's screens are fast, and the interviewer moves on when you're done. So it pays more to get the first three right quickly than to be thorough on the fifth.
Engagement, retention and funnel metrics on users, posts, sessions and events. The level is medium to hard, and you're expected to use window functions and cohort logic.
| Question | Difficulty |
|---|---|
| Share of Active Users | Medium |
| Successful Lower Priced Product | Medium |
| Top Three Classes | Medium |
| Same Brand Purchases | Medium |
| Recommendation System | Medium |
| Most Lucrative Products | Medium |
| Highest Sales with Promotions | Medium |
| VR App Sessions | Medium |
| Rank Variance Per Country | Hard |
| Retention Rate | Hard |
| Popular Posts | Hard |
| First and Last Day Promotion Results | Hard |
| Product Families | Hard |
| Sales Percentage Week's Beginning and End | Hard |
| Daily Revenue | Hard |
Product and metrics. You define a success metric for a product, decide whether to launch, judge whether an ad worked, and measure the health of Groups. Interviewers mark you down if you reach for an experiment template before clarifying the goal. The analytical case is the same logic in reverse. You find out why two numbers disagree, list the causes, and pick what to check first. Talk in trade-offs.
| Case study | Difficulty |
|---|---|
| Returning User | Easy |
| McD's Big Macs | Medium |
| Too Many Friends | Medium |
| Facebook Groups in the US | Medium |
| Fake News on Facebook | Hard |
It's lighter than the data scientist loop, but it's still there. You'll cover what an A/B test is for, what sample size depends on, and how to read a conversion rate. You won't derive anything. You'll be asked whether a number can be trusted.
| Question | Difficulty |
|---|---|
| Expectation of Variance | Medium |
| Mean, Median Age in Mexico | Medium |
| Used by Different Groups | Medium |
| Serving ads within Newsfeed | Medium |
| Sample Size | Medium |