Candidates report a Meta Data Scientist process built around a SQL and product sense screen, then a loop mixing SQL, product case studies, statistics and behavioral questions. Here is what to prepare.
$178K
Avg. Base Comp
$350K
Avg. Total Comp
4-5 rounds
Typical Rounds
Not reported
Process Length
Candidates describe the Meta Data Scientist interview as a screen followed by a loop, with SQL and product sense carrying most of the weight. Several report a technical screen that pairs SQL with a product case, often built on social products such as group calls, bad-content detection or notifications. One candidate also describes CoderPad with no way to run queries, so write SQL accurately without executing it.
The loop is where reports diverge. Four candidates describe four rounds, typically covering SQL, product or analytical case work, statistics and probability, and a behavioral conversation. One senior candidate instead described coding, system design, experiment design and behavioral rounds, so expect some variation by level.
The case questions repeatedly chain a launch decision to experiment design. Candidates were asked how to decide whether to build a group calling feature, which target metric to track, and how to handle a result where usage rises but daily actives fall. Others were asked about network effects, spillover between treatment and control users, and stratified randomization. Practice moving from clarifying questions to metrics to test design without losing structure, and say your randomization unit and test duration before proposing metrics.
Pace is a recurring theme. Interviewers move quickly and probe with follow-ups, so prioritize a few metrics and reach a recommendation. Reported outcomes are mixed, including candidates who passed screens, some who did not advance, and one offer.
Synthesized from 73 candidate reports by our editorial team.
Had an interview recently?
Share your experience. Unlock the full guide.
Real interview reports from people who went through the Meta process.
Share your own interview experience to unlock all reports, or subscribe for full access.
Sourced from candidate reports and verified by our team.
Topics based on recent interview experiences.
Featured question at Meta
Write a query that returns all neighborhoods that have 0 users.
| Question | |
|---|---|
| 2nd Highest Salary | |
| Comments Histogram | |
| Employee Salaries | |
| Subscription Overlap | |
| Merge Sorted Lists | |
| Experiment Validity | |
| Liked Pages | |
| 500 Cards | |
| Session Difference | |
| Last Transaction | |
| Random SQL Sample | |
| Like Tracker | |
| Weighted Keys | |
| Button AB Test | |
| Search Ratings | |
| Flight Records | |
| Emails Opened | |
| Average Order Value | |
| Largest Salary by Department | |
| Top 3 Users | |
| Swipe Precision | |
| Notification Deliveries | |
| Decreasing Comments | |
| Find the Missing Number | |
| Recurring Character | |
| Scrambled Tickets | |
| Longest Streak Users | |
| Impression Reach | |
| Bank Fraud Model |
Synthesized from candidate reports. Individual experiences may vary.
Candidates report starting with a recruiter call or outreach about their background and which roles fit. One candidate was offered two positions and chose a product data scientist track. Timing between application and the first call varies, and one candidate heard from a recruiter about two months after applying.
Candidates report a roughly 45-minute screen that typically combines SQL with a product or metrics case. Reported SQL used joins, subqueries, CTEs, ratios and window functions on social-product tables, with difficulty ranging from easy-medium to medium-hard. One candidate said CoderPad did not run queries. Product questions covered group calls, notifications and detecting bad content.
In loop-stage reports, candidates describe layered product cases. Examples include whether to launch group calling, why a metric moved, and measuring a content classifier. Follow-ups often probe randomization, network interference, guardrail metrics and trade-offs. Candidates say structure matters, so clarify the goal, choose metrics, then design the test.
Several candidates describe a dedicated SQL round with multiple questions, often involving window functions and joins. Some also report probability and statistics topics such as Bernoulli distributions, Bayes' rule and the Central Limit Theorem. These rounds may be separate from the product cases, depending on the loop and level reported.
Candidates report a behavioral conversation in the loop. Questions included constructive feedback from a stakeholder, project requirements changing midway, managing multiple stakeholders, and the scale and impact of past work. Prepare specific examples. Other candidates described a research-background conversation instead, so the exact format may vary.
