OfferLoop
Sign inOpen app

INTERVIEW GUIDE

Meta Data Scientist Interview: Questions & Process

Meta's Data Scientist (Analytics) interview is product-analytics heavy, not ML-research heavy. Expect a technical screen with SQL and analysis, then an onsite covering SQL/coding, applied statistics and A/B testing, product-sense and metrics, and behavioral. They want someone who can define the right metric, write the query, and turn data into a product decision.

The interview process

1. Recruiter screen — ~30 min call
Tests: background, the analytics-vs-ML distinction, team fit, and timeline
2. Technical screen — 45-60 min
Tests: live SQL plus a short analytical/product question — joins, aggregation, and translating a vague ask into a query
3. Analytical execution — 45 min
Tests: applied statistics and experimentation: A/B test design, sample size, significance, and interpreting noisy results
4. Analytical reasoning / product sense — 45 min
Tests: defining and choosing metrics, diagnosing a metric drop, and recommending a product action
5. Behavioral — 45 min
Tests: cross-functional collaboration, influencing decisions with data, and handling ambiguity

Questions you're likely to get

Technical

  • Given a users table and a logins table, write SQL for daily active users over the last 30 days.
  • Find the second-highest engagement score per region using window functions.
  • Compute next-day retention from an events table.
  • How would you estimate the probability that two random users share at least one friend?
  • Explain p-values, statistical power, and what increasing sample size does to each.

Role-specific

  • Daily active users dropped 5% week over week. How do you find out why?
  • What metric would you use to measure the success of Reels, and why?
  • Design an A/B test for a new 'save post' button. What's your primary metric and guardrails?
  • A test shows a 1% lift with p=0.06. Do you ship it? Walk me through your thinking.
  • How would you detect whether a spike in messages sent is real or a logging bug?

Behavioral

  • Tell me about a time your analysis changed a product or business decision.
  • Describe a time you had to push back on a stakeholder using data.
  • Tell me about a time you worked with incomplete or messy data.

Practice these problems live

Relevant LeetCode problems for the Meta Data Scientist loop. Start a live, AI-run coding interview on any of them — or open the problem on LeetCode.

▶ Combine Two Tables· Easy · LeetCode #175
Combine Two Tables checks the basic LEFT JOIN instinct every Meta SQL screen assumes.
▶ Employees Earning More Than Their Managers· Easy · LeetCode #181
Employees Earning More Than Their Managers tests self-joins, common in metric queries.
▶ Game Play Analysis IV· Medium · LeetCode #550
Game Play Analysis IV is the canonical next-day retention query Meta-style problems mirror.
▶ Immediate Food Delivery II· Medium · LeetCode #1174
Immediate Food Delivery II tests conditional rates and aggregation like real funnel metrics.
▶ Department Top Three Salaries· Hard · LeetCode #185
Department Top Three Salaries forces window functions and ranking under pressure.
▶ Two Sum· Easy · LeetCode #1
Two Sum covers the light Python/DSA fluency Meta sometimes probes alongside SQL.

How to answer (worked examples)

Daily active users dropped 5% week over week. How do you find out why?
Structure before you guess. First confirm it's real (logging, seasonality, holidays). Then segment: by platform, geography, new vs returning, app version, and date of the break. Form a hypothesis from where the drop concentrates (e.g., one OS after a release), then validate with a query. Close with the decision: escalate a bug, or explain an expected shift. Meta rewards a methodical funnel over a lucky guess.
Design an A/B test for a new 'save post' button.
Name the hypothesis and the primary metric (e.g., saves per DAU, or downstream retention), not just clicks. Define randomization unit, guardrail metrics (latency, time spent, reports), sample size and runtime for adequate power, and what result would make you ship, iterate, or kill. Mention novelty effects and network interference. Meta's experimentation bar is high, so showing you protect against false positives is the signal.

What Meta looks for

FAQ

Is Meta Data Scientist a machine learning role?

Mostly no. The Data Scientist (Analytics) track is product analytics — SQL, experimentation, and metrics. ML modeling lives more in the Machine Learning Engineer and Research Scientist tracks.

How important is SQL?

Very. You'll write live SQL in the screen and likely onsite. Be fast and correct with joins, window functions, and date logic — it's a gate, not a nice-to-have.

How much statistics do I need?

Solid applied stats: hypothesis testing, p-values, power, confidence intervals, and A/B test design. You don't need measure theory — you need to reason correctly about experiments.

What is the product-sense round really testing?

Whether you can define the right metric, diagnose a change, and connect data to a product decision. Practice talking through metric trees and root-cause analysis out loud.

How long is the process?

Typically 4-6 weeks from screen to offer, depending on scheduling and team matching.

Sound like a data scientist, not a query bot

Meta scores how you reason about metrics and experiments out loud, not just whether your SQL runs. Rehearse the product-sense and A/B test rounds with OfferLoop's realtime voice coach before the loop.

Practice this interview out loud →

Related

All Meta interviews →Meta Software Engineer interview →Meta Data Analyst interview →Meta Data Engineer interview →Meta Machine Learning Engineer interview →

OfferLoop is an independent interview-practice tool and is not affiliated with, endorsed by, or sponsored by Meta. All company names and trademarks are the property of their respective owners.

Interview formats vary by team, level and year, and this guide is compiled from general knowledge of publicly discussed hiring processes — treat it as preparation material, not an official description of Meta's current process.