Meta Interview Questions — Page 70

Meta·Data Scientist·Technical Phone Screen

May 2026

SQL-heavy technical screen for a Data Scientist role at Meta, focused on user engagement metrics using call and user profile tables. Two questions, both variations on the same join-and-aggregate theme. Nothing too wild but the devil is in the details with the date filtering.

Using a calls table and a user profile table, write a query to find what percentage of French users participated in at least one video call the previous day.
Write a query to calculate total video call duration divided by the number of daily active users in the United States for today.

“The join is straightforward but I tripped up on the DISTINCT count part.”

View Post

Meta·Data Scientist·Technical Phone Screen

May 2026

Meta DS technical screen, two SQL questions back to back on a shared schema. Both involved deduplication logic which is where things got tricky for me.

Given a views table with possible duplicate same-day rows per viewer, write a SQL query that returns the count of distinct posts that had at least one view from an unconnected user lasting more than 60 seconds, within a specified 7-day window.
For the same 7-day window, compute the average reactions per post for Friend-labeled viewers and Unconnected-labeled viewers separately. Reactions should be attributed to a view using a join on post_id, viewer_id, and date, and same-day duplicate views should be resolved by keeping the longest duration. The denominator should be distinct posts with at least one view from that group, including posts with zero reactions.

“The dedup part is what caught me.”

View Post

Meta·Data Scientist·Onsite - Product Sense / Strategy

May 2026

Meta Data Scientist onsite that was essentially a product pitch exercise disguised as a data role interview. One big open-ended question, no SQL, no stats, just 'convince a skeptical engineering team to build this feature' and you have five minutes.

You believe a feature that pins conversations with many unread messages to the top of a user's inbox will improve responsiveness. In roughly 3-5 minutes, pitch a skeptical engineering team on building and launching it. Your pitch needs to cover: the problem with baseline metrics, why pinning mechanistically helps and what order-of-magnitude impact you'd expect, how you'd handle engineering concerns like ranking complexity, real-time updates, caching, client performance, edge cases for users with thousands of chats, and accessibility, a rollout plan from flag to kill-switch, success metrics and latency/error guardrails, how you'd detect and handle regressions for power users versus low-unread users, and finally ownership, effort estimate, and the minimum viable slice you'd ship first.

“This was a lot to hold in your head at once.”

View Post

Meta·Data Scientist·Onsite - Product Sense / Strategy

May 2026

Brutal product case for a DS role at Meta. Five interconnected parts, all expecting you to think like a PM, an economist, and an experimentation researcher simultaneously. The depth required was genuinely surprising.

You're evaluating a Venmo-style payments feature inside a large messaging app. What's the single most important business objective for the first 90 days, and what 2-3 concrete metrics would you use to track it? At least one metric should act as a guardrail protecting core messaging health.
Propose two different monetization models for this payments feature and explain how you'd measure their causal impact on ARPU without the numbers being inflated by promotional subsidies.
Design a beta rollout for this feature that accounts for network interference and leakage. What's your randomization unit, how do you define exposure, what happens when a non-assigned user tries to access the feature through an assigned friend, and how do you estimate intent-to-treat vs treatment-on-treated effects?
If a randomized experiment isn't feasible, write out a difference-in-differences specification for evaluating this feature. Define your treatment and control groups, the pre and post windows, what you're actually estimating, what parallel trends checks you'd run, how you'd cluster your standard errors, and name two falsification tests.
What are the top three risks for launching this feature across product, fraud and regulatory, and ecosystem dimensions? For each, name a measurable leading indicator and a mitigation you'd ship before full launch.

“I went straight to transaction volume and stumbled when they pushed back asking me to be more precise about units and definitions.”

View Post

Meta·Data Scientist·Technical Phone Screen

May 2026

Meta DS interview focused entirely on product analytics for a video platform. Two meaty open-ended problems back to back, no coding, just a lot of 'how would you think about this' with a PM framing the whole thing.

How would you build a metric framework for 'engagement' on a video platform, including what you'd optimize, what you'd use to diagnose movement, and what guardrails you'd put in place?
Comment volume tends to follow a heavy-tailed distribution. How would you analyze that distribution, monitor it over time, and design an experiment to increase healthy commenting?

“This felt broad in a way that made me nervous.”

View Post

Meta·Product Manager·Onsite - Product Sense / Strategy

May 2026

Meta PM interview with a product strategy case built around Zoom for Business. Pretty conceptual, no data given, just had to reason through why the product exists and what success looks like.

You're a PM at Zoom for Business. Why does this product exist, and what would you define as its north star metric?
What specific goals would you set for your team if you were leading this product?

“I spent too long on the 'why does it exist' part and kind of rushed the north star.”

View Post

Meta·Data Scientist·Technical Phone Screen

May 2026

Meta Data Scientist interview with a SQL-heavy technical screen. One question, three tables, and a deadline that made me second-guess every JOIN I've ever written.

Given three related tables for ads, impressions, and conversions, write a SQL query that returns the top 5 ads by conversion count over the past 30 days. Output should include ad_id, advertiser_id, and total_conversions, sorted descending, with ads that have zero conversions excluded.

“The LEFT JOIN direction tripped me up for a second.”

View Post

Meta·Technical Program Manager·Onsite - System Design / Architecture

May 2026

Meta TPM system design round, one question about building a video sharing platform with some kind of authenticity verification layer. Pretty open-ended and I wasn't totally sure how deep to go on the trust/integrity side versus the core sharing infrastructure.

Design a video sharing application that includes mechanisms to verify whether uploaded videos are authentic.

“I spent the first chunk of time on the standard stuff, upload pipeline, storage, CDN, and kind of forgot the whole point was the authenticity angle.”

View Post

Meta·Data Scientist·Technical Phone Screen

May 2026

Meta DS interview with a pretty involved SQL problem covering multi-account user logic and notification analytics. Two-part question, both parts required in a single session. The schema wasn't trivial and the edge cases were the whole point.

Given a notifications table and a daily snapshot table mapping persons to accounts, write a single SQL query returning three counts in one row: how many people have exactly two active accounts last month, exactly three, and more than three. Active means last_action_date falls within August 2025, using only the most recent snapshot date.
Among people with two or more active accounts in August 2025, what percentage have at least one unread notification on any of their accounts during that same period? Return a single value rounded to two decimal places, and handle the case where the denominator could be zero.

“The tricky bit is that one person can own multiple userids, so you have to group by person_id first to count active accounts per person, then pivot that into three buckets in one row using conditional aggregation.”

View Post

Meta·Software Engineer·Technical Phone Screen

May 2026

Meta SWE coding round, one question the whole time, sparse vector dot product. Pretty clean problem if you've seen it before but there's a real design decision buried in it that they want you to talk through.

Given two sparse vectors of equal length where most values are zero, design a data structure and write an efficient method to compute their dot product. Be ready to discuss time and space complexity.

“The naive approach is obvious but they're not asking for that.”

View Post