Meta Interview Questions — Page 232

Meta·Data Scientist·Onsite - Multi Round

Jul 2026

Brutal Meta DS onsite focused entirely on experiment design for a Group Call feature with network interference. One long, multi-part question that spiraled into graph theory, sample size math, and spillover bias. Walked out not totally sure how I did.

How would you define the exposure unit and the estimand for an A/B test of a Group Call feature where network interference is expected? For example, should you target intent-to-treat at the cluster level or a per-user average treatment effect?
How would you construct clusters from a social graph for this experiment? Walk through how you'd weight edges for strong ties, choose a clustering rule like connected components after thresholding, and prevent cluster overlap.
If you randomize at the cluster level, how do you handle users who appear on edges between treatment and control clusters? What role do holdout buffers or partial saturation designs play?
What primary metric and two guardrail metrics would you choose for this feature? Critically assess 'time spent per user per day' as a success metric, including its pros, cons, and manipulation risks, and propose at least one alternative primary metric.
Compute the design effect for cluster randomization given an average cluster size of 20 and an ICC of 0.05. How does the design effect change if the average cluster size doubles but the ICC halves?
If you ignored network effects and randomized by individual user instead of cluster, under what conditions would the naive treatment effect estimate be biased downward versus upward? Give intuition for positive and negative spillovers and a real-world example of each.

“This is where I started to feel the ground shift.”

View Post

Meta·Data Scientist·Technical Phone Screen

Jul 2026

Meta DS interview focused entirely on ads ranking experimentation, three parts back to back. The questions were meaty and interconnected, which I liked, but the depth required across all three caught me off guard more than I expected.

A new ad recommendation algorithm has been built and the team believes it performs better. How would you design an experiment to evaluate whether that's actually true?
What unit of randomization would you choose for this ads experiment, and what interference risks does that choice create?
Walk me through how you'd actually run this experiment: ramp plan, duration, sample size, and what checks you'd do before calling a result.
If a 50/50 experiment split isn't feasible, what are two or more alternative approaches you'd consider, and what are the tradeoffs?
If you're not allowed to run any controlled experiment at all, how would you assess whether the new algorithm is better and how would you handle recommendations for cold-start users?

“This is the kind of question where you can go broad fast and lose the thread.”

View Post

Meta·Software Engineer·Technical Phone Screen

Jul 2026

Meta SWE coding round with a friend recommendation problem. Pretty standard randomization + filtering exercise but the follow-up about reproducibility with a seed tripped me up a bit.

Implement a random friend recommendation function that takes a user, a list of all users, an optional count k, and an optional seed. It should return up to k users who are not already friends with the given user and not the user themselves, with no duplicate results.

“The filtering part was easy enough, just exclude the user's own id and anyone already in their friend list, then sample from what's left.”

View Post

Meta·Software Engineer·Onsite - System Design / Architecture

Jul 2026

System design round at Meta for a software engineering role. The whole thing was basically one giant question about building a ticketing platform, and it went deep fast. Lots of ground to cover and I don't think I got through all of it cleanly.

Design a high-concurrency ticket purchasing system for events with seat maps, covering seat selection, holds with expiry, oversell prevention, wait queues, payments with idempotency, refunds, anti-bot measures, audit logs, and real-time availability updates.

“This was the whole interview.”

View Post

Meta·Data Scientist·Onsite - Behavioral / Leadership

Jul 2026

Behavioral round for a Data Science role at Meta. Two questions, both pretty standard inclusion and conflict territory, but Meta tends to push harder on specifics than most places so you can't just give a vague answer and coast.

How do you make teammates feel included, especially someone who is new, quiet, remote, or from a different background than yours?
Tell me about a conflict you had with a coworker or someone on a partner team. What caused it, how did you handle it, and what came out of it?

“I had a decent story about onboarding a remote contractor but I kept second-guessing whether it was specific enough.”

View Post

Meta·Data Scientist·Technical Phone Screen

Jul 2026

A Meta DS interview that went deep into experiment statistics, way deeper than I expected. The whole session was basically one long clustered experiment design problem broken into six parts, and I was sweating through most of it.

Given a baseline exploration rate, a target minimum detectable effect, and cluster-level randomization with a known intra-cluster correlation, derive the required number of users and clusters per arm. Show the formulas including the design effect adjustment.
What exact hypothesis test would you use for the primary metric in a clustered experiment, and when would you prefer a nonparametric alternative?
Define the confidence interval you would report for this experiment and explain how you'd interpret it alongside practical significance thresholds.
You're peeking at the primary metric weekly for four weeks. Design a sequential testing plan and show the adjusted alpha at each look.
You're tracking three guardrail metrics alongside the primary. How do you control for multiple testing while preserving power on the primary, and what's your decision logic for combining primary and guardrail results?
If 10% of control users are accidentally exposed to the treatment, what direction does this bias the intent-to-treat estimate and how would you correct for it?

“The design effect formula itself isn't hard once you remember DE = 1 + (m-1)*ICC, but I fumbled the algebra when converting the inflated variance back to cluster counts.”

View Post

Meta·Software Engineer·Technical Phone Screen

Jul 2026

Meta SWE coding round, two LeetCode problems back to back. Nothing too exotic but the second one definitely required more graph intuition than I expected.

Given a binary tree, find the length of its longest path between any two nodes (the diameter).
Given a binary matrix with exactly two islands, find the minimum number of 0s you need to flip to connect them.

“Felt pretty manageable.”

View Post

Meta·Data Scientist·Technical Phone Screen

Jul 2026

Meta DS interview with a meaty experiment design question that blends tradeoff framing, CLV thinking, mitigation strategy, and post-launch monitoring all in one. Dense question, not a lot of room to breathe.

An experiment shows strong lifts in notification metrics (views, CTR, notification-driven actions) but a meaningful drop in usage of non-notified accounts among multi-account users. How do you frame this tradeoff for a PM and cross-functional partners, estimate longer-term user value and churn risk, propose mitigations and a phased rollout, and design a post-launch monitoring plan?

“This is basically four questions stapled together and they want you to move through all of them without losing the thread.”

View Post

Meta·Software Engineer·Technical Phone Screen

Jul 2026

Meta SWE coding round, one question the whole time: implement Union Find with path compression and rank-based union. Pretty focused session, no fluff.

Implement a Union Find (Disjoint Set) data structure supporting find with path compression and union by rank, then use it to process a sequence of union and find operations.

“I knew Union Find going in but blanked on the exact rank logic mid-implementation.”

View Post

Meta·Data Scientist·Technical Phone Screen

Jul 2026

Meta DS interview, got a pretty involved SQL question for an ads analytics context. One question, lots of moving parts, felt like a take-home but done live.

Given an ad impressions table and a conversions table, write a SQL query that attributes each conversion to the most recent eligible impression within 7 days (same user and ad, impression before conversion). For January 2024 impressions, return per campaign and per day: impression count, distinct users, attributed conversions, conversion rate, total spend, total attributed revenue, and average hours from impression to conversion.

“This one took me a minute to even parse.”

View Post