← Meta Interview Insights

Meta·Data Scientist·Technical Phone Screen·Senior

Senior
Jun 2026

Summary

Brutal Meta DS interview question covering basically every dimension of experiment design you can think of. The kind of question where you realize halfway through your answer that you've already forgotten three of the ten sub-parts.

Questions Asked (1)

Q1

Design a rigorous experiment to evaluate a pinned-unread chat feature. Your answer should address: randomization unit and cross-contamination prevention, eligibility and pre-exposure criteria, primary and secondary metrics plus guardrails, power and sample size given heavy-tailed engagement, test duration and seasonality, novelty effects and ramp plan, network effects in group chats and SUTVA risks, handling users who switch between multiple accounts, instrumentation and data quality checks, and a pre-registered analysis plan including variance reduction and how to handle conflicting metric signals.

A/B Testing & ExperimentationProduct Analytics & MetricsSystem Design
Author's notes

This one is a monster.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Structure your answer around the experiment lifecycle: design (randomization, eligibility, metrics), execution (power, duration, ramp, instrumentation), and analysis (pre-registration, variance reduction, conflict resolution). Emphasize how you mitigate unique challenges like network effects in group chats and heavy-tailed engagement. Conclude with a decision framework for interpreting conflicting signals.

Pro tip: Propose using a cluster-randomized design at the group-chat level to address SUTVA violations, and pre-register a switchback or holdout to validate long-term effects. Show you understand Meta's scale by discussing how to handle multi-account users via a unified user graph and how to detect novelty via pre-period behavior.

1. Define the experiment design

Specify randomization unit (user or group chat), eligibility criteria (e.g., active users with unread messages), and pre-exposure filters to exclude users already exposed to the feature. Address cross-contamination by using cluster randomization for group chats and isolating treatment at the user level for 1:1 chats.

2. Select metrics and guardrails

Choose primary metric (e.g., unread message click-through rate), secondary metrics (e.g., session duration, messages sent), and guardrails (e.g., user retention, notification opt-outs). Ensure metrics are sensitive to the feature and aligned with company goals.

3. Plan power, sample size, and duration

Account for heavy-tailed engagement by using variance reduction techniques (CUPED, stratification) and robust power analysis (e.g., bootstrap or quantile-based). Determine test duration to cover seasonality and novelty effects, with a ramp plan (e.g., 1% -> 5% -> 50%) to monitor early signals.

4. Address network effects and SUTVA

For group chats, use cluster randomization (by group) to prevent spillover. For multi-account users, unify accounts to avoid contamination. Consider switchback designs or holdouts to measure long-term effects and validate SUTVA assumptions.

5. Pre-register analysis and handle conflicts

Pre-register the analysis plan including variance reduction, multiple testing corrections, and decision rules for conflicting metrics (e.g., primary wins but guardrail degrades). Use sequential testing or Bayesian methods to allow early stopping if needed.

Key Points to Mention

  • Randomization unit: user-level for 1:1 chats, group-level for group chats to prevent contamination.
  • Eligibility: users with unread messages in the past X days, excluding those who opted out or are in other experiments.
  • Metrics: primary (unread message CTR), secondary (session time, messages sent), guardrails (retention, notification opt-outs).
  • Power: use CUPED with pre-period engagement to reduce variance; bootstrap for heavy-tailed metrics.
  • Duration: run for at least 2 weeks to capture weekly seasonality; ramp gradually to detect novelty.
  • Network effects: cluster randomization for group chats; unify multi-account users via user graph; consider switchback for SUTVA.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.