Facebook Interview 2026: SQL Questions, Coding & Response Time

SQL questions at Meta haven’t become easier just because engineers have access to better tooling. In 2026, data science and data engineering candidates still face interview questions modeled on real Facebook users' data: monthly active users, content engagement, ad performance, and cohort retention. The schemas are realistic, and interviewers expect you to reason about both correctness and business implications.

Common themes you’ll encounter include computing engagement metrics per user, identifying power users based on comments and reactions, calculating 7-day or 30-day retention rates, and analyzing ad CTR across segments. These problems test your ability to write clean SQL and explain why your solution matters for product decisions.

SQL Concepts You Must Master

Meta interviewers probe depth on these core SQL skills:

• Joins: Inner, left, and semi-joins across user activity, posts, and engagement tables

• GROUP BY vs HAVING: Aggregating metrics and filtering aggregated results

• Window functions: ROW_NUMBER(), DENSE_RANK(), LAG(), LEAD() for ranking and retention calculations

• Date truncation and partitions: Working with DATE_TRUNC, date arithmetic, and partitioned tables

• Query optimization: Understanding explain plans, indexing strategies, and performance tradeoffs on billion-row tables

A Concrete Example: Retention Query

Here’s a simplified example of a LAG-based retention query you might see:

WITH user_activity AS (

  SELECT 

    user_id,

    DATE_TRUNC('day', activity_date) AS activity_day

  FROM activity_log

  GROUP BY user_id, DATE_TRUNC('day', activity_date)

),

with_prev AS (

  SELECT 

    user_id,

    activity_day,

    LAG(activity_day, 1) OVER (PARTITION BY user_id ORDER BY activity_day) AS prev_day

  FROM user_activity

)

SELECT 

  activity_day,

  COUNT(DISTINCT CASE WHEN activity_day = prev_day + INTERVAL '1 day' THEN user_id END) AS retained_users,

  COUNT(DISTINCT user_id) AS total_users

FROM with_prev

GROUP BY activity_day

ORDER BY activity_day;

When writing this in an interview, you’d talk through the logic: partition by user, order by date, use LAG to find the previous active day, then identify users who returned exactly one day later. At Meta’s scale, you’d also note that this query would need partitioning by date range and potentially approximate counting for performance.

Realistic Schemas to Practice

Meta’s SQL questions often use schemas like:

• user_activity: user_id, activity_type, activity_date, session_id

• ads_events: ad_id, user_id, event_type (impression, click, conversion), event_time

• stories_impressions: story_id, viewer_id, view_time, completion_rate

• reels_engagement: reel_id, user_id, action_type, action_time

Practice writing queries that join multiple tables, handle null values, and compute metrics like integer division for percentages without floating-point errors.