Home โ€บ Case Study โ€บ Meta Senior Data Analyst Interview Questions and Answers...
Case Study Company Interview Questions Data Analyst FAANG

Meta Senior Data Analyst Interview Questions and Answers 2026

Meta’s Senior Data Analyst interview is one of the most experimentation-heavy in the industry. You will be expected to go beyond writing correct SQL โ€” interviewers want to see how you think about product metrics, design experiments for a two-sided platform, and diagnose sudden metric drops under ambiguity. This guide covers all four rounds with real questions from 2025โ€“2026 interviews, complete answers, and insider notes on what each round is actually scoring.

www.TheDataMonk.com
Meta
Senior Data Analyst
Product Analytics / Growth & Experimentation
SQL
Python
A/B Testing
Experimentation
Tableau / Looker
Statistics
Salary Range
โ‚น45โ€“75 LPA
5โ€“9 Years Exp
Domain: Product Analytics / Growth
Location: Bangalore
Interview Rounds: 4โ€“5 Rounds
Team: Core Data Science, Growth
Tools: SQL, Python, Tableau, Spark
Hiring Season: Janโ€“Apr, Augโ€“Oct
Interview Process
1
Round 1 ยท 60 min
Advanced SQL

DAU / WAU stickiness

Retention cohorts

Friend graph analysis

Content engagement rate

Medium โ€“ Hard
2
Round 2 ยท 50 min
Product Analytics

North Star for Reels

DAU drop diagnosis

Feed ranking metrics

Metric trade-off analysis

Medium โ€“ Hard
3
Round 3 ยท 50 min
A/B Testing & Stats

Experiment design

Network effects & SUTVA

Statistical significance

Power analysis

Hard
4
Round 4 ยท 45 min
Case Study

Marketplace analytics

New feature metrics

Launch decision framework

Cross-product trade-offs

Hard


Round 1 โ€” Advanced SQL

SQL Coding Round โ€” Product & Growth Focus
Meta SQL questions are always set around user behaviour โ€” DAU, retention, content engagement, friend graphs. You are expected to write clean, scalable SQL and explain the business meaning of every result. Interviewers watch whether you filter, aggregate, and join in the right order.
Questions 5 questions
Duration 60 minutes
Difficulty Medium to Hard
Key Topics Window FunctionsRetention CohortsDAU / WAU / MAUFriend GraphContent Engagement
Q1 โ€” Calculate DAU, WAU, MAU and Stickiness Ratio
Medium12 min

The growth team wants to track daily, weekly, and monthly active users and compute the stickiness ratio (DAU/MAU) for the last 90 days. A user is “active” if they have at least one event in the events table on that day.

Tables:

  • events(user_id, event_type, event_date)

๐ŸŽฏ What the interviewer tests
Whether you use COUNT(DISTINCT user_id) correctly and whether you understand that DAU, WAU, MAU require different time-window aggregations. Many candidates compute DAU correctly but fail on WAU because they use a fixed 7-day window instead of a rolling one. Stickiness = DAU/MAU โ€” a value above 20% is generally considered healthy for a social product.

-- DAU
SELECT
    event_date,
    COUNT(DISTINCT user_id) AS dau
FROM events
WHERE event_date >= CURRENT_DATE - INTERVAL '90 days'
GROUP BY event_date;
-- WAU (rolling 7-day window)
SELECT
    event_date,
    COUNT(DISTINCT user_id) OVER (
        ORDER BY event_date
        RANGE BETWEEN INTERVAL '6 days' PRECEDING AND CURRENT ROW
    ) AS wau
FROM (
    SELECT DISTINCT event_date, user_id FROM events
    WHERE event_date >= CURRENT_DATE - INTERVAL '90 days'
) t;
-- MAU and stickiness
WITH dau_cte AS (
    SELECT event_date, COUNT(DISTINCT user_id) AS dau
    FROM events
    WHERE event_date >= CURRENT_DATE - INTERVAL '90 days'
    GROUP BY event_date
),
mau_cte AS (
    SELECT
        DATE_TRUNC('month', event_date)::DATE AS month,
        COUNT(DISTINCT user_id) AS mau
    FROM events
    WHERE event_date >= CURRENT_DATE - INTERVAL '90 days'
    GROUP BY DATE_TRUNC('month', event_date)
)
SELECT
    d.event_date,
    d.dau,
    m.mau,
    ROUND(100.0 * d.dau / NULLIF(m.mau, 0), 1) AS stickiness_pct
FROM dau_cte d
JOIN mau_cte m
  ON DATE_TRUNC('month', d.event_date) = m.month
ORDER BY d.event_date;
๐Ÿ’ก Key insight
Stickiness above 20% is the benchmark for healthy social products (Facebook historically ~50โ€“60%). If stickiness drops while MAU holds steady, it means existing users are visiting less frequently โ€” a retention signal, not an acquisition problem. Mention this business context to stand out.
๐Ÿ” Common Follow-Up Questions
  • How would you compute a 28-day rolling MAU instead of calendar-month MAU?
  • If DAU drops but MAU is flat, what does that tell you about user behaviour?
  • How would you segment stickiness by user cohort (new vs returning)?

Q2 โ€” 30-Day User Retention Cohort Analysis
Hard18 minalso asked at Google, Snap

Build a retention cohort table: group users by the week they first logged in, then for each subsequent week (Week 0 to Week 4), calculate what percentage of the cohort was still active.

Tables:

  • logins(user_id, login_date)

๐ŸŽฏ What the interviewer tests
This is Meta’s most commonly repeated SQL question. The two steps most candidates get wrong: (1) computing “weeks since first login” using DATEDIFF divided by 7, and (2) using the cohort size (Week 0 count) as the denominator โ€” not the current week count. Week 0 retention is always 100% by definition.

WITH first_login AS (
    SELECT user_id, MIN(login_date) AS cohort_date
    FROM logins
    GROUP BY user_id
),
weekly_activity AS (
    SELECT
        l.user_id,
        DATE_TRUNC('week', f.cohort_date)::DATE    AS cohort_week,
        FLOOR(
            EXTRACT(EPOCH FROM (l.login_date - f.cohort_date))
            / 604800                                -- seconds in 7 days
        )                                           AS weeks_since_start
    FROM logins l
    JOIN first_login f ON l.user_id = f.user_id
    WHERE FLOOR(
        EXTRACT(EPOCH FROM (l.login_date - f.cohort_date)) / 604800
    ) BETWEEN 0 AND 4
),
cohort_size AS (
    SELECT cohort_week, COUNT(DISTINCT user_id) AS total
    FROM weekly_activity WHERE weeks_since_start = 0
    GROUP BY cohort_week
)
SELECT
    wa.cohort_week,
    cs.total           AS cohort_size,
    wa.weeks_since_start,
    COUNT(DISTINCT wa.user_id) AS active_users,
    ROUND(
        100.0 * COUNT(DISTINCT wa.user_id) / cs.total, 1
    ) AS retention_pct
FROM weekly_activity wa
JOIN cohort_size cs ON wa.cohort_week = cs.cohort_week
GROUP BY wa.cohort_week, cs.total, wa.weeks_since_start
ORDER BY wa.cohort_week, wa.weeks_since_start;
๐Ÿ’ก Key insight
Present this as a heatmap in practice (cohort weeks as rows, Week 0โ€“4 as columns, retention % as values). A “smile” shape (high Week 0, sharp drop at Week 1, recovery at Week 4) is typical of notification-driven re-engagement. Mention that Meta uses 28-day retention as its primary product health metric, not 7-day.
๐Ÿ” Common Follow-Up Questions
  • How would you visualise this cohort table as a heatmap? What would the axes be?
  • Which cohort has the best Week-4 retention โ€” what product actions improved it?
  • How would you adapt this query if “active” means posting content, not just logging in?


Top Interview Prep Resources

Used by 12,000+ candidates who cleared DA interviews at product companies.

BESTSELLER

2200 Most Asked Analytics Interview Questions
  • 2,200+ questions across all topics
  • 23 analytics topics covered end-to-end
For those who want to master interviews across all analytics domains โ€” SQL, Python, stats, case studies.
โ‚น1,999
โ‚น7,999

Buy Now

SQL MASTERY

Ace Any SQL Interview โ€” 200 Must-Know Questions
  • 220+ questions, no fluff
  • SQL advanced queries, data models, DBMS most asked topics
For anyone who wants to crack any SQL round โ€” from fresher to advanced level.
โ‚น799
โ‚น2,499

Buy Now

Q3 โ€” Content Engagement Rate by Post Type
Medium12 min

Calculate the engagement rate (likes + comments + shares divided by impressions) per post type (photo, video, reel, story) for the last 30 days. Also show which post type has the highest engagement rate and flag any type below 2%.

Tables:

  • posts(post_id, user_id, post_type, created_at)
  • interactions(interaction_id, post_id, user_id, type, interaction_date) โ€” type is ‘like’, ‘comment’, or ‘share’
  • impressions(post_id, impression_date, impression_count)

WITH interaction_counts AS (
    SELECT
        p.post_type,
        COUNT(i.interaction_id) AS total_interactions
    FROM posts p
    LEFT JOIN interactions i ON p.post_id = i.post_id
    WHERE p.created_at >= CURRENT_DATE - INTERVAL '30 days'
    GROUP BY p.post_type
),
impression_counts AS (
    SELECT
        p.post_type,
        SUM(im.impression_count) AS total_impressions
    FROM posts p
    LEFT JOIN impressions im ON p.post_id = im.post_id
    WHERE p.created_at >= CURRENT_DATE - INTERVAL '30 days'
    GROUP BY p.post_type
)
SELECT
    ic.post_type,
    ic.total_interactions,
    im.total_impressions,
    ROUND(
        100.0 * ic.total_interactions / NULLIF(im.total_impressions, 0), 2
    ) AS engagement_rate_pct,
    CASE
        WHEN 100.0 * ic.total_interactions / NULLIF(im.total_impressions, 0) < 2
        THEN 'Below threshold'
        ELSE 'Healthy'
    END AS status
FROM interaction_counts ic
JOIN impression_counts im ON ic.post_type = im.post_type
ORDER BY engagement_rate_pct DESC;
๐Ÿ’ก Key insight
Reels typically have 3โ€“5x higher engagement rates than static photos on Meta platforms because the algorithm surfaces them more aggressively. If the interviewer asks why reels dominate, connect it to the feed ranking model โ€” time-spent and completion rate are stronger ranking signals for video than for photo, creating a virtuous cycle.
๐Ÿ” Common Follow-Up Questions
  • How would you weight interactions differently โ€” should a share count more than a like?
  • If reels have 5x engagement but photos have 10x more impressions, which is more valuable to Meta?

Q4 โ€” Identify Power Users: Top 10% by Activity Score
Medium12 minalso asked at LinkedIn, Twitter/X

Define an activity score for each user as: (posts ร— 3) + (comments ร— 2) + (likes ร— 1) in the last 28 days. Identify the top 10% of users by this score โ€” these are “power users”. Show their user_id, score, and percentile rank.

Tables:

  • posts(post_id, user_id, created_at)
  • interactions(interaction_id, user_id, post_id, type, interaction_date)

WITH user_scores AS (
    SELECT
        user_id,
        SUM(CASE WHEN activity = 'post'    THEN 3
                 WHEN activity = 'comment' THEN 2
                 WHEN activity = 'like'    THEN 1 END) AS activity_score
    FROM (
        SELECT user_id, 'post'    AS activity FROM posts
        WHERE created_at >= CURRENT_DATE - INTERVAL '28 days'
        UNION ALL
        SELECT user_id, type      AS activity FROM interactions
        WHERE interaction_date >= CURRENT_DATE - INTERVAL '28 days'
          AND type IN ('comment', 'like')
    ) all_activity
    GROUP BY user_id
),
ranked AS (
    SELECT
        user_id,
        activity_score,
        NTILE(10) OVER (ORDER BY activity_score DESC) AS decile,
        ROUND(
            100.0 * RANK() OVER (ORDER BY activity_score DESC)
                  / COUNT(*) OVER (), 1
        ) AS percentile_rank
    FROM user_scores
)
SELECT user_id, activity_score, percentile_rank
FROM ranked
WHERE decile = 1
ORDER BY activity_score DESC;
๐Ÿ” Common Follow-Up Questions
  • How would you track whether power users this month were also power users last month?
  • What product interventions would you test to convert mid-tier users into power users?


Mentorship Program
Stop the Grind. Get Mentored.
3 mock interviews ยท personalised feedback ยท resume review ยท referral support
Starter Bundle
โ‚น2,999
one-time

  • โœ“ 2200 Questions eBook
  • โœ“ Ace Any SQL Interview eBook
  • โœ“ 1 Mock Interview Session
  • โœ“ Resume Review

Get Started

Elite Bundle
โ‚น9,999
one-time

  • โœ“ Everything in Complete
  • โœ“ 3 Mock Interviews
  • โœ“ 1:1 Mentor Calls
  • โœ“ Offer Negotiation Support
  • โœ“ Priority WhatsApp Access

Go Elite


Round 2 โ€” Product Analytics & Metrics

Product Sense & Metrics Design Round
This round tests your product intuition. You will be asked to define metrics for a product, investigate a sudden change in a key metric, or make a launch/no-launch decision based on data. The best answers are structured, show trade-off thinking, and connect data to user behaviour.
Questions 3 questions
Duration 50 minutes
Difficulty Medium to Hard
Key Topics North Star MetricMetric TreeRoot Cause AnalysisTrade-off Analysis
Q1 โ€” Define the North Star Metric for Instagram Reels
Medium14 min

Meta is investing heavily in Reels. Define a North Star Metric (NSM) for Reels, defend your choice against at least two alternatives, and describe the metric tree below it.

๐ŸŽฏ What the interviewer tests
Whether you understand that a good NSM must be (a) measurable, (b) a leading indicator of revenue, and (c) actionable by the product team. Most candidates say “time spent on Reels” โ€” which is fine but not differentiated. Senior-level answers explain why their metric captures creator and consumer value simultaneously.

Strong answer framework:

  • Proposed NSM: Daily Reels Completion Rate โ€” percentage of Reels that users watch to โ‰ฅ80% of length. Captures both content quality and feed relevance simultaneously.
  • Why not Time Spent: Time spent can be inflated by auto-play loops; completion rate requires genuine attention. A user watching 50 Reels to 90% completion is more valuable than watching 200 Reels to 30%.
  • Why not Reels Created: Supply-side metric; doesn’t measure whether consumers find value. Can be gamed by incentivising creator volume regardless of quality.
  • Metric tree: Completion Rate = (Reels shown ร— Watch-through rate) where Watch-through = f(feed ranking quality, creator quality, content diversity).
  • Guardrails: Overall time on platform (don’t cannibalise Stories/Feed), creator churn rate (supply health), share of Reels with negative feedback.
๐Ÿ” Common Follow-Up Questions
  • If Reels completion rate goes up but overall Instagram DAU drops, is that a success?
  • How would the NSM change if Meta wanted to monetise Reels vs grow creator supply?

Q2 โ€” Facebook DAU Dropped 5% Last Monday. Diagnose.
Hard20 minalso asked at Snap, Pinterest, LinkedIn

DAU fell from 2.1 billion to 2.0 billion between Sunday and Monday with no planned deployments. Walk through a structured diagnostic process.

MECE Diagnostic Framework โ€” DAU Drop
Data / Measurement Issue
Logging pipeline failure ยท Instrumentation bug ยท Timezone miscalculation ยท Dashboard filter error ยท A/B test traffic splitting side effect
Platform / Technical Issue
App crash on specific OS version ยท Auth / login failure ยท Slow load times increasing bounce ยท Silent background code deploy ยท CDN / server outage in specific region
External / Seasonal Factors
Monday morning pattern (Monday is naturally lower than Sunday) ยท Competitor launch ยท Major news event reducing screen time ยท Regulatory action in key market ยท Holiday in top country
Product / Algorithm Change
Feed ranking model update reducing session triggers ยท Notification frequency change ยท Content policy enforcement wave ยท New privacy settings reducing data personalisation

Diagnostic steps in order:

  1. Rule out measurement error first โ€” check if other metrics (MAU, sessions) dropped equally. If only DAU dropped, suspect logging.
  2. Segment by geography โ€” global drop vs. one country. A single country points to a regional issue (regulation, competitor, outage).
  3. Segment by platform โ€” iOS vs Android vs web. A platform-specific drop points to a client-side bug or OS update incompatibility.
  4. Check the Monday baseline โ€” DAU is structurally lower on Mondays vs Sundays. Compare to prior Monday, not prior day.
  5. Review deployment logs โ€” even without planned deploys, check for config changes, ML model refreshes, or content policy updates.
  6. Correlate with session quality โ€” if users who did log in had shorter sessions, it’s a product/content issue, not an access issue.
๐Ÿ’ก Key insight
The single most common mistake in this question is jumping to a hypothesis before eliminating measurement error. Always validate the data before investigating the cause. In Meta’s actual data stack, a broken Hive pipeline or a misconfigured Scuba query has caused “DAU drops” that turned out to be entirely artificial. Say this explicitly โ€” interviewers love it.


Round 3 โ€” A/B Testing & Experimentation

Experimentation Design & Statistical Analysis
Meta runs thousands of experiments simultaneously. This round tests whether you can design a rigorous experiment, handle violations like network effects and novelty bias, interpret statistical results correctly, and make a confident launch recommendation even under uncertainty.
Questions 3 questions
Duration 50 minutes
Difficulty Hard
Key Topics Experiment DesignNetwork Effects / SUTVAPower Analysisp-value Interpretation
Q1 โ€” Design an Experiment for a New Feed Ranking Algorithm
Hard22 min

Meta’s feed team wants to test a new ranking model that prioritises content from close friends. Design the full experiment โ€” unit of randomisation, metrics, duration, and how you’d handle network effects.

๐ŸŽฏ What the interviewer tests
The key challenge here is that feed ranking experiments on a social network violate SUTVA โ€” treating one user’s feed changes what content creators produce, which changes what other users see. Senior candidates are expected to identify this, explain why user-level A/B testing is insufficient, and propose a cluster-based randomisation instead.

Complete experiment design:

  • Hypothesis: Prioritising close-friend content increases 28-day retention and reduces “negative feedback” (hide, unfollow) by โ‰ฅ5%.
  • Unit of randomisation: Not individual users โ€” use graph clusters (ego networks). Users whose close friends are in the same cluster are randomised together to prevent cross-contamination. Alternatively, use geo-based holdout markets.
  • Primary metric: 28-day retention (% of users still active after 28 days).
  • Secondary metrics: Daily session length, posts created (creator-side), negative feedback rate (hide/unfollow), Reels watch time (guardrail โ€” don’t sacrifice video engagement).
  • Duration: Minimum 4 weeks to capture full novelty bias decay. Feed algorithm changes often show a spike in Week 1 (novelty) that normalises by Week 3โ€“4.
  • Sample size: Run a power calculation with ฮฑ=0.05, power=0.80, and MDE of 0.5% on 28-day retention. Given Meta’s scale, this likely requires only 1โ€“2% traffic allocation.
  • Risks to flag: Creator supply-side effects (creators optimise content for close friends โ†’ less broad reach), advertiser impact (if feed changes reduce ad inventory viewed), novelty bias inflating early results.
๐Ÿ” Common Follow-Up Questions
  • The experiment shows +2% retention but -3% ad revenue. Do you ship it?
  • How would you detect if novelty bias is inflating Week 1 results?
  • What is the difference between a holdout experiment and a standard A/B test?

Q2 โ€” Interpret A/B Test Results: Conflicting Metrics
Hard16 minalso asked at Airbnb, Uber

An experiment to reduce notification frequency shows: DAU +1.2% (p=0.03), session length -2.1% (p=0.01), 7-day retention +0.8% (p=0.09), unsubscribe rate -15% (p<0.001). Should you ship?

๐ŸŽฏ What the interviewer tests
There is no single correct answer โ€” this is a trade-off question. The interviewer wants to see you reason about which metrics matter most for the stated goal, acknowledge the non-significant retention result honestly, and make a directional recommendation rather than hiding behind ambiguity. Saying “I need more data” without a clear direction is a red flag at the senior level.

Analysis framework:

  • What the data says: Users are visiting more often (DAU +) but spending less time per visit (session -). Fewer are unsubscribing from notifications (strong signal of reduced notification fatigue). 7-day retention improvement is directionally positive but not significant at 95%.
  • Reconciling DAU up + session length down: This is consistent with a “healthier habit” โ€” users check the app briefly and intentionally rather than being repeatedly pulled in by notifications. Net engagement (DAU ร— session length) is roughly flat or slightly positive.
  • Decision: Ship, with monitoring. The unsubscribe rate drop (-15%, highly significant) is a strong long-term retention signal โ€” users who unsubscribe from notifications churn at 3x the rate of subscribed users. Prioritise long-term retention over short-term session length.
  • Guardrail check: Confirm ad impressions per DAU did not drop more than 1% (session shortening could hurt ad revenue). If it did, consider a partial rollout.


Round 4 โ€” Case Study & Business Strategy

Business Case & Cross-Product Strategy Round
The final round evaluates whether you can think at the product strategy level. You will be given an ambiguous business problem โ€” a new feature launch, a cross-product trade-off, or a market entry decision โ€” and expected to structure a data-driven recommendation end to end.
Questions 1 case study
Duration 45 minutes
Difficulty Hard
Key Topics Launch DecisionMarket SizingCross-Product Trade-offsMetric Design
Case โ€” Should Meta Launch a Paid Verification Tier in India?
Hard40 min

Meta is considering launching Meta Verified (paid blue checkmark) in India at โ‚น699/month. Evaluate the opportunity, size the market, define the success metrics, and recommend whether to launch.

MECE Framework โ€” Meta Verified India Launch
Revenue Opportunity
India MAU ~400M. Verified target = creators + SMBs = ~5% = 20M potential. 1% conversion = 200K subscribers ร— โ‚น699/mo = โ‚น140 Cr ARR. Conservative but meaningful at Meta’s India cost base.
Strategic Value
Creator monetisation flywheel โ€” verified creators produce more content โ†’ more engagement โ†’ more ad revenue. Differentiates from YouTube (free verification via 100K subs). Direct subscription revenue reduces ad dependency.
Risks
Trust erosion โ€” users associate blue tick with authenticity, not payment. Price sensitivity in India (โ‚น699 is high vs local ARPU). Potential backlash from unpaid verified accounts (journalists, politicians). Competitor (YouTube, ShareChat) may undercut.
Success Metrics
Subscriber count growth rate ยท Revenue per verified user ยท Verified creator content output vs unverified ยท Organic user trust score (survey) ยท Churn rate at Month 3 ยท Cross-sell to Meta Business Suite

Recommendation: Launch in a limited pilot โ€” target the top 0.1% of creators (โ‰ฅ50K followers) with a 3-month free trial, then convert to paid. This builds supply-side proof before mass rollout. Monitor trust perception scores and creator content output weekly. Full launch decision at Month 4 based on pilot data.


4-Week Plan to Crack Any Product Company Interview

This plan works for Meta, Google, Amazon, Airbnb โ€” any product-based company. It covers ~950 questions across two eBooks and includes three mock interviews. No shortcuts, no cramming.

Week 1
Foundation

Build your base โ€” SQL, Python, case study frameworks

Master window functions, CTEs, cohort logic. Revise pandas basics, hypothesis testing. Learn the MECE framework for case studies. Read The Data Monk’s company-specific interview pattern posts to understand what is being asked before you prepare.

Week 2
Question Bank

2200 Most Asked Analytics Interview Questions โ€” deep dive

Cover the SQL, Python, and case study sections of the 2200 questions eBook. Focus on company-specific question sets. The best way to prepare for an interview is to see exactly what gets asked โ€” this eBook gives you that edge.

Week 2
End
Book your first mock interview with The Data Monk. Get real feedback on your SQL approach, case study structure, and communication. โ†’ Book on Topmate
Week 3
SQL Mastery

Ace Any SQL Interview โ€” end to end, no excuses

Pick up the Ace Any SQL Interview eBook and work through every question. Hard window function problems, optimisation, and real company SQL patterns. A second mock interview at the end of this week.

Week 3
End
Second mock interview โ€” focus specifically on SQL and experimentation. โ†’ Book on Topmate
Week 4
Final Polish

Revise first 5 chapters of the 2200 questions eBook + timed practice

Revisit the first five chapters of the main eBook under timed conditions โ€” 15 minutes per SQL question, 20 minutes per case study. Simulate real interview pressure. A final mock interview to close out your preparation.

Week 4
End
Final mock interview โ€” full simulation with feedback across all rounds. โ†’ Book on Topmate
In 4 weeks you will have covered ~950 questions across two eBooks and had 3 mock interviews with real feedback. That is more preparation than 95% of candidates who walk into these interviews. โ€” Or skip the individual steps and take up our mentorship program below.

Scroll to Top