Meta Analytics Engineer Interview Questions (2026)

1. 1Recruiter screen30 minEasy

   Process overview, level calibration, and prep guidance — Meta recruiters actively coach on round formats.

2. 2Technical screen45 minHard

   Two DSA problems in 45 minutes on a plain shared editor with no autocomplete or execution.

3. 3Coding round ('Ninja')45 minHard

   Two more problems at loop difficulty; clean near-compilable code and verbalized complexity analysis expected.

4. 4System design ('Pirate')45 minHard

   Design a Meta-scale product system (feed, Stories, chat) with emphasis on read-heavy fan-out, caching, and data modeling.

5. 5Behavioral ('Jedi')45 minMedium

   Deep past-experience discussion on conflict, growth, and impact aligned to Meta values; graded as a real signal round.

1. Q1

   Design an A/B test for a new Facebook Feed ranking or recommendation change. Define hypothesis, primary metric, guardrails, randomization unit, and launch decision rule

   MediumStatistics & Experimentation RoundA/B TestingMeta-specific

   Context: Context: Meta wants to balance engagement, creator health, and advertiser value.

   How to answer: A strong answer will define a clear, directional hypothesis, such as 'The new ranking algorithm will increase user engagement (e.g., time spent) without negatively impacting user well-being.' The primary metric should be a key engagement or retention metric like 'Total Time Spent' or 'DAU/MAU ratio.' Guardrail metrics are crucial and include 'Negative Feedback Rate,' 'User Reported Problems,' and 'Friendship Removal Rate' to detect adverse effects. The randomization unit should be 'User ID' to ensure consistent experience, and the launch decision rule should involve statistically significant positive movement on the primary metric, no significant negative movement on guardrails, and a pre-defined minimum effect size.

2. Q2

   For Instagram Reels, should randomization happen at user, session, device, advertiser, or country level? Explain the tradeoffs

   MediumStatistics & Experimentation RoundA/B TestingMeta-specific

   Context: Consider cross-device behavior, interference, marketplace effects, and operational feasibility.

   How to answer: For Instagram Reels, randomization should primarily happen at the user level. This ensures that a user consistently experiences either the control or treatment group, preventing contamination and interaction effects that could arise from switching between groups. Randomizing at a higher level like country or advertiser would reduce statistical power and make it difficult to detect smaller effects, while lower levels like session or device could lead to a single user experiencing multiple treatments, invalidating the experiment's results. The tradeoffs involve balancing statistical power, preventing contamination, and ensuring a consistent user experience.

3. Q3

   Choose primary and guardrail metrics for a WhatsApp Business experiment aimed at improving meaningful social interaction rate. What metrics would prevent a harmful launch?

   MediumStatistics & Experimentation RoundA/B TestingMeta-specific

   Context: Include user experience, partner health, revenue, reliability, and long-term retention considerations.

   How to answer: A strong candidate would identify 'Meaningful Social Interaction Rate' (MSI Rate) as the primary metric, defined as the proportion of active users sending or receiving messages with non-business contacts. Guardrail metrics should focus on preventing negative impacts on user experience and platform health. Key guardrails include 'Total Messages Sent/Received' (overall engagement), 'User Retention' (e.g., D7, D30), 'Block/Report Rate' (user dissatisfaction), and 'App Crashes/Errors' (technical stability). The candidate should explain why each guardrail is important for preventing a harmful launch, such as ensuring the experiment doesn't boost MSI at the expense of overall usage or user safety.

4. Q4

   During a Marketplace experiment, the treatment/control split is 52/48 instead of 50/50. How would you diagnose sample ratio mismatch?

   MediumStatistics & Experimentation RoundA/B TestingMeta-specific

   Context: Assume assignment logs, exposure logs, and eligibility filters may disagree.

   How to answer: To diagnose sample ratio mismatch (SRM), I would first verify the randomization unit and ensure consistent logging across treatment and control. Next, I'd perform a chi-squared test or z-test on the observed counts of the randomization unit (e.g., users, sessions) in treatment vs. control to statistically assess the deviation from the expected 50/50 split. If the p-value is below a significance threshold (e.g., 0.05), it indicates a statistically significant SRM. Finally, I would investigate potential causes such as implementation bugs in the randomizer, caching issues, or external system dependencies affecting assignment or logging.

5. Q5

   The Stories experiment is trending positive after two days. A PM wants to stop early and launch. How do you handle peeking and sequential testing?

   MediumStatistics & Experimentation RoundA/B TestingMeta-specific

   Context: Discuss pre-specified stopping rules, alpha spending, business urgency, and risk.

   How to answer: A strong candidate would first explain the fundamental problem of peeking: it inflates Type I error rates, leading to false positives. They would then discuss sequential testing methodologies like Always Valid Inference (AVI) or using an O'Brien-Fleming boundary to adjust p-values or confidence intervals for continuous monitoring. The answer should also cover practical considerations at Meta, such as pre-registering experiment duration, minimum detectable effect (MDE), and the importance of statistical power, while emphasizing the need for robust decision-making over speed.

6. Q6

   A new Ads Manager feature shows a large week-1 lift in meaningful social interaction rate, but the effect fades by week 4. What could explain this and how would you design the test duration?

   MediumStatistics & Experimentation RoundA/B TestingMeta-specific

   Context: Discuss novelty, learning effects, seasonality, and durable impact.

   How to answer: The fading effect could be due to novelty effect, where users initially engage more with a new feature but revert to baseline as the novelty wears off. Another explanation is selection bias, if early adopters are more engaged users who eventually normalize. To design the test duration, I would consider the feature's expected user lifecycle and the time needed for user behavior to stabilize, typically 4-8 weeks, while also monitoring for novelty effects. I would also analyze user cohorts based on their adoption time to differentiate between novelty and true long-term impact.

7. Q7

   In a marketplace-like Facebook Feed feature, treatment users may affect control users. How would network effects or interference bias the experiment?

   MediumStatistics & Experimentation RoundA/B TestingMeta-specific

   Context: Examples include advertiser supply, content inventory, delivery capacity, or pricing pressure.

   How to answer: Network effects, or 'interference,' in an A/B test occur when the treatment group's actions influence the control group's experience, or vice-versa. This typically biases the experiment towards a null result, underestimating the true treatment effect. Specifically, if treatment users generate more content or engagement that control users can see, the control group benefits from the treatment without being exposed to its direct mechanics. Conversely, if treatment users 'steal' engagement from control users, the control group might appear worse off than it truly is, potentially overestimating the treatment effect. This invalidates the core assumption of independent user experiences between groups.

8. Q8

   meaningful social interaction rate is a low-frequency event for Instagram Reels. How would you set up an experiment with enough power without waiting too long?

   MediumStatistics & Experimentation RoundA/B TestingMeta-specific

   Context: Discuss proxy metrics, variance reduction, larger samples, longer windows, and risk of metric gaming.

   How to answer: To address the low-frequency nature of meaningful social interactions (MSI) on Instagram Reels, a strong candidate would propose using a proxy metric that is highly correlated with MSI but occurs at a much higher frequency. This proxy metric should be sensitive to the changes being tested and available within a shorter timeframe. Additionally, they might suggest increasing the sample size significantly, running the experiment for a longer duration if feasible, or considering a more sensitive experimental design like a switchback or matched-pair design if applicable to Reels. Finally, they should discuss power analysis to quantify the required sample size and duration for both the proxy and the ultimate MSI metric.

9. Q9

   Design a geo or country-level experiment for WhatsApp Business. When is this better than user-level randomization, and what are the analytical downsides?

   MediumStatistics & Experimentation RoundA/B TestingMeta-specific

   Context: Use matched markets, pre-period balancing, spillover checks, and fewer experimental units.

   How to answer: A strong candidate would design a geo-level experiment by defining distinct geographic units (e.g., countries, states, or DMAs) as the unit of randomization, then randomly assigning these units to treatment or control groups. This approach is superior to user-level randomization when there are network effects, spillover effects, or when the intervention itself is inherently geo-specific (e.g., a regulatory change or a marketing campaign targeting a region). However, geo-level experiments suffer from lower statistical power due to fewer experimental units, increased risk of selection bias if geo-units are not truly comparable, and challenges in controlling for confounding variables that vary by region.

10. Q10

    The Marketplace experiment lifts meaningful social interaction rate overall, but only for new users and only in one country. How would you evaluate heterogeneous treatment effects?

    HardStatistics & Experimentation RoundA/B TestingMeta-specific

    Context: Balance pre-planned segments with exploratory slicing and multiple testing risk.

    How to answer: A strong candidate would first acknowledge the initial finding and the need for subgroup analysis to understand the heterogeneous treatment effects. They would then propose specific statistical methods like interaction terms in regression models (e.g., OLS or logistic regression depending on the metric) to quantify the effect modification by user tenure and country. The candidate should also discuss the importance of multiple comparisons correction (e.g., Bonferroni, FDR) when analyzing many subgroups and consider power implications for smaller segments. Finally, they would suggest visualizing these effects (e.g., forest plots) and potentially exploring causal mechanisms for the observed heterogeneity.

11. Q11

    Treatment improves meaningful social interaction rate but worsens feed load time for Stories. Walk through a launch recommendation

    HardStatistics & Experimentation RoundA/B TestingMeta-specific

    Context: Make a decision under conflicting metrics and quantify tradeoffs for stakeholders.

    How to answer: A strong recommendation would involve a phased rollout strategy, starting with a small-scale experiment to validate the positive impact on meaningful social interaction (MSI) and quantify the negative impact on feed load time (FLT) more precisely. This would be followed by a deep dive into user segments to identify if certain groups are disproportionately affected by FLT or benefit more from MSI, potentially leading to a targeted rollout. Finally, the recommendation should include a plan for iterating on the treatment, exploring technical optimizations to mitigate FLT while preserving MSI gains, and defining clear success metrics and a rollback plan.

12. Q12

    How would you design ramp-up, holdback, and post-launch monitoring for a successful Ads Manager A/B test?

    HardStatistics & Experimentation RoundA/B TestingMeta-specific

    Context: Include ramp stages, persistent holdback, alert thresholds, rollback criteria, and owner accountability.

    How to answer: Design ramp-up by gradually increasing experiment exposure (e.g., 1%, 5%, 10%) while monitoring key health metrics (e.g., error rates, latency, CPU usage) for anomalies before each stage. Implement a holdback group (e.g., 1-5% of total eligible population) that remains on the old experience indefinitely, allowing for long-term impact assessment and quick rollback if unforeseen issues arise. Post-launch monitoring involves establishing a dashboard with both experiment metrics (e.g., CTR, CPA, ROAS) and system health metrics, setting up automated alerts for significant deviations, and defining clear rollback criteria and procedures.

13. Q13

    Midway through the Facebook Feed test, tracking for WhatsApp Business changed. How would you decide whether the experiment results are still usable?

    HardStatistics & Experimentation RoundA/B TestingMeta-specific

    Context: Compare instrumentation versions, affected traffic share, raw logs, and sensitivity analyses.

    How to answer: First, I would assess the nature of the change to WhatsApp Business tracking: was it a bug fix, a new feature, or a complete overhaul? Then, I would analyze the timing of the change relative to the experiment's start and the observed impact on key metrics for both control and treatment groups. I'd specifically look for any differential impact on the experimental groups that correlates with the tracking change. If the change was minor and impacted both groups equally, the results might still be usable; otherwise, a re-run or advanced statistical adjustments would be necessary.

14. Q14

    Two overlapping experiments on Instagram Reels both affect ad CTR and revenue per user. How would you detect and manage interaction effects?

    HardStatistics & Experimentation RoundA/B TestingMeta-specific

    Context: Discuss experiment registry, factorial design, exclusion rules, and interaction terms.

    How to answer: A strong candidate would first emphasize the importance of pre-analysis, including checking for known interactions and defining primary metrics. They would then propose statistical methods like ANCOVA or regression analysis with interaction terms to detect significant interaction effects on ad CTR and revenue per user. For management, they would suggest strategies such as sequential rollout, re-randomization, or creating a combined treatment group to directly measure the interaction. Finally, they would discuss the trade-offs of each management strategy regarding statistical power and time to insight.

15. Q15

    Meta's Facebook Feed revenue suddenly drops 10% week over week. Structure a business case to diagnose the issue and identify the most likely drivers

    MediumProduct Analytics & Business CaseBusiness CasesMeta-specific

    Context: Consider traffic, conversion, pricing, mix, supply/inventory, outages, marketing, and seasonality.

    How to answer: A strong business case for a 10% WoW revenue drop in Facebook Feed would start with clarifying questions about the scope and recent changes. The diagnosis would then follow a structured approach, segmenting revenue by key dimensions like geography, device, ad type, and user demographics to pinpoint the largest contributing segments. Concurrently, investigate potential internal system issues (e.g., data pipeline, ad delivery bugs) and external factors (e.g., policy changes, competitor actions, macroeconomic shifts). Finally, formulate hypotheses for the most likely drivers based on the data, prioritize them, and outline next steps for deeper investigation and potential mitigation.

Practice these with instant AI feedback in a live mock interview → Start a Meta Analytics Engineer mock

Be fast and correct on coding; for design, drive the conversation; prepare impact-focused behavioral stories

Indicative Analytics Engineer pay in India: ~₹9–40 LPA (role-level range, not a Meta-specific figure).

Other roles at Meta

Analytics Engineer interviews at other companies