← Meta Interview Insights

Meta·Data Scientist·Onsite - Multi Round·Senior

SeniorPrefer not to say
Jun 2026

Summary

A Meta DS onsite that was basically one giant experiment design question stretched across five sub-parts. The kind of question where you think you've covered everything and then they hit you with 'but what about interference' and you realize you haven't.

Questions Asked (5)

Q1

Your team is launching a Group Calls feature in an app that currently only supports 1:1 calls. Pick one primary metric and 3 to 5 guardrails for measuring its impact. Also explain when total call volume is a misleading metric and when it might actually be fine to use, and propose rate-based or user-normalized alternatives.

A/B Testing & ExperimentationProduct Analytics & Metrics
Author's notes

I went with call success rate per active user as my primary metric.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by defining a primary success metric that directly captures the new value of group calls, such as the percentage of active users who participate in a group call. Then select guardrails that monitor potential negative side effects on the existing 1:1 experience, overall engagement, and system health. Finally, discuss the limitations of total call volume and propose normalized alternatives like calls per user or per DAU.

Pro tip: Frame your primary metric as a 'success metric' that reflects the feature's core value, and choose guardrails that are leading indicators of long-term harm—this shows you think about both growth and sustainability. Also, explicitly state that you would validate the metric's sensitivity through an A/B test or holdout.

1. Define the primary success metric

Choose a metric that directly measures adoption and value creation from group calls, such as the percentage of active users who initiate or join a group call within a given period.

2. Select guardrail metrics

Pick 3-5 metrics that ensure the feature doesn't harm existing user behavior or system performance, e.g., 1:1 call frequency, overall call duration, user retention, crash rate, and notification opt-outs.

3. Explain when total call volume is misleading

Total call volume can increase simply because group calls count as one call but involve multiple users, or due to a small number of power users, masking changes in per-user behavior.

4. Propose normalized alternatives

Suggest rate-based or user-normalized metrics such as calls per DAU, group calls per user, or percentage of users making group calls, which control for user growth and power-user effects.

5. Discuss when total call volume is acceptable

Total call volume might be fine for capacity planning or infrastructure scaling, where aggregate load matters more than per-user behavior.

Key Points to Mention

  • Primary metric should directly measure group call adoption and value, e.g., % of active users in group calls.
  • Guardrails should cover existing 1:1 call behavior, overall engagement, retention, and technical health (e.g., crashes, latency).
  • Total call volume is misleading when group calls inflate counts without reflecting per-user engagement or when power users dominate.
  • Normalized alternatives: calls per DAU, group calls per user, or % of users making group calls.
  • Total call volume is appropriate for infrastructure capacity planning and aggregate load monitoring.
  • Consider segmenting by user cohorts (e.g., new vs. existing) to detect heterogeneous effects.

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

Q2

How would you choose a randomization unit for this experiment, and what rollout plan would you use to reduce network interference? Compare graph cluster randomization, ego-clusters, and geo-based switchbacks on bias, variance, and engineering cost.

A/B Testing & ExperimentationTechnical Trade-offs
Author's notes

This is where I started to sweat a little.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by framing the choice of randomization unit as a trade-off between bias, variance, and engineering cost, then systematically compare graph cluster randomization, ego-clusters, and geo-based switchbacks. Emphasize that the optimal choice depends on the experiment's goals, the strength of interference, and available resources, and propose a rollout plan that balances validity with practicality.

Pro tip: Acknowledge that no method is universally best; instead, recommend a hybrid or staged approach—e.g., pilot with ego-clusters for quick signal, then validate with graph clusters or switchbacks—and always quantify interference via A/A tests or holdout groups.

1. Define the interference problem and objectives

Clarify how interference manifests (e.g., social contagion, shared resources) and what the experiment aims to measure. Establish whether the goal is to estimate global or local effects, and the acceptable levels of bias and variance.

2. Compare methods on bias, variance, and cost

For each method, analyze bias (e.g., graph clusters reduce bias but may not eliminate it; ego-clusters balance bias and variance; switchbacks control temporal interference but may introduce time trends), variance (graph clusters often have high variance due to few clusters; ego-clusters lower variance; switchbacks can have low variance if well-designed), and engineering cost (graph clusters require full graph and complex randomization; ego-clusters are simpler; switchbacks need careful time segmentation and may disrupt user experience).

3. Select the randomization unit based on trade-offs

Choose the method that best aligns with the experiment's constraints and goals. For example, if interference is strong and the graph is known, graph cluster randomization may be preferred despite higher cost; if quick iteration is needed, ego-clusters offer a good balance; if interference is primarily temporal (e.g., supply-side), geo-based switchbacks are suitable.

4. Design a rollout plan to mitigate interference

Propose a phased rollout: start with a small pilot using the chosen method, monitor for interference via A/A tests or holdout groups, and gradually scale. Consider techniques like cluster-level randomization with buffer zones, time-based switchbacks with washout periods, or ego-cluster randomization with network exposure modeling.

5. Validate and iterate

After rollout, validate assumptions (e.g., no spillover) using diagnostics like network autocorrelation or difference-in-differences. Be prepared to adjust the randomization unit or rollout plan based on observed interference and variance.

Key Points to Mention

  • Graph cluster randomization: reduces bias from network interference but increases variance and engineering complexity due to graph partitioning and cluster-level assignment.
  • Ego-clusters: randomizes each node with its immediate neighbors, balancing bias and variance, but may still have residual interference and requires careful handling of overlapping clusters.
  • Geo-based switchbacks: alternates treatment over time within geos, controlling temporal interference, but can be confounded by time trends and requires sufficient time periods.
  • Bias-variance trade-off: graph clusters minimize bias at the cost of variance; ego-clusters offer a middle ground; switchbacks can have low variance but potential bias from time effects.
  • Engineering cost: graph clusters need full network data and complex randomization; ego-clusters are simpler but need network info; switchbacks require time-based randomization and may disrupt user experience.
  • Rollout plan: pilot with A/A tests, use holdout groups, consider staggered rollout, and monitor interference metrics to adjust design.

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

Q3

If cluster-based randomization isn't feasible, what quasi-experimental design would you use instead? Walk through the assumptions and diagnostics you'd run.

A/B Testing & ExperimentationProduct Analytics & Metrics
Author's notes

Went with invitation-gated treatment as my main proposal: only users who receive a group invite token are exposed, which gives you a natural instrument.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by acknowledging that cluster randomization is often infeasible due to network effects, interference, or operational constraints, and propose a quasi-experimental design like difference-in-differences (DiD) or synthetic control. Then, systematically walk through the key assumptions (e.g., parallel trends, no spillovers) and the diagnostics you'd run to validate them, emphasizing how you'd handle violations.

Pro tip: Show maturity by discussing how you'd combine quasi-experimental methods with sensitivity analyses (e.g., placebo tests, robustness checks) to quantify uncertainty, and mention that you'd pre-register the analysis plan to avoid p-hacking.

1. Choose the quasi-experimental design

Select a design like difference-in-differences (DiD), synthetic control, or interrupted time series based on the treatment assignment mechanism and data availability. Justify why it's the best fit given the constraints.

2. State the key assumptions

Clearly articulate the assumptions required for causal inference, such as parallel trends (for DiD), no interference/spillovers, and correct model specification. Explain how each assumption relates to the design.

3. Run diagnostics to test assumptions

Describe specific diagnostics: pre-trend tests (event study plots), placebo tests (e.g., fake treatment dates), and checks for spillovers (e.g., comparing treated and control units' characteristics). Mention statistical tests and visualizations.

4. Address violations and sensitivity

If assumptions are violated, discuss remedies like propensity score matching, synthetic control, or instrumental variables. Conduct sensitivity analyses (e.g., Rosenbaum bounds, different control groups) to assess robustness.

5. Interpret results and limitations

Quantify the treatment effect with confidence intervals, and clearly communicate limitations and the strength of causal evidence. Suggest follow-up experiments if possible.

Key Points to Mention

  • Difference-in-differences (DiD) and synthetic control as alternatives to cluster randomization
  • Parallel trends assumption and how to test it with pre-treatment data
  • Placebo tests and falsification checks to validate design
  • Handling spillovers/interference via buffer zones or modeling
  • Sensitivity analysis (e.g., Rosenbaum bounds, robustness to control group selection)
  • Pre-registration and transparency to mitigate p-hacking concerns

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

Q4

How do you define exposure and triggers for this experiment? What are your exclusion criteria, and how do you prevent contamination between treatment and control groups?

A/B Testing & ExperimentationProduct Analytics & Metrics
Author's notes

Creators vs joiners distinction is important here and I made sure to flag it early.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by defining the target population and the precise moment users become eligible for the experiment, then explain how you assign them to treatment or control. Next, outline your exclusion criteria to remove invalid or biased users, and finally describe the technical and analytical safeguards you use to prevent contamination between groups.

Pro tip: Emphasize that you document all decisions before launch and monitor for contamination post-launch using metrics like cross-group exposure rates; this shows you understand that experimentation is an ongoing process, not a one-time setup.

1. Define Exposure and Triggers

Specify the exact user action or system event that triggers inclusion in the experiment, and define what constitutes 'exposure' to the treatment (e.g., seeing the new feature). Ensure the trigger is aligned with the experiment's goal and is measurable.

2. Establish Exclusion Criteria

Identify users or events to exclude, such as bots, internal employees, users who previously participated in similar experiments, or those with incomplete data. Justify each exclusion based on potential bias or interference.

3. Prevent Contamination

Implement randomization at the appropriate unit (e.g., user, session) and use techniques like unique user IDs, cookie-based bucketing, or network isolation to prevent users from being exposed to both treatment and control. Consider spillover effects and use cluster randomization if needed.

4. Monitor and Validate

After launch, check for contamination by analyzing exposure logs, ensuring users are consistently in one group, and monitoring for unexpected cross-group interactions. Use holdout groups or A/A tests to validate the setup.

Key Points to Mention

  • Randomization unit (user, session, device) and its implications for contamination
  • Exclusion criteria such as bots, internal users, and users with prior exposure
  • Techniques to prevent contamination: unique identifiers, bucketing, network effects isolation
  • Metrics to detect contamination: exposure rate, cross-group conversion, sample ratio mismatch (SRM)
  • Documentation and pre-registration of exposure, triggers, and exclusions
  • Consideration of spillover effects and use of cluster randomization when necessary

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

Q5

How would you pre-register this analysis? Specify your minimum detectable effect given plausible baselines, how you'd handle seasonality, and what you'd do about novelty and ramp-up effects.

A/B Testing & ExperimentationProduct Strategy
Author's notes

Pre-registration piece was fine: lock in primary metric, guardrails, sample size calculation, and analysis plan before launch.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by framing the pre-registration as a way to ensure rigor and avoid p-hacking, then walk through the key components: defining the metric and hypothesis, calculating the minimum detectable effect (MDE) based on plausible baselines, addressing seasonality through design and analysis, and mitigating novelty and ramp-up effects via run-in periods and extended measurement. Emphasize that pre-registration includes specifying these details in advance to maintain statistical validity.

Pro tip: Show that you understand the business context: at Meta, even small effect sizes can matter at scale, so balance statistical power with practical significance when choosing MDE. Also, mention that pre-registration should include a plan for multiple comparisons and sequential testing if applicable.

1. Define hypothesis and primary metric

Clearly state the null and alternative hypotheses, and select a primary metric that is sensitive to the change and aligned with business goals. Specify guardrail metrics to monitor for unintended consequences.

2. Determine baseline and MDE

Estimate the baseline metric value from historical data or a pilot. Choose an MDE that is the smallest effect size that would justify the cost of implementation, considering statistical power (typically 80%) and significance level (typically 5%).

3. Account for seasonality

Use historical data to identify seasonal patterns and either stratify randomization by time or include time as a covariate in the analysis. Alternatively, run the experiment long enough to cover full seasonal cycles.

4. Mitigate novelty and ramp-up effects

Implement a ramp-up period where the treatment is gradually rolled out, and exclude the initial days from analysis to let user behavior stabilize. Use a holdback group to measure long-term effects and detect novelty decay.

5. Pre-register analysis plan

Document all decisions—metric definitions, MDE, sample size, duration, exclusion criteria, and statistical tests—in a pre-registration document before data collection. This prevents post-hoc rationalization and ensures transparency.

Key Points to Mention

  • Minimum detectable effect (MDE) calculation using power analysis, with assumptions about baseline variance and desired power.
  • Seasonality handling through stratified randomization, time as a covariate, or ensuring experiment duration covers seasonal cycles.
  • Novelty effect mitigation via run-in periods, excluding early data, and using holdback groups for long-term measurement.
  • Ramp-up effects addressed by gradual rollout and analyzing data only after stabilization.
  • Pre-registration benefits: reduces p-hacking, increases reproducibility, and aligns stakeholders on success criteria.
  • Consideration of multiple testing corrections and sequential analysis if peeking at data.

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