← Meta Interview Insights

Meta·Data Scientist·Technical Phone Screen·Intermediate

Intermediate
Jun 2026

Summary

Meta data scientist interview with a statistics and product analytics focus. Three questions covering descriptive stats, probability, and metric divergence diagnosis. The probability one was straightforward but the last question is where things got interesting and a bit uncomfortable.

Questions Asked (3)

Q1

Given a distribution of video view counts, how would you compute and interpret the mode, median, and mean? What relationship do you expect between these three for this type of data, and which would you actually report?

Product Analytics & Metrics
Author's notes

This felt like a warmup but I almost tripped on the 'which should you report' part.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by defining each metric and explaining how to compute them from a video view count distribution. Then discuss the expected right-skewed shape and the typical mean > median > mode relationship, and finally recommend which metric to report based on the business question and robustness to outliers.

Pro tip: Mention that the mean is sensitive to viral videos, so the median is often more representative for typical performance, but the mean is still useful for total engagement or revenue calculations.

1. Define and compute each metric

Explain that the mode is the most frequent view count, the median is the middle value when sorted, and the mean is the sum divided by the number of videos.

2. Describe the expected distribution shape

Note that video view counts typically follow a right-skewed (power-law) distribution with a long tail of highly viewed videos.

3. Interpret the relationship

State that for right-skewed data, the mean is pulled to the right, so mean > median > mode, and explain why this happens.

4. Recommend which metric to report

Advise that the median is often best for typical performance, while the mean is useful for total impact; the mode is rarely informative for continuous view counts.

5. Connect to business context

Tie the choice of metric to the specific product question, such as measuring typical user engagement versus total watch time.

Key Points to Mention

  • Right-skewed distribution due to viral videos
  • Mean is sensitive to outliers; median is robust
  • Mode may be less useful for continuous data with many unique values
  • Mean > median > mode for right-skewed data
  • Report median for typical performance, mean for total impact
  • Consider log transformation or trimmed mean for analysis

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

Q2

If each video a user watches has a 10% chance of being shared independently, what's the probability that at least one out of ten videos gets shared?

Product Analytics & Metrics
Author's notes

Complement rule, done.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Recognize that the question asks for the probability of at least one success in 10 independent trials with a 10% success rate. Use the complement rule: calculate the probability of no shares (0.9^10) and subtract from 1. Then compute the numerical value and interpret it in a product context.

Pro tip: After computing the probability, briefly discuss how this simple model might be extended to real-world scenarios, such as non-independent sharing behavior or varying share rates, to show depth and business acumen.

1. Clarify assumptions

Confirm that each video's sharing event is independent and that the probability of sharing is constant at 10% for each video.

2. Define the event

Let X be the number of videos shared out of 10. We want P(X ≥ 1).

3. Use complement rule

Compute P(X = 0) = (1 - 0.1)^10 = 0.9^10, then P(X ≥ 1) = 1 - 0.9^10.

4. Calculate numerical value

Compute 0.9^10 ≈ 0.3487, so P(X ≥ 1) ≈ 0.6513 or 65.13%.

5. Interpret and contextualize

Explain that there is about a 65% chance that at least one video is shared, and discuss implications for virality or engagement metrics.

Key Points to Mention

  • Independence assumption and its limitations in real user behavior
  • Complement rule for 'at least one' probability problems
  • Binomial distribution framework (n=10, p=0.1)
  • Calculation of 0.9^10 and final probability
  • Interpretation in terms of product metrics (e.g., expected shares, virality)
  • Potential extensions: varying share rates, dependencies, or larger n

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

Q3

You notice that average watch time per video is going up while average share rate per video is going down at the same time. What could explain this divergence and how would you test your hypotheses, separating causal from correlational explanations?

Product Analytics & MetricsA/B Testing & ExperimentationRoot Cause Analysis
Author's notes

This is the one I wish I'd structured better from the start.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by decomposing the metrics and considering possible shifts in user behavior, content mix, or platform changes that could drive watch time up and shares down. Then outline a structured plan to test each hypothesis, using experiments and causal inference methods to distinguish correlation from causation. Finally, discuss how you would validate findings and recommend actions.

Pro tip: Always tie metrics back to the product's north star and consider whether the divergence is actually a problem—sometimes increased watch time with lower shares could indicate a shift to more passive consumption, which might be fine if it aligns with business goals.

1. Clarify the metrics and context

Define exactly how watch time and share rate are calculated, and check if the divergence is consistent across segments (e.g., user demographics, content types, platforms).

2. Generate hypotheses

Brainstorm potential explanations: changes in content recommendation algorithms, seasonal effects, new user influx, UI changes, or shifts in content format (e.g., longer videos).

3. Prioritize and test hypotheses

For each hypothesis, design observational analyses (e.g., cohort analysis, regression) and experiments (e.g., A/B tests) to test the relationship, ensuring to control for confounders.

4. Separate causation from correlation

Use methods like instrumental variables, difference-in-differences, or propensity score matching to establish causality, and validate with experiments where possible.

5. Synthesize and recommend

Summarize findings, quantify the impact, and propose actionable recommendations, considering potential trade-offs and long-term effects.

Key Points to Mention

  • Metric decomposition and segmentation to identify if the divergence is uniform or driven by specific groups.
  • Potential confounders such as seasonality, external events, or platform changes.
  • Use of A/B tests to establish causality, e.g., testing a new ranking algorithm.
  • Causal inference techniques (e.g., difference-in-differences, instrumental variables) when experiments are not feasible.
  • Consideration of user intent and content type: passive consumption vs. active sharing.
  • Alignment with business goals: is the divergence necessarily bad? Could it indicate a shift in user behavior that is acceptable or even beneficial?

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