Meta·Data Scientist·Technical Phone Screen·Senior

Senior

Jun 2026

Summary

Meta DS interview with a meaty product case around estimating revenue potential for a new Shopping tab. Single question but it had like five sub-parts, so the whole session was basically one long back-and-forth working through the math and methodology.

Questions Asked (1)

Your team is evaluating a new Shopping tab in a social photo app. Organic posts and in-app search are the only channels in scope. Using available data (product page views, add-to-carts, purchases, average order value, and organic impression counts), estimate the monthly revenue potential if the tab launches to 100% of users. Walk through your full conversion funnel and the algebra behind it, explain how you'd get unbiased CTR and product page view rates from existing organic entry points, describe how you'd adjust for cannibalization of purchases that would have happened anyway through other surfaces, explain how you'd scale experiment results to the full user population while accounting for user heterogeneity, and finish with a sensitivity analysis on your key assumptions.

Product Analytics & MetricsA/B Testing & ExperimentationPricing & Monetization

Author's notes

This was a lot.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by defining a clear conversion funnel from organic impressions to purchases, then estimate each stage's rate using existing organic data (e.g., search and feed) while correcting for biases. Scale the per-user revenue to the full population, adjust for cannibalization, and run sensitivity analyses on key assumptions like CTR and AOV.

Pro tip: Emphasize that you would validate your estimates with a holdout experiment or a small-scale launch to measure incremental lift, rather than relying solely on observational data. This shows you understand the difference between correlation and causation, which is critical for a Data Scientist at Meta.

1. Define the funnel and gather baseline metrics

Outline the conversion funnel: organic impressions → clicks (CTR) → product page views → add-to-carts → purchases. Use existing organic channels (e.g., in-app search, feed) to estimate each stage's rate, ensuring the data is representative.

2. Estimate unbiased rates from organic entry points

Identify potential biases (e.g., selection bias, position bias) in organic data and adjust using methods like propensity score weighting, stratified sampling, or instrumental variables. For CTR, consider using search results as a proxy but adjust for intent differences.

3. Scale to full population and adjust for cannibalization

Multiply per-user revenue by the total user base, but account for user heterogeneity by segmenting users (e.g., by activity level, demographics) and weighting accordingly. Estimate cannibalization by measuring overlap with existing purchase paths and applying a discount factor.

4. Conduct sensitivity analysis and validate with experiments

Perform sensitivity analysis on key assumptions (CTR, conversion rates, AOV) to understand revenue range. Propose a holdout experiment or phased rollout to measure true incremental impact and refine estimates.

Key Points to Mention

Funnel stages: impressions → clicks → product page views → add-to-carts → purchases, with average order value (AOV) to compute revenue.
Bias correction techniques: propensity score matching, stratification, or using search as a natural experiment to estimate CTR.
Cannibalization adjustment: measure overlap between new tab and existing surfaces, use difference-in-differences or holdout groups to isolate incremental purchases.
Scaling with heterogeneity: segment users by engagement or demographics, weight segments by population share, and consider network effects.
Sensitivity analysis: vary CTR, conversion rates, and AOV by ±20% to show revenue range and identify most impactful assumptions.
Validation: propose an A/B test or phased rollout to measure actual lift and compare with estimates.

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