Google·Data Scientist·Technical Phone Screen·Senior

Senior

Jun 2026

Summary

Google data scientist interview with a meaty case question about comparing two competing snack shops over a 14-day window. One question, but it unpacked into five sub-parts covering metric design, data collection, causal identification, regression modeling, and decision guardrails. Dense.

Questions Asked (1)

Two snack shops operate side by side at a school entrance. You have 14 days to figure out which one will generate more profit over the next quarter. Design a full measurement and analysis plan: define your primary metric and why it's the right one, specify the minimum data you'd collect, describe your causal identification strategy given real-world confounders like exam weeks and mid-study price changes, write out the regression you'd fit including fixed effects and controls, and state your explicit decision rule with guardrails for things like stockouts.

A/B Testing & ExperimentationProduct Analytics & MetricsData Modeling

Author's notes

This one sprawled fast.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by defining a clear primary metric (e.g., daily profit per shop) and justify why it aligns with the business goal. Then outline a data collection plan that captures key variables, and propose a causal identification strategy using fixed effects and controls to handle confounders. Finally, specify a regression model and a decision rule with guardrails.

Pro tip: Emphasize that you would pre-register your analysis plan and decision rule to avoid p-hacking, and mention that you'd check for heterogeneous treatment effects across days (e.g., exam vs. non-exam) to see if the profit difference is consistent.

1. Define Primary Metric and Justify

Choose daily profit (revenue minus costs) as the primary metric because it directly measures the business outcome. Explain why other metrics like revenue or customer count are insufficient.

2. Data Collection Plan

Specify the minimum data: daily profit, sales, prices, costs, foot traffic, day of week, exam schedule, promotions, stockouts, and weather. Collect for both shops over 14 days.

3. Causal Identification Strategy

Use a difference-in-differences or fixed effects model to control for time-invariant shop characteristics and common time trends. Include day fixed effects to absorb exam weeks and other temporal confounders.

4. Regression Model Specification

Fit: Profit_it = α + β*Shop_i + γ_t + δ*X_it + ε_it, where Shop_i is a dummy for shop A, γ_t are day fixed effects, and X_it includes controls like price, promotions, stockouts, and foot traffic.

5. Decision Rule and Guardrails

Decide shop A generates more profit if β is positive and statistically significant (p<0.05) and the 95% CI lower bound exceeds a minimum practical effect (e.g., $10/day). Guardrails: exclude days with stockouts >10% or run sensitivity analysis.

Key Points to Mention

Primary metric: daily profit, not revenue, because it accounts for costs.
Use fixed effects for days to control for exam weeks and other time-varying confounders.
Include control variables: price changes, promotions, stockouts, foot traffic, weather.
Pre-register the analysis plan and decision rule to ensure validity.
Check for heterogeneous effects: does the profit difference vary by exam vs. non-exam days?
Guardrails: handle stockouts by excluding affected days or using instrumental variables.

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