DoorDash·Data Scientist·Technical Phone Screen·Senior

SeniorPrefer not to say

May 2026

Summary

DoorDash DS interview with a single very deep open-ended case question about measuring and forecasting the price gap between their platform and in-store menus. No behavioral rounds mentioned, just this beast of a problem. Left the room feeling like I'd only scratched the surface.

Questions Asked (1)

Design an end-to-end analysis to measure and forecast the gap between DoorDash platform menu prices and the same items' in-store prices. Cover how you'd collect and clean the data, build a price index, match identical items across sources, quantify uncertainty without an A/B test, forecast the gap using exogenous drivers like CPI and fuel costs, and size the study including minimum detectable effect under clustering and autocorrelation.

A/B Testing & ExperimentationProduct Analytics & MetricsData Modeling

Author's notes

This question is five questions duct-taped together and they want you to treat it like one coherent story.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by framing the business problem and defining the target estimand: the platform-to-in-store price gap for identical items. Then walk through a full pipeline: data collection and cleaning, item matching, index construction, uncertainty quantification, forecasting with exogenous drivers, and power analysis for sizing. Emphasize practical trade-offs and validation at each stage.

Pro tip: Don't just list methods—show you understand the data-generating process: in-store prices are often collected via crowdsourcing or receipts, which introduces selection bias and measurement error. Propose a validation subsample with ground-truth audits to correct for these biases.

1. Data Collection and Cleaning

Identify sources: DoorDash platform data (menu prices, item IDs, restaurant IDs, timestamps, location) and in-store prices (crowdsourced receipts, store visits, third-party data). Clean by deduplicating, handling missing values, and standardizing units/currencies. Document data quality issues and potential biases.

2. Item Matching and Price Index Construction

Match identical items across sources using a combination of exact identifiers (e.g., GTIN, SKU) and fuzzy matching on name, description, size, and modifiers. Validate matches with a manually labeled sample. Build a price index (e.g., Laspeyres or Fisher) to track the gap over time, weighting by order volume or item popularity.

3. Uncertainty Quantification and Forecasting

Quantify uncertainty without A/B tests using bootstrapping, Bayesian hierarchical models, or variance estimation from observational data. Forecast the gap using time series models (e.g., ARIMA, Prophet) with exogenous drivers like CPI, fuel costs, and local wages. Validate with backtesting and sensitivity analysis.

4. Study Sizing and Power Analysis

Determine sample size needed to detect a minimum detectable effect (MDE) under clustering (e.g., by restaurant or region) and autocorrelation (time series). Use formulas for clustered designs (intraclass correlation) and adjust for temporal correlation (effective sample size). Simulate power under different scenarios.

Key Points to Mention

Data sources and biases: platform data vs. in-store data collection methods (crowdsourcing, receipts) and how to mitigate selection bias.
Item matching: use of unique identifiers, fuzzy matching, and manual validation to ensure identical items.
Price index: choice of index formula (Laspeyres, Paasche, Fisher) and weighting scheme (order volume, revenue).
Uncertainty without A/B test: bootstrapping, Bayesian models, and accounting for measurement error.
Forecasting: use of exogenous variables (CPI, fuel costs) and time series methods; validation via backtesting.
Power analysis: clustering (ICC), autocorrelation, and simulation for MDE under realistic conditions.

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