Practice real data science interviews tuned by the interviewers who actually run them.

Cold food complaints have spiked 34% over the past 3 weeks across 12 metro areas. The VP of Customer Experience wants a root cause analysis. Two recent operational changes — batched delivery and prep time inflation — may be interacting in unexpected ways.

Average delivery time increased by 4.2 minutes last month across the platform. No major product changes were shipped. The operations team is alarmed and wants to know what's driving the increase and whether it requires immediate action.

Consumer order cancellations are up 28% in specific geographic markets over the past two weeks. The pattern is concentrated in suburban areas during dinner rush hours. You need to diagnose the root cause and recommend immediate actions.

DoorDash's Sponsored Listings product claims merchants see a 2.3x order lift, but the Growth team suspects selection bias is inflating the number. You need to audit true ad incrementality and determine whether the $18M/month ad product is delivering real value.

DoorDash launched Store Search — cross-restaurant item search (e.g., 'pad thai' returns results from multiple restaurants). Engagement looks great: 38% adoption, 15% of orders from search. But the VP of Product challenges whether it's creating new demand or just redistributing existing orders.

You've been asked to define the core metrics framework for understanding marketplace demand at DoorDash. This goes beyond simple order counts — you need metrics that capture demand health, unmet demand, and early warning signals for demand-supply imbalances.

Leadership wants a single health scorecard that captures the entire DoorDash marketplace across consumers, merchants, and Dashers. You need to define the metric hierarchy, handle conflicting metrics across sides, and design something stakeholders can actually use.

DoorDash launched alcohol delivery as a new vertical. You need to define success metrics that account for the unique characteristics of alcohol delivery — different purchase patterns, regulatory requirements, and potential cannibalization of food orders.

An experiment improves consumer conversion by 3% but hurts Dasher acceptance rate by 2%. You need to recommend whether to launch. This tests your ability to reason about two-sided marketplace trade-offs without defaulting to a binary decision.

DoorDash is redesigning the Dasher payout structure. The current system has complaints about unpredictability and perceived unfairness. You need to design a new system that improves Dasher satisfaction while maintaining platform economics.

DoorDash's grocery delivery vertical has grown rapidly, but the restaurant delivery team is concerned about cannibalization. You need to determine whether grocery is additive to the platform or stealing from restaurant orders.

DoorDash runs an ad marketplace where restaurants pay for sponsored placement in search results. You need to measure whether these ads actually drive incremental orders or just cannibalize organic traffic, while balancing the interests of merchants, consumers, and the platform.

DoorDash is launching in Austin (sprawling, car-dependent) and Portland (compact, bike-friendly). The current dispatch algorithm optimizes solely for distance-to-restaurant and was trained on suburban markets. It systematically under-assigns bike Dashers in dense areas. Design an experiment to test city-optimized dispatch.

DoorDash launched Peak Alerts — push notifications urging offline Dashers to go online during high-demand periods. Headline metrics look great: 28% more Dashers during alerts, 62% fewer unfulfilled orders. But deeper analysis may reveal that Peak Alerts cannibalize off-peak supply instead of creating net new supply.

The subscription team wants to test a $2/month price increase for DashPass (from $9.99 to $11.99). You're the data scientist tasked with designing this experiment. Pricing experiments have unique challenges — long-term churn effects aren't visible in short tests, and existing vs. new subscribers react very differently.

The marketing team wants to test a 5% off delivery fee coupon to drive consumer orders. But you know that many orders would have happened anyway. How do you measure the true incremental impact when marketplace dynamics create spillover effects?

You're testing a new real-time batching algorithm that assigns 2 orders to a single Dasher trip. This affects supply, demand, and timing across the entire local marketplace. Standard A/B testing breaks down because of interference effects — you need a switchback design.

DoorDash sends push notifications to offline Dashers during high-demand periods, asking them to go online. The operations team wants to know if this is actually effective at increasing supply, or if Dashers would have come online anyway.

DoorDash is testing a new restaurant ranking algorithm that prioritizes delivery speed and proximity over ratings and popularity. Early data looks promising for delivery times, but there are concerns about fairness to merchants and consumer discovery.

The hypothesis is that late deliveries cause customer churn. But you can't randomly assign late deliveries — you need to find a clever identification strategy to prove or disprove the causal relationship and size the business impact.

Wrong orders are one of DoorDash's biggest customer pain points and refund cost drivers. You've been asked to design a machine learning system that minimizes wrong orders — but you first need to clarify what 'minimize' means and where in the pipeline to intervene.

DoorDash is considering expanding into 5 international markets. You need to build a prioritization framework that balances demand potential, operational feasibility, competitive landscape, and unit economics — then design a pilot to validate before committing.

You need to estimate consumer lifetime value (LTV) from first 30 days of behavior. This model will inform acquisition spend, retention investment, and growth strategy. The challenge: early behavior is noisy, promotional effects inflate initial usage, and different customer segments have wildly different trajectories.

DoorDash's refund costs are growing faster than order volume. Customer support suspects a significant portion of refund claims are fraudulent — users claiming missing or wrong items that were actually delivered correctly. You need to design a detection system and quantify the financial impact.