Diagnose Cold Food Deliveries with Key Metrics Analysis
Company: DoorDash
Role: Data Scientist
Category: Analytics & Experimentation
Difficulty: medium
Interview Round: Onsite
##### Scenario
A food-delivery platform is receiving a spike in complaints that delivered meals arrive cold.
##### Question
How would you diagnose why customers receive cold food? What key metrics would you break down and monitor? Design an A/B test to evaluate a solution that keeps food hot.
##### Hints
Think supply-chain steps, time-in-transit, batching, packaging; define leading/lagging metrics, success criteria, and assignment unit.
Quick Answer: DoorDash analytics prompt on diagnosing cold food deliveries, covering complaint metrics, lifecycle timestamps, batching, packaging, weather, segmentation, data quality, A/B testing, and operational guardrails.
A food-delivery platform is receiving a spike in customer complaints that delivered meals arrive cold. You are the Data Scientist asked to diagnose the issue and evaluate a solution intended to keep food hot.
Constraints & Assumptions
Complaints are a noisy lagging signal; include leading metrics and direct measurements where possible.
Consider the full path from kitchen completion to customer handoff.
Segment by merchant, cuisine, distance, weather, batching, courier mode, packaging, and market load.
Design an experiment that measures customer benefit and operational trade-offs.
Clarifying Questions to Ask Guidance
How is "cold food" reported: complaint reason, refund request, rating text, photo, or temperature sensor?
Did complaint volume rise, complaint rate rise, or both?
Were there recent changes to batching, dispatch, packaging, merchant prep, courier supply, or weather?
What solution are we testing: insulated bags, packaging, dispatch timing, batching limits, merchant throttling, or routing?
What a Strong Answer Covers Guidance
Diagnostic framework mapping timestamps: prep complete, courier arrival, pickup, transit, drop-off, and total time at temperature risk.
Key metrics: cold complaint rate, refund rate, rating, repeat rate, cook-to-pickup time, transit time, batching/stacking, detours, distance, courier mode, weather, and packaging compliance.
Root-cause decomposition by merchant, cuisine, market, daypart, weather, order size, courier mode, and stack position.
Data-quality checks for reason-code changes, support-policy changes, denominator shifts, and complaint selection bias.
Experiment design with treatment/control, unit of randomization, eligibility, primary metric, guardrails, sample size, and duration.
Guardrails such as delivery time, cost, courier earnings, batching efficiency, merchant burden, cancellation, and customer satisfaction.
Follow-up Questions Guidance
How would you distinguish food getting cold at the restaurant from coldness during transit?
Which segment would you target first?
What if the fix reduces cold complaints but increases delivery time?