Marketplaces, gig & delivery Data Scientist Interview Loop
Marketplaces run the most uniform data scientist interview of any industry. There's a recruiter call, then a technical screen that pairs 30 to 45 minutes of timed SQL with a marketplace case. Then come four to five interviews of 45 to 60 minutes: SQL and coding, statistics and experiments, a marketplace case, and behavioral. Uber adds a Bar Raiser who shapes your level. DoorDash matches you to a team after the screen, and two out of three of its rounds include a case, with interviewers who interrupt as you go. General big tech prep misses what they all test: a change on one side of the market moves the other.
Companies behind this pattern
Practice by round
Live SQL in a plain editor, timed, with little room for mistakes. The tables are trips, orders, deliveries, drivers and riders. Expect joins, aggregations and filters on time-windowed data, then one problem that needs a subquery or a window function.
Marketplace-level difficulty: extremely late deliveries, average order cost during rush hours, the share of rides by weather and hour, Lyft driver churn, host popularity and rental prices, and first-time orders. The follow-up is always what the number tells a marketplace manager.
The marketplace case, which at DoorDash and Uber is most of the loop. Questions include regional delivery delays, cities with higher or lower demand, estimated time to arrive, adding a waiting period, accepting rides, and a framework for judging a product launch. Find the cause across both sides, and name the trade-offs before the interviewer does.
| Case study | Difficulty |
|---|---|
| Heatmap Telling Taxi Drivers | Easy |
| Favorite Product or App | Easy |
| Main Outcome Metrics | Easy |
| Accepting Rides | Easy |
| Uber Black Rides | Medium |
| Cities with Higher or Lower Demand | Medium |
| Customer Engagement and Disengagement | Medium |
| Product Launch Evaluation Framework | Medium |
| Estimated Time to Arrive | Hard |
| Adding Waiting Period | Hard |
Experiments where the groups affect each other, which marketplaces go deeper on than any other industry. Expect A/B testing with spillover effects, and what to do when rider and driver results conflict. Next to that are a binomial distribution, a p-value for a fair coin, an anomaly in a distribution, and the chance that discount coupons get used.