Uber Data Analyst Interview Loop
The Uber data analyst interview starts with a 30-minute recruiter call that already checks how comfortable you are with SQL. Next is a 60-minute semi-technical screen with a hiring manager. Then there are around four onsite interviews: SQL, a product or operations case, statistics and experimentation, and behavioral. Uber uses one structure for L3 and L4 and decides your level afterwards, so there's no easier version for less experienced candidates. Many of these roles sit in operations, like city, logistics, delivery and support, and the cases show it.
What decides it
Asking why. Analysts who report that cancellations went up get a polite no. Analysts who explain what's driving it get through.
Practice by round
SQL basics inside the manager screen. You'll get joins, aggregations and a ranking question on trips and drivers. Then there's a short pandas problem, and a business question to see whether you ask why before you answer.
Trips completed within a window, actual versus predicted arrival, total orders per status and service, revenue loss, distance per dollar, average trip distance by city, and how well advertising channels work. Uber's follow-up is always the same. What does this tell an operations manager, and what would you check before believing it? Answer both.
| Question | Difficulty |
|---|---|
| Employees' Years In Service | Medium |
| Days Without Hiring/Termination | Medium |
| Total Monetary Value Per Month/Service | Medium |
| Percentage Of Revenue Loss | Medium |
| Actual vs Predicted Arrival Time | Medium |
| Customers with Large Orders | Medium |
| Sum Of Transportation Numbers | Medium |
| Completed Trip within 168 Hours | Hard |
Operations cases on Uber's own questions. They include accepting rides, congestion and driving conditions, which metrics bring in customers, and how to judge Uber Black. The projects cover city supply and demand, partner business modeling, and whether new drivers will start driving. Segment the problem, form a hypothesis, and say what you'd check first. Be clear about the two-sided market too. An incentive that fixes drivers costs riders, and the interviewer will ask about it.
| Data project | Difficulty |
|---|---|
| Insights from City Supply and Demand Data | Easy |
| Partner Business Modeling | Easy |
| Registrations by Email Provider | Easy |
| Will New Drivers Start Driving | Medium |
| Each User's Third Transaction | Medium |
Statistics at the analyst level. You'll spot an anomaly in a distribution, read a test result, and use enough probability to say whether a difference is real.
| Question | Difficulty |
|---|---|
| Anomaly in Distribution | Easy |
| SVM Classifier | Medium |
| Guardrails for a Two-Sided Change | Medium |
| Switchback Versus User-Level Randomization | Medium |
| When Rider and Driver Results Conflict | Hard |