Lyft Data Scientist Interview Loop
Lyft runs two data science tracks, Decisions and Algorithms, and the loop is the same shape for both. A recruiter call, a one-hour technical phone screen, and a virtual onsite of four to six interviews. The phone screen has no coding. It is probability, experiment design and business questions about the marketplace. The onsite adds a business case, a technical case, live coding and a conversation about your experience. There is no take-home.
What decides it
The phone screen. It is sixty minutes with one or two interviewers on probability and marketplace reasoning, and most candidates who get a no get it there. Practice saying the reasoning, not just the answer.
Practice by round
SQL and Python on ride data: average fare by city, cancellation rate by city, pickup wait by hour, tipped ride share. The coding round comes at the onsite, where Decisions candidates manipulate a rideshare dataset and Algorithms candidates solve a programming problem end to end.
| Question | Difficulty |
|---|---|
| Find the average distance traveled in each hour | Easy |
| Bikes Last Used | Easy |
| Average Fare by City | Easy |
| Driver Ride Summary | Easy |
| High-Value Riders | Easy |
| Cancellation Rate by City | Easy |
| Pickup Wait by Hour | Easy |
| Tipped Ride Share | Easy |
Harder queries on the same data, like actual versus predicted arrival time and churn rate of drivers. Lyft’s coding round is about handling data well rather than algorithm puzzles.
| Question | Difficulty |
|---|---|
| City with Most Customers | Medium |
| Actual vs Predicted Arrival Time | Medium |
| Most Profitable City of 2021 | Medium |
| Find The Combinations | Medium |
| Churn Rate Of Lyft Drivers | Medium |
| Start Dates Of Top Drivers | Medium |
| Distances Traveled | Medium |
| Vehicle Gear Usage | Medium |
| Find the fraction of rides for each weather and the hour | Hard |
The business case: which cities have higher or lower demand, whether to add a waiting period before a rider can cancel, and how to estimate time to arrive. The projects on driver lifetime value and demand for electric bicycles are the same questions worked with data.
| Case study | Difficulty |
|---|---|
| Heatmap Telling Taxi Drivers | Easy |
| Cities with Higher or Lower Demand | Medium |
| Estimated Time to Arrive | Hard |
| Adding Waiting Period | Hard |
| Data project | Difficulty |
|---|---|
| Driver Lifetime Value | Easy |
| Demand for Electric Bicycles | Easy |
Probability and experiment design, which is most of the phone screen. Binomial and conditional probability, the expected cost of a discount coupon, whether a coin is fair, and how to test a change for drivers when drivers and riders share one market.
| Question | Difficulty |
|---|---|
| Circle in Polar Coordinates | Easy |
| Lifetime Driver | Easy |
| Binomial Distribution | Easy |
| Cost of Discount Coupons | Medium |
| Discount Coupons Usage | Medium |
| Two Coupons | Medium |
| P Value for the Coin to be Fair | Medium |
| Switchback Versus User-Level Randomization | Medium |