Explain the Data and Models Behind Google Maps Arrival-Time Estimates
Company: Google
Role: Product Manager
Category: Machine Learning
Difficulty: medium
Interview Round: Other
During a product manager interview at Google, the interviewer asks a technical question: when someone asks Google Maps for directions, how does the app arrive at the estimated time of arrival (ETA) it shows? Explain what information the estimate is built from, how that information becomes a predicted travel time for a route, and how the estimate stays current once the person is on the move.
```hint Think in pieces
A route is a chain of smaller pieces. Ask what it takes to predict the time for one piece at the moment the traveler will actually reach it, and where each input for that prediction comes from.
```
### Clarifying Questions
- Should I focus on driving, or also cover public transit, walking and cycling, which rely on different data?
- Do you mean the estimate shown before departure, the one that updates during turn-by-turn navigation, or both?
- How deep should I go: a product-level explanation of the inputs and logic, or the modeling approach as well?
- Should I treat the route as already chosen, or also explain how travel-time predictions decide which route is suggested?
### What a Strong Answer Covers
- Breaking a trip into road segments or transit legs whose predicted times, counted from the departure time, add up to the ETA.
- The main data sources (historical patterns, live aggregated speeds, incident and closure reports, road attributes, transit schedules and live feeds) and what each one contributes.
- How historical and live signals are combined into a forward-looking prediction, including the role of machine learning and the handling of roads with little data.
- How the estimate and the route are refreshed during the trip, and how predicted versus actual arrival times are used to measure and improve accuracy.
- Product judgment about why accurate, stable ETAs matter to users, the main sources of error, and the privacy of the location data involved.
### Follow-up Questions
- A stadium event ends and thousands of cars leave at once. Why might nearby ETAs be wrong for the first few minutes, and how could the system catch up faster?
- How would you measure ETA accuracy, and would you rather the app's errors lean early or late?
- Traffic patterns change suddenly across a whole city, as they did during pandemic lockdowns. What happens to a model trained on history, and what would you change?
Overview: A Google product manager interview question asking how Google Maps works out the estimated arrival time it shows for a trip. It tests whether a candidate can reason about the data behind a travel-time estimate, how the prediction stays current during the trip, and how its accuracy would be judged, all in clear product terms.