AI & Technical question

How does Google Maps compute ETA?

Practice this question out loud. An AI interviewer asks it, follows up like a real interviewer would, and scores your answer. Type or speak.

What this question tests

Tests technical fluency in explaining a real-time prediction system, covering inputs, modeling and the accuracy-latency tradeoff.

State the core inputs: historical traffic patterns by road segment and time of day, plus real-time traffic data from other users' location signals.
Explain the routing layer: a shortest-path or fastest-path algorithm over the road graph, weighted by predicted travel time per segment rather than raw distance.
Describe the prediction model: a machine learning model trained on historical and live data to predict segment-level travel time, accounting for factors like weather and events.
Note the real-time correction: the ETA updates continuously as the trip progresses, incorporating the driver's actual observed speed and any new congestion.
Mention the tradeoff: more frequent updates improve accuracy but cost more compute and battery, so updates are throttled and smoothed rather than instantaneous.

Correctly separates the routing algorithm, finding the path, from the prediction model, estimating time per segment, rather than conflating the two.
Names both historical and real-time crowdsourced data as inputs, the actual basis of Google Maps' real system.
Mentions the continuous re-estimation during a trip, not just a single upfront ETA calculation.

Describing only a static shortest-distance calculation without addressing time prediction or real-time traffic.
Ignoring that the ETA updates continuously during the trip based on actual observed progress.

How would you handle ETA prediction in an area with little historical traffic data?
How would you evaluate whether an ETA model is accurate enough to ship?