Technology

Google Maps

DeepMind + Google Maps: 50% More Accurate ETAs

50% improvement in ETA accuracy
1B+ users benefiting daily
Deployed in 97% of Google Maps regions
GNN architecture now standard

The Problem

ETA predictions in Google Maps needed to account for complex, non-linear traffic patterns. Traditional models used averages and couldn't capture cascading traffic effects.

The Solution

DeepMind partnered with Google Maps to apply Graph Neural Networks (GNNs) to traffic prediction. The model learns road segment dependencies and how congestion propagates through the network.

The Outcome

ETA accuracy improved by 50% in cities worldwide. Used by 1B+ users daily. Model handles 97% of predictions globally.

Key Metrics

  • 50% improvement in ETA accuracy
  • 1B+ users benefiting daily
  • Deployed in 97% of Google Maps regions
  • GNN architecture now standard
TechnologyTrafficGraph Neural NetworksMapsGoogle