The Roads Management Insights data models for trip duration and speed reading are built by combining different information sources:

- **Aggregated maps data:** The most critical source is aggregated, anonymized
  data from Google Maps, which allows Google Maps to calculate the real-time
  speed of vehicles on roads around the world.

- **Historical traffic data:** Over time, the aggregated user data is used to
  build historical traffic patterns, which help the system understand the
  "normal" traffic for a specific road at any given time and day of the week.

- **Supplemental data:** Historical data is combined with other data,
  including third-party information from partners like local Departments of
  Transportation, as well as real-time user feedback from Maps users reporting
  incidents like crashes or construction.

AI combines these information sources together to understand current conditions
with real-time data, and to provide baseline predictions with historical data.
This fusion is key for how routes are predicted, for example:

- Short routes depend largely on **current** , **real-time** information
- Longer routes use **advanced AI modeling** , where nearby segments are predicted using **real-time data** , while more-distant segments rely more heavily on **historical patterns.**
- Roads with limited real-time signals rely more heavily on its **historical
  data** to predict slowdowns.