Plan road interventions, then prove they worked

A city decides where to widen, where to re-phase a signal, where to add a flyover, from origin destination surveys that capture a single day every few years. The 1954 Detroit study set that paradigm, and most planning still works the same way: a sample, frozen in time, that says nothing about the corridor the day after the work is done.

So interventions ship on intuition and complaints, and nobody can say with data whether the last one helped. The corridors that hurt commuters every evening rarely sit at a camera junction, so the camera network never sees them. In Pune, monitoring every road surfaced 34 chronic bottleneck corridors that never appeared in the camera system.

Trafficure normalizes 15,000 or more road segments per city against the road graph and computes a baseline from around 90 days of history. Each segment carries its own normal, so an intervention is compared against the corridor's own past rather than a citywide average.

Because the network refreshes every two minutes and analysis runs a full cycle in under 120 seconds, the before and after is not a once a year survey. It is a continuous record the planning team can return to whenever a change goes in.

Trafficure is built on Google Maps Roads Management Insights, the aggregated anonymized movement of more than a billion smartphones, refreshed roughly every two minutes. The data is anonymized with k-anonymity applied at source, so it describes roads rather than identifiable people.

A corridor's before and after is read from that continuous feed, so the comparison is the road itself over time rather than a sample of vehicles on a single survey day.