Preventing Incidents Before They Happen With Predictive Analytics
City safety boards had long relied on incident reports to identify dangerous intersections, meaning action only followed after someone had already been hurt. In 2025, a predictive foot-traffic modeling initiative flipped that sequence, combining historical incident data with live pedestrian counts to flag high-risk crossings before they produced another statistic.
The Challenge
Signal timing across the city had been set once and rarely revisited, even as pedestrian volumes and traffic patterns shifted over time. Safety boards could point to intersections with a history of incidents, but had no forward-looking way to identify the next one before it happened.
Our Approach
- Combined historical incident records with live pedestrian and vehicle counts across 40 intersections.
- Built a risk-scoring model to flag crossings trending toward dangerous conditions before an incident occurred.
- Retimed signals at flagged intersections in coordination with the city's traffic engineering team.
Key Results
- 40 intersections retimed based on predictive risk scoring.
- High-risk crossings flagged proactively, ahead of incident data alone.
- A repeatable model now used to prioritize the city's ongoing signal-retiming programme.
The shift from reactive to predictive safety planning has changed how the city allocates its retiming budget, prioritizing intersections trending toward risk rather than waiting for incident reports to justify the spend.
