This paper presents a live evaluation of incident detection within SCOOT® 8AI in Manchester. The work forms part of an ongoing collaboration between Transport for Greater Manchester (TfGM) and TRL Software – and addresses a difficult practical problem:
how can an incident detection be evaluated when there is no complete incident log or manually confirmed ground truth?
A matchday event on 28 January 2026 provides a real-world case study. Police escorted supporters from Manchester city centre towards the Etihad Stadium, crossing the southern part of the Ancoats region and creating a recognisable disturbance in the traffic data. We use that event to examine the temporal and spatial signature of disruption, then consider the wider behaviour of the incident detection. The result is not a claim of measured incident accuracy. It is a practical method for reducing noisy detector output into a smaller set of physically credible candidate incidents for operator attention, supported by evidence that the strongest filtered events follow recognisable traffic patterns.