Journal
Collision clusters: staged patterns versus a bad junction
Motor books throw up maps with hot spots. A roundabout on a ring road, a retail park exit, a particular slip road: collisions repeat because traffic repeats. Treating the hot spot as a fraud cluster wastes SIU and annoys honest policyholders who will be asked odd questions.
A staged-collision pattern, when we have seen one worth writing down, usually needs more than geography. It needs a pairing: the same recovery firm, the same hire desk, the same medical-legal, arriving in a tight window, often with similar passenger counts and similar damage descriptions. Geography alone is the junction.
The extract rarely proves staging. It can show that a pairing is unusual against the rest of the book. The briefing should use that word — unusual — and list what an investigator would still need: a statement inconsistency, a repairer who never saw the car, a hire that started before the reported time of loss.
We also chart time of day. Thursday-night and weekend clusters have different honest explanations (shift workers, leisure traffic) and different dishonest ones. Mixing them in a weekly total hides both.
If your geocoding is the postcode of the policyholder, not the collision, the map is a map of who buys your policy. We will not draw it. We will ask for a collision location field, even a messy one.
The honest sentence for a committee is often: this junction is expensive; this pairing is worth five files of SIU time; the rest is coincidence until someone reads the statements. That sentence is less exciting than a heat map. It is also harder to reverse in six months.