How the Equillis Risk Score is built.
Every band shown in Equillis comes from published UK police data and a documented, rank-based model. This page explains the source, the approach and the limits so you can decide how much weight to place on a score. The detailed model mechanics, coefficients and calibrations are proprietary and shared under NDA with underwriting and procurement teams carrying out formal validation.
One national corpus, refreshed monthly
Equillis is built on street-level crime records published by UK police forces via data.police.uk under the Open Government Licence v3.0. We ingest the vehicle-crime categories — theft of a vehicle, theft from a vehicle, interference and tampering — geocode them to the published anonymised point locations, and hold them in a single national corpus of over a million records.
The corpus is refreshed each time a new monthly dataset is published. Every scored surface in the product shows the dataset version and the month it covers, so you always know how current a figure is.
Police point locations are deliberately snapped to anonymised map points, typically the centre of a nearby street or landmark. That means individual points are indicative rather than exact addresses — which is why Equillis scores areas, corridors and sites rather than pinpoints.
Rank-based, exposure-adjusted scoring
The Equillis Risk Score ranks an entity — a corridor section, a service area, or a postcode district — against comparable locations across the UK network. It combines recent recorded activity with long-run exposure, adjusted for the type of location and the amount of evidence behind it. The score always lands between 0 and 100.
Because the score is rank-based, it stays stable as the corpus grows. Adding new data sharpens the ordering rather than inflating every band. A score reflects relative risk: how exposed a location looks compared with the rest of the network, not a fixed probability that an incident will happen there.
The exact model structure, weights, smoothing parameters and calibration rules are proprietary. We share them under NDA with underwriting, procurement and validation teams.
What each band means
Bands are relative positions against the scored network, not absolute thresholds. Critical is deliberately scarce — only a small fraction of scored sections reach it at any one time.
| Band | Interpretation |
|---|---|
| Low | Recorded vehicle crime is light relative to the rest of the network. |
| Elevated | Measurably busier than average. Worth avoiding for unattended overnight stops. |
| High | Consistently among the busier sections nationally. |
| Critical | Among the highest-intensity sections in the country right now. |
For corridors the underlying measure is intensity rather than a raw count, so a long rural section is not penalised for its length, and a corridor that has quietened down is not judged on last year's activity. The model also accounts for local background levels and freight-relevant context, so strategic freight routes are compared against the places freight actually stops, not against residential car-crime hotspots.
What the score cannot tell you
- It measures recorded crime. Under-reporting is real, particularly for attempted entry and fuel theft.
- Point locations are anonymised, so scores are meaningful at area and corridor level, not at a single address.
- It is not a prediction of whether your vehicle will be targeted tonight; it is a comparison of relative exposure.
- Forces publish on slightly different schedules, so the most recent month can be incomplete for some areas.
Ask us to show our working
If you are assessing Equillis for underwriting, procurement or operational sign-off and want to interrogate the model against your own claims or incident history, get in touch — we will walk through the calculation on your own sites and lanes.