Does the score actually predict anything?
A risk score is only worth using if last year's bands told you something about this year's incidents. We band every freight corridor using the older half of our data, then measure what actually happened over the following period — and publish the result whether it flatters us or not.
How this test is run
- Take the last 24 months of recorded vehicle crime along every scored motorway and strategic A-road.
- Use only the older 12 months to rank corridors and assign a band — no knowledge of what came next.
- Count what was actually recorded on each corridor over the following 12 months.
- Report the mean monthly incidents per corridor in each band, the lift against the network average, and the rank correlation between the two periods.
Banded on 2024-07 – 2025-06, measured on 2025-07 – 2026-06
| Band assigned | Corridors | Incidents / corridor / month | Lift vs average | Share of incidents |
|---|---|---|---|---|
| Critical | 6 | 17.2 | 2.42× | 60.6% |
| High | 5 | 8.6 | 1.21× | 25.2% |
| Elevated | 5 | 4.3 | 0.60× | 12.5% |
| Low | 8 | 0.4 | 0.05× | 1.7% |
A lift above 1.00 means corridors in that band went on to record more vehicle crime per month than the network average. A Spearman correlation near 1.00 means the ordering of corridors was broadly stable between the two periods — the ranking carried forward rather than being noise.
What this test does not show
This is a frequency test, not a severity test. Police records count a broken van window and a high-value load theft identically, so the bands describe how often incidents occur — not how expensive they are.
It is also not exposure-normalised: busier corridors carry more traffic as well as more crime. Treat the bands as relative intensity across the network, not as a per-vehicle probability.
Coverage gaps in the underlying police data carry through into the test. The full detail is on the methodology page.