Methodology · scoring
Safety score
Lower severity-weighted crime → higher score. Each Police.uk category is weighted by its impact on personal safety — a robbery counts 20× more than a shoplift — before percentile-ranking nationally.
Formula
The maths
safety = 100 − percentile_rank_nationwide(Σ rate_i × severity_weight_i)
Weights
Why each input gets the weight it does
| Input | Weight | Rationale |
|---|---|---|
| Violent crime, robbery, weapons | 1.00 | Direct threat to body or coercion. Anchors the index. |
| Burglary | 0.70 | Home invasion — high anxiety even when no contact occurs. |
| Theft from the person | 0.60 | Mugging / pickpocketing — confrontational but lower body-harm risk than violent crime. |
| Criminal damage / arson | 0.40 | Visible disorder; affects perceived safety walking the streets. |
| Drugs, public order | 0.35 | Visible disorder, lower harm tier. |
| Anti-social behaviour | 0.30 | By far the most-reported category; mostly noise / rowdiness. |
| Vehicle crime, other theft, other crime | 0.20 | Opportunistic property crime. Annoying, but doesn't make residents feel unsafe. |
| Bicycle theft | 0.10 | Low fear factor. |
| Shoplifting | 0.05 | Commercial theft. Inflates totals in retail districts but barely affects residents — a Tesco shoplift on Oxford Street shouldn't drag down the safety score for the people living above it. |
Data sources
What feeds this score
- Police.uk Open Data — 14 incident categories per local areaGeocoded incident → Local area
- ONS local-area population estimatesLocal area
- MHCLG English IMD 2019 — Crime sub-score (Greater Manchester proxy)Local area
Limitations
What this score doesn't capture
- The severity weights are editorial, not regression-derived. They reflect a defensible ordering of categories by impact on perceived personal safety; the absolute values matter less than their ratio.
- Greater Manchester Police stopped publishing to data.police.uk in mid-2020. We fill GM with an IMD-derived proxy: the national percentile rank of the 2019 IMD Crime Score, inverted to match the 0–100 'higher-is-safer' direction. Every page where this proxy is used carries a visible note.
- We still don't adjust for transient population — a city-centre area with 200 residents and a busy nightlife strip can look unsafe by per-1,000-residents normalisation. Severity weighting reduces but doesn't eliminate this distortion: weapons and violent-crime rates are still high in nightlife districts because they really do happen there, even if the offenders aren't local.
Last methodology change: 29 April 2026. See /changelog for prior revisions, or contact hello@placetrics.co.uk with corrections.