Crime — methodology

1. Data sources, licence and freshness

The crime section draws from two sources: the live Police.uk public API and a small set of Home Office-derived reference statistics held in our database for comparison.

• Street-level crime reports — fetched live from the Police.uk crime data API. Each call returns reported crimes within a 1-mile radius of the supplied latitude/longitude for a single calendar month. We request a rolling 24-month window, in 8-wide concurrent batches, with a per-call timeout of around 7 seconds.
• Outcomes — fetched live from the Police.uk /outcomes-at-location endpoint for the most recent published month, to give a directional indication of how recent reports were resolved.
• Local authority and national averages — Home Office Police Recorded Crime, loaded into our database as annual rates per 1,000 residents per local authority, plus an England-wide average.

Licence: Police.uk data is published under the Open Government Licence v3.0. Contains data from data.police.uk, the Home Office, and contributing police forces.

Freshness: Police.uk publishes one calendar month at a time, on a delay of around two months. We conservatively skip the three most recent months and start the 24-month window from the most recent month that is reliably published. If you generate a report in early May, the most recent month included will typically be February.

2. What we do

• Resolve the postcode to a centroid via Postcodes.io, then query Police.uk for every month in the 24-month rolling window using that point.
• Aggregate the totals into a single category breakdown (anti-social behaviour, burglary, violent crime, vehicle crime, etc.) ordered by frequency.
• Build a monthly trend line so the recent shape of reported crime is visible at a glance, and a recent-90-day breakdown for the latest signal.
• Compute a direction (rising / falling / stable) by comparing the most recent half of the window to the earlier half. We flag the trend as unreliable and hide the percentage when the sample size is too small for the change to be meaningful (e.g. 2 → 4 crimes is noise, not a trend).
• Where local authority population and area data is available, compute an approximate annualised crimes per 1,000 residents rate inside the 1-mile disc using local authority population density as a proxy denominator. We then compare it like-for-like to the LA and national rates.
• Where the per-1,000 rate cannot be computed, fall back to a directional comparison against the LA rate ÷ 12 and clearly mark it as approximate.
• Surface each missing month or missing comparison as an explicit data gap so you can see where the picture is partial.

3. What we don't do

• We do not measure unreported crime. Police.uk only reflects what the public reported and the police recorded. National victimisation surveys consistently show this is a fraction of all incidents.
• We do not show crime at a single address. Police.uk anonymises locations to a "snap point" (typically the centre of a street, junction or area), and our 1-mile disc is deliberately broad. Two adjacent streets within the same disc can have very different lived experiences.
• We do not characterise people, demographics, or "rough" / "safe" areas. We report category counts, trend, and rate. Interpretation is up to the reader.
• We do not provide live or near-live alerts. The most recent month in the window is normally two to three months behind today.
• We do not assess specific offence severity. A "violent crime" category covers a very wide spectrum, and Police.uk does not publish severity.
• We do not include British Transport Police, Action Fraud, or MoD Police data. Police.uk's public API is geographic territorial-force data only.

4. Known limitations

Read the crime section as a directional summary, not a measurement instrument.

• The window is 24 months. We previously fetched up to 60 months but cut to 24 to stay reliably within our generation budget and to keep the trend grounded in the recent past, where reporting practices and category definitions are most consistent. A 24-month window is enough to see a real shift but is not a long-term historical baseline.
• Snap-point geography. Police.uk locations are deliberately approximate to protect victim privacy. A reported crime "on" a street may have occurred in the wider area; the headline figure inside our 1-mile disc is the relevant unit of analysis, not the street-by-street pin map.
• Rural vs urban density. The same 1-mile disc captures very different population sizes in a city centre versus a hamlet. Where we can compute a per-1,000-residents rate using local authority density, we prefer that figure for comparisons; where we can't, we say so and mark the comparison as directional only.
• Density proxy is approximate. Our per-1,000-residents rate uses local-authority-wide population density as the denominator inside the disc. This assumes uniform density across the LA, which is rough but materially better than comparing raw counts. LSOA-level density is on our roadmap.
• Recording practices change over time. Crime category definitions and recording standards evolve, so very long-window comparisons should be treated with care — another reason we cap the window at 24 months.
• Partial months happen. Police.uk occasionally fails to return a month within our timeout. We record that as a zero-count entry and surface a data gap so the trend is not silently distorted.
• Outcomes data is partial. Many cases remain "under investigation" or "no further action"; the outcomes summary is directional, not exhaustive.

5. What this is not a substitute for

6. Further reading

• Methodology overview — how the whole report is built.
• Report disclaimer — limits of liability and the scope of all sections.
• Data sources & attribution — full licence detail for every source we use.
• police.uk — the public-facing service for street-level crime and your neighbourhood policing team.