Street level crime data UK: every incident row, delivered as one table

The query reads like a hunt for a file; it is really a routing question between three record families that happen to live under one roof. data.police.uk - the crime-and-policing publication of the UK's Single Online Home National Digital Team - carries all three: the crime table, one anonymised row per incident; the outcomes table, the police and court result that followed; and stop-and-search at the level of the individual search. Contributions arrive from the 43 geographic forces of England and Wales, the British Transport Police and the Police Service of Northern Ireland, with the Ministry of Justice feeding the outcomes layer.

The scale surprises people in both directions. A whole month of national incident data compresses to roughly 57 megabytes, and one May 2026 snapshot cut around a single Midlands city-centre coordinate returned 1,593 records - small enough to move, dense enough to rank every street in a town. Each row carries a category enum in which burglary sits beside robbery, vehicle crime and criminal-damage-arson, so territory work reads straight off the rows rather than out of a press release.

Inside Datadory's security & alarm services pool, that record scores 9 out of 10 against a slice average of 7.23 across 13 datasets and a catalog average of 7.81 across 1,744 - the anchor of the UK half of the pack.

One row per incident - the shape a delivery lands in, pinned to live records:

# street-level incidents - one row per crime
category      : anti-social-behaviour         month        : 2026-05
location_type : Force                         id           : 135145675
latitude      : 52.631711                     longitude    : -1.144995
street        : On or near Thorpe Street

# the same incident spine, with its latest recorded outcome attached
category         : criminal-damage-arson       month        : 2026-05
street           : On or near Marquis Street
outcome_category : Investigation complete; no suspect identified
outcome_date     : 2026-05
persistent_id    : b5da11998ae66a2187077e9330b652f366ed55fd41e96ae37697f20de57575d1

# stop-and-search - one row per individual search
type                      : Person search       datetime : 2026-05-30T16:30:00+00:00
age_range                 : 18-24               gender   : Other
officer_defined_ethnicity : White
legislation               : Misuse of Drugs Act 1971 (section 23)

A dozen fields divide cleanly into four jobs. Identity: category names the offence and persistent_id is the 64-character identifier that stays with a crime across republications, where the plain integer id is not guaranteed constant. Time and place: month fixes the year-month, the latitude-longitude pair gives the anonymised point, and the street id and name locate it in human terms - "On or near Thorpe Street" is deliberate anonymisation, not missing data. Provenance: location_type separates territorial-force records from British Transport Police records, and the flat-file layout adds the reporting force plus the 2021-boundary LSOA code and name. Result: outcome_status.category and outcome_status.date carry the latest outcome, which is what turns a burglary map into a detected-versus-undetected map without a second file.

Wider views ride alongside on request: neighbourhood teams, events and priorities, force profiles with senior officers, crimes-at-location cuts, and the reference dimensions - the full 44-force roster and the crime-category list - each folded into the same dictionary once named.

• Geography: England, Wales and Northern Ireland, organised by police force - 44 of them counting British Transport Police and the PSNI - and drillable down to streets and 2021-boundary Lower Layer Super Output Areas. Scotland appears only where the British Transport Police operates, which matters if your territory map crosses the border.
• Time: monthly snapshots running from July 2023 through June 2026 as published at research time, with older months retained in the archive - a three-year seasonality read and a single-month hotspot sweep draw from the same well.
• Granularity: individual incidents, not aggregates. Every row is one crime or one search, anonymised to a point snapped onto a street and its containing LSOA - fine enough for catchment analysis, coarse enough to be publishable at all.

That stack - force rollups that nest honestly down to individual offences, with the LSOA as the census-grade join key - is why this record anchors UK demand modelling in the pool: LSOA-scale counting needs no resampling because the grain was decided upstream.

• Territory targeting. Rank postcodes, towns and force areas by residential burglary intensity before committing dealer-recruiting budget; the LSOA column sorts territories without another vendor spreadsheet. The sales and growth teams playbook covers the motion end to end.
• Patrol and alarm-response planning. Route response capacity toward the streets where incidents actually cluster, and set branch siting against neighbourhood evidence instead of force-wide averages.
• Market sizing. Cut burglary and vehicle crime by force area and catchment to size installation and monitoring demand; pair the offence base with Census establishment counts and the market researchers workflow shows the sequence.
• Outcome-rate analysis. Because each crime travels with its latest outcome, attach rates by force become arithmetic on one table - the ratio that separates markets where police respond from markets where residents buy hardware instead. Quant researchers model it explicitly; see the investors and quants patterns.
• Model entity spines. Force, LSOA and street form the geographic spine that joins incidents to demographics, property registries and competitor locations; the data scientists patterns detail the joins.

Three coverage edges change what a claim can rest on, and all three are knowable before a build starts.

• The PSNI contributes no stop-and-search records, so any search-rate analysis is England, Wales and British Transport Police only.
• Outcomes do not ride on British Transport Police or PSNI records, so detected-versus-undetected comparisons must exclude both forces rather than treat absence as zero.
• Scotland appears only through the British Transport Police - Scottish territorial forces sit outside the collection, so a GB-wide map has a visible seam at the border.

The fourth edge is a design choice rather than a gap: locations are approximate by construction. Anonymised points snapped to streets and LSOAs are what make premises-level grain legally publishable nowhere in the UK estate - and the same convention governs US city equivalents, where LAPD rounds addresses to the hundred block. If a model needs parcel precision, name it when requesting the sample and the answer states plainly what the grain supports.

Two companion records complete the UK picture without changing the workflow: the NSI company finder lists more than 2,000 approved installers with approval tier and police-response region, and the SSAIB approved company directory covers the UK and Republic of Ireland across roughly three dozen certification schemes. The crime layer sizes demand; the certification registers verify supply.

The comparison decides which side of the Atlantic a pipeline starts on, and it is a question of job, not of quality.

Read the table as four different instruments. If the question is which streets in Leeds saw burglaries last month, only the UK feed answers it - nothing else in the pool publishes individual incidents at street and LSOA grain for a whole nation. If it is how did burglary trend in Ohio since 1975, the FBI Crime Data Explorer owns the long aggregate view, because decades-deep counting is what its six-field agency-year-offense rows are built for. If it is block-level demand modelling in one American city, the LA ledger supplies 1,004,894 geocoded incidents on a stable 28-column schema - inside a fixed 2020-to-2024 window that will never revise underneath you. And if the question is the crime nobody reported, only the National Crime Victimization Survey measures it - roughly 240,000 people in about 150,000 household interviews a year, with a flag recording whether police were told.

Grain wins arguments that history cannot. A force-level count can tell a UK alarm business that burglary fell; only the incident rows tell it where to send the next van. The two sides meet head to head in data.police.uk vs FBI Crime Data Explorer, and the US-side question gets its own treatment in is there a free api for US crime statistics.

Because the incidents were never the hard part - the seams are. Three record families mean three schemas, two identifier conventions, an outcome stream that lags its crimes and a force roster whose members populate different columns. Receiving them is a query.

Datadory normalises before delivery: one field dictionary across the shelf, force plus LSOA plus street as the join keys, months typed as dates, categories held constant, outcomes already joined to their crimes and provenance riding beside every value so any row traces back to the snapshot it came from. Hotspot math becomes a WHERE clause instead of an afternoon with a spreadsheet.

This page covered the UK street-level shelf - what the rows contain, how far they reach and where the edges sit. Work through these next:

1. data.police.uk - the product page: full field dictionary, verified sample rows and coverage chips for the 44-force footprint.
2. Security & Alarm Services data guide - all 13 pooled records mapped, from UK incident rows to NAICS 561621 market sizing.
3. Best security alarm services datasets - the scored short version of the leaders.
4. data.police.uk vs FBI Crime Data Explorer - the grain-versus-history argument settled row by row.
5. Is there a free API for US crime statistics - the transatlantic counterpart question.
6. [Security & alarm services data hub](" + IND + ") - every field dictionary and coverage profile in one inventory, ending at a sample request.