How the figures are calculated

Every figure here comes from the monthly street-level files published through data.police.uk by the Single Online Home National Digital Team, on behalf of police forces. Case results come from the Ministry of Justice where they are supplied. We accept dated monthly archives, keep their bytes, and calculate from the copies we kept.

Counting is the part that needs care. A later release can republish the same crime with a different month or a different category, so we hold each identifier once inside the twelve-month window and keep its latest published month. Where one month offers two versions of the same record, we take the last in a fixed alphabetical order, because no revision time is supplied and the true order inside a month is unknown. Exact repeats are not treated as changes. Anti-social behaviour rows carry no identifier, so their row counts are kept as published.

A record count is exact and carries no interval. Shares carry a Wilson 95 percent interval and their own denominator. A difference between two shares has no interval of its own; the two shares keep theirs.

Case results use the twelve months before the window, exclude anti-social behaviour, and take the last known status at the release date. Results need at least 50 records with a result before any share appears.

Each crime identifier is counted once in September 2025 to August 2026. Republished identifiers: 22,189 which is 0.4% of the window. Above 2 percent, the build stops for review. At least one force month arrived without a results file. Those results are unavailable, not zero.

Each crime identifier is counted once in September 2025 to August 2026. Republished identifiers: 22,189 which is 0.4% of the window. Above 2 percent, the build stops for review. At least one force month arrived without a results file. Those results are unavailable, not zero.