Where every figure comes from, and what it can carry.

Except where the table above states otherwise, the datasets here are public sector information licensed under the Open Government Licence v3.0 ↗. Census 2021 estimates are Crown copyright, sourced through Nomis at the University of Durham. Place lookups use Nominatim, whose data is © OpenStreetMap contributors and available under the Open Database License ↗. Companies House data is used under the free company data product terms rather than the OGL. None of these publishers has reviewed or endorsed this site.

A record is published only when it can be checked against its source. A school needs an Ofsted identifier (URN), a name, a postcode, a recorded outcome and a link to the published report. A care location needs a CQC location identifier, a name, a postcode, at least one registered service type and a link to its CQC page. An area brief needs an official area code and name, Census population evidence and at least three of the four joined evidence layers. The same predicate decides whether a record appears in a directory, whether its page resolves and whether its address is in the sitemap, so those three can never disagree.

Addresses are built from the official identifier plus a slug derived from the current source name, never from a stored slug. A record whose name changes in the source therefore moves to a new address rather than sitting at a stale one, and the identifier in the URL always resolves.

A search resolves a postcode to its ward, LSOA, MSOA and local authority, and each figure is then shown at the level its source actually publishes. Census measures are local-authority figures, so they describe a whole council area rather than a street. Reported collisions are local-authority totals for the stated year, with no population or road-length rate inferred from them. Transport nodes are counted inside an approximate 10 km latitude and longitude box around the resolved point, which is a rough proximity count and not a travel time. Charity records are matched on postcode district. Ofsted and CQC use their own local-authority attribution, which is the education authority and the CQC-recorded authority respectively, and neither is guaranteed to match the council that collects the bills at that address.

Ofsted has inspected under several frameworks across the period this release covers, so a graded judgement, an ungraded “remains” result and a monitoring outcome are not one scale and are not mapped onto one. Around 295 school records in the held release carry no published inspection outcome, and those pages say so. The CQC directory does not include the quality rating, and roughly a quarter of registered locations carry no latest-check date at all; a missing date is a gap in the register, not a finding about the service. Provisional road-collision years are labelled provisional because they are revised. Census 2021 is a single point in time and is now several years old. Nothing here is a live status check: a page describes a published record on the date that record was published.

Each dataset is reloaded on its publisher’s cadence, which for these sources ranges from daily to once per Census. Every record stores the release it came from and when it was loaded, and that release string is printed on the page rather than kept in a changelog. Sitemap dates are the real load dates of the underlying data, not the date the sitemap was requested. Corrections reach this site only through the source: see About for how that works.