Public records. Evidence. Accountability.United Kingdom ● Explore the record
The CartularyA record of the public realm

The Cartulary

Methodology

Our standards for sources, calculations, information requests and responsible publication.

Public records should be understandable, traceable and open to correction. This page sets the standards for evidence published by The Cartulary. Each collection must also explain its own coverage, calculations and limitations.

What we examine

Our scope is how British public institutions spend, decide and perform, and how they respond when asked to account for their work. The Ledger concerns public money. The Atlas concerns places and outcomes. The Docket concerns information requests and responses. Briefings explain what the evidence supports. The Evidence Library provides the source record across these areas.

Coverage is collection-specific. A record about one authority, period or jurisdiction must not be presented as a finding about the whole country.

Start with the source

Use original publications wherever possible: official statistics, procurement notices, spending releases, registers, reports, recorded decisions and disclosed correspondence. Identify the publisher and link to the original record. A secondary account can point to evidence; it does not replace checking that evidence.

For every published evidence record, record the source URL, publisher, relevant period, publication date where available, retrieval date and reuse conditions. For processed data, also record the transformation or calculation method, its version and the last verification date. Where information is unavailable, say so.

Copies of documents should retain their original content. Any extract, transcription or redacted copy must be labelled. Where a preserved file and verification hash are available, publish those details; never imply that a linked document has been independently archived when it has not.

Describe the evidence precisely

Separate what a source records from an allegation, an interpretation and a conclusion. An official record establishes what its publisher reported; it may still contain errors, omissions or later revisions.

Use these labels consistently:

  • Official data: a value reported by the named public source, with its date and definition.
  • Cartulary calculation: a derived value, with its inputs, method and rounding explained.
  • Source statement: a claim attributed to its speaker or document.
  • Analysis: an interpretation supported by identified evidence and stated limitations.
  • Not available in this collection: a coverage gap, not a finding that the information does not exist.
  • Awaiting disclosure: use only after a request has actually been submitted, with a link to its record.

A blank cell is not zero. A missing publication is not proof of concealment. An unanswered question is not an established allegation.

Make comparisons reproducible

Show units, periods, geographical boundaries and definitions beside the data. Explain denominator choices for population rates. Distinguish a stock at a point in time from activity over a period. Check for revisions, missing values and changes in collection methods before comparing figures.

Use shared identifiers to link records where possible. A similar name alone is not enough to establish that two companies, people or organisations are the same. Flag uncertain matches and keep them out of conclusions that require a confirmed identity.

Show the date of the underlying observation separately from the date a page was updated. A page refresh does not make an old dataset current. Modelled rates must be visibly labelled and must not be described as live measurements.

Scrutiny indicators are questions to examine

Any procurement or transparency indicator must publish its definition, coverage and calculation. An unusual value or procedural feature can justify further examination. It does not, by itself, establish corruption, misconduct or poor value.

Geographical overlap does not establish causation. Data about asylum support, nationality, immigration status and crime must retain their distinct definitions. Do not infer a relationship between people or events from separate area-level datasets.

Request information with care

Before submitting a request, check existing publications and earlier requests, identify the records needed and narrow the scope. A person must approve the request and the appropriate information-rights route before submission.

Each Docket entry must identify the authority, applicable regime and jurisdiction, actual submission date, correspondence, status and next action. Record the basis of any displayed deadline and any clarification or extension affecting it. Do not apply a single deadline rule to every request. Escalations require review of the response and the relevant procedure.

Publish in the public interest

Material must have a clear connection to the question being examined. Review personal information, source permissions and restrictions before publication. Limit unnecessary identifying detail, especially about children or private individuals.

Where a publication makes a material adverse factual claim about an identifiable person or organisation, seek a relevant response and represent it fairly. Explain material limitations and unresolved uncertainty alongside the finding.

If material is redacted or withheld, explain the reason at a useful level without exposing the information being protected. Link corrections to the affected record so that later readers can understand what changed.

Interactive evidence

The console uses three visible provenance labels: ● Official for published figures (forecasts are explicitly identified), ◐ Modelled for calculations, and ◇ Disclosed for information provided through a request. Each calculation links to inputs and method. Moving counters can be paused and respect reduced-motion preferences.

The spending wheel keeps OBR forecasts and HM Treasury service outturn in separate views. The Atlas retains historical council codes and boundary versions, reports unmapped counts, and identifies changes in source coverage. A gap in a source is not treated as zero. FOI charts retain the source’s definitions and the distinction between allocations and expenditure.

Read the console calculations and their limits.

Historical arrests and ethnicity

The separate Arrests & ethnicity collection covers England and Wales, 2020/21–2022/23. Arrests are not convictions or unique people. The filters keep time periods, demographic totals and ethnicity classification schemes separate; missing values are not assumed to be zero. Source notes explain recording changes, population denominators and the limits of comparisons.

Privacy edits to disclosed documents

Reviewed FOI copies conceal requester-identifying details, correspondence links and unique case references. Response letters are presented as page images and flattened PDFs, with redactions labelled. The substantive response remains unchanged. Original provenance is retained privately for editorial verification; it is not included in public downloads.