Openness comes as a PDF: what format says about how usable this data really is
Two things become visible once you look past individual releases to the shape of the dataset as a whole: what format the actual data arrives in, and which part of the law gets reached for to stop it arriving at all. Neither is uniform.
PDF is the format of UK government transparency
Across every attachment we've archived and analysed - more than 35,000 files - 86% are PDFs. Genuinely machine-readable CSV, the format you could drop straight into a spreadsheet or a script without a conversion step, is 2.5%. Word documents and spreadsheets make up most of the rest - together over a tenth of everything released.
This matters because "the data was released" and "the data is usable" are different claims. A statistical table trapped in a scanned or exported PDF has technically satisfied a disclosure obligation while remaining, in practice, something a journalist or researcher has to manually retype before they can do anything with it. Format is itself a transparency choice, even when nobody frames it that way.
Which exemption gets used tracks the subject, not just the requester
Cross-referencing exemption citations against our topic tags shows real variation by subject, not just noise:
| Topic | Security / law enforcement exemptions | "Not held" |
|---|---|---|
| Counter-Terrorism | 12.5% | 16.7% |
| Firearms & Weapons | 0.0% | 22.7% |
| Prisons & Probation | 5.1% | 12.8% |
| Immigration & Asylum | 0.7% | 3.9% |
| Police Misconduct & Complaints | 0.0% | 8.8% |
Counter-Terrorism is the one topic here where security-related exemptions (sections covering law enforcement, investigations, and security bodies) show up often - unsurprising in substance, but useful as a sanity check that the exemption data reflects something real rather than being noise. Firearms & Weapons requests, by contrast, are essentially never stopped by a security exemption in our sample - when they don't get answered, it's because the information wasn't held, not because it was actively withheld. Immigration & Asylum, the largest single topic in this tracker (see our earlier piece on it), has the lowest "not held" rate of any topic here by a wide margin - consistent with it running mostly through the Home Office's well-established central process rather than scattered smaller bodies.
Some of these topic samples are on the small side (20-40 documents for the rarer ones), so we'd treat the smaller percentages as suggestive rather than settled - Counter-Terrorism's 16.7% "not held" rate, for instance, is 4 releases out of 24, not a robust base rate.
Reading this responsibly
Format figures cover every attachment we've archived to date, regardless of topic. Exemption figures are keyword-presence rates - a document mentioning a section number isn't necessarily one where that section was the operative reason for withholding, and topics with small sample sizes shouldn't be read as precisely as the ones with hundreds of documents behind them.