All 84 curated data pages in the 7 NDL collections were tested against 20 checks taken from the data asset management (DAM) policy, the critical data asset (CDA) metadata guidance, the data quality action plan (DQAP) guidance, the NDL's own curation criteria, FAIR principles and inclusion. 8 of the NDL's own chart data files were checked with 49 data quality rules. Every check is prioritised with the GOV.UK data quality issues framework, using the low end of each target range, as the framework's v1.0 tables do.
Most gaps appear on nearly every page, so one change to the page template fixes them everywhere. Each row treats a check as a quality rule measured across the pages it applies to. The black mark shows the framework target.
| Check | Source | Pages | Pass · part · fail | Performance vs target | % | Priority | Fix within |
|---|
Select a page name to see the evidence behind each result.
Collections differ by less than 11 points, which confirms that the gaps come from the shared page template rather than from individual curators.
The "Download the chart data" files are the NDL's own data products. Each was checked with rules for structure, validity, timeliness and whether the measure is described. Each also has a suggested trust level on the IOPC data trust scale, which says whether the data is fit to use externally. The NDL republishes data it doesn't produce, so the decision tree route is: measure, caveat and tell users, and pass fixes to the publisher.
Only 7 of 25 pages about people say which groups the data can be broken down by. A disabled person or a member of an ethnic minority group can't tell from the page whether data about people like them exists.
48 pages cover only part of the UK. 45 of them don't point to data for every other nation, and 28 point to none. The Early years collection is the model to copy: it links all four nations.
27 pages offer no open machine-readable format. 8 offer only proprietary spreadsheets, and 19 offer only PDFs, web pages or dashboards. These are harder to use with screen readers and on low bandwidth.
Each page's text was read on 14 October 2026, and its details were recorded in a structured form. Only what the NDL page itself shows was counted, because that is where users decide. Methodology is often one click away on the publisher's site. The licence, classification and owner-role checks are structural: no page has a field for them, so they apply strictly only to designated CDAs. Of 20 chart files, 8 were checked; UK house prices was too large to read whole. The code, data and rules are open, so the full assessment can be re-run with one command (python scripts/run_ndl_assessment.py).