Where every figure on this site comes from

How the pipeline works

There are four steps, and they always run in the same order. First, a fetch stage pulls each dataset from its publisher — an approved-programme list, a pay circular, a bursary scheme’s published rules, a workforce statistics release — and stores the raw response with the date it was retrieved. Second, a normalisation stage maps each publisher’s own vocabulary onto the site’s: a course is tied to a profession slug and a provider slug, an employer is tied both to the ITL1 region that course facets use and to the NHS region its publisher assigns it, and a pay point is tied to a nation and a pay year.

Third, everything is validated against a schema in which provenance is compulsory. Any externally sourced value is wrapped in a structure that requires a source identifier, and a value that arrives without one fails the build rather than being published unattributed. The same schema makes almost every field optional, which is the deliberate half of the design: the pipeline omits what it cannot verify instead of defaulting it. Fourth, the validated result is written as a single build artefact, and the site is rendered entirely from that file — so every page in this deployment is guaranteed to be describing the same dataset as every other page.

What “not published” means

It means the field is genuinely absent from the source. A course with no tuition fee shown has no published fee in the data the regulator or provider released; a profession with no registrant count has none the register publishes separately. It does not mean zero, it does not mean the value is small, and it is not a placeholder for something the site will guess at later.

Two related labels are worth knowing. Where a figure covers more than one nursing field — the occupational classification keeps a single SOC code for more than one field, and the NMC’s own headline nurse count does not split cleanly by field because a nurse can hold registration in more than one — the page marks it as a shared figure and names the group, because splitting it would be invention. And where a source permits linking but not republishing, the site links out instead of copying the value across.

The registry sorts every source into three states. 4 are open — Open Government Licence or an equivalent — and can be reused with attribution, which is what the source block at the foot of each page provides. 7 are conditional: reusable, but only on stated terms such as a required attribution line, a link back to the publisher, or registration for an API key.

Because the status is stored with the source rather than remembered by a person, a restricted source cannot quietly end up feeding a page: the validator refuses a build in which it does. The source registry lists all of them, grouped by status, with the conditions written out.