Guide
Building software that depends on school term dates
Updated
Term dates look like reference data and behave like user-generated content: thousands of independent publishers, no schema, no change notifications, and a real cost when a record is wrong.
Talk to a specialist Six sources on coverage, authority and machine readability, with the statutory reason no seventh exists.
Enumerate before you aggregate
The Department for Education's Get Information about Schools register is machine readable, comprehensive for England, and downloadable in bulk. It holds establishment records rather than calendars.
That gives you the denominator and, crucially, each school's governance type, which under section 32 determines whether its dates come from the council or from its own governing body or trust. Doing this first turns a vague scraping problem into a bounded one.
Design for staleness explicitly
Every date you hold should carry where it came from and when it was last confirmed. Without those two fields you cannot distinguish a date verified this morning from one scraped two years ago, and neither can your users.
Surface both. A date shown with its source and its age is honest; a date shown bare implies a confidence nobody can support.
Decide what wrong looks like
For a marketing calendar, a wrong half term is an annoyance. For anything scheduling childcare, transport or staffing around a school day, a wrong inset day fails a family on the morning.
That difference should drive the design. A high-stakes feature needs a confirmation step against the school's own page, or it needs to not exist.
Four nations, four models
England, Wales, Scotland and Northern Ireland publish differently, and gov.uk's tool covers England and Wales only.
Decide the scope early. Retrofitting Scotland and Northern Ireland into a model built around English councils is a rewrite rather than an extension, because the structures and the terminology differ.
Say what your coverage is
Nobody in this market publishes a verifiable coverage figure, and the reason is that the denominator is every school in the UK and the register of those is a separate dataset.
If you build something, state your coverage against that register and state your freshness policy. It is a low bar and almost nothing on the market currently clears it, which is the actual gap here.