Roadmap

Source schools from official government data, not each LA's own directory

Idea

Use GIAS (Get Information about Schools) bulk data as the canonical school source, and an algorithm to surface candidate out-of-area schools per LA, instead of manually scraping each LA's own "out of district" category by hand.

Bradford's Local Offer has an "Out of district education provision" category - 85 schools Bradford places children in outside the district, mostly independent/specialist SEND schools reached via EHCP. Importing it (August 2026) meant one GIAS lookup per school via the browser - real but slow, and it only works because Bradford happens to maintain that list themselves. Other LAs may not.

GIAS publishes a daily-updated bulk download of every school in England (URN, address, LA, phase, type, status) - a local cache of this would replace one-lookup-per-school with one download, and make URN a real join key instead of a stored string.

The harder half is deciding which schools are actually relevant to a given LA - and it splits into two genuinely different problems:

Mainstream schools near a border - a family near the edge of an LA whose nearest school is technically in the next authority. This isn't really an LA-level property at all; it's about where a specific family lives. Better served by a live "schools near this postcode" search than a permanent tag.
Specialist/independent placements - the Witherslack, Acorn, Cambian, Mary Hare School kind of entry. These aren't distance-driven (Mary Hare is 200+ miles from Bradford) - what actually determines relevance is whether an LA places children there, which only the LA's own knowledge (or directory) currently holds. Geography can't manufacture that signal from scratch.

So the realistic shape: use a GIAS-backed geographic pass (radius, or bordering LAs, weighted toward special/independent/non-maintained-special types) to generate candidates worth checking - the same shape as the Import queue - not an algorithm that auto-decides inclusion. A proprietor-name grouping (Witherslack Group, Acorn Group etc. all share a field in GIAS) is a free extra: confirming one site is relevant to an LA is a hint worth checking against its sibling sites nationally.

Worth building once several more LAs are being onboarded and the one-school-at-a-time browser lookup pace becomes the actual bottleneck - not before. At 4-5 LAs the manual per-LA-directory pattern still worked fine.