Skip to content

Verified catalogue operations

The Careers catalogue is a platform-operator workflow for approved school, pathway, subject-combination, occupation and requirement data. The first release is API-only: it does not add an import screen to a school’s ShuleOne account. Use only the protected platform integration and an authorized staff identity.

Record all of the following for every immutable dataset snapshot:

  • a stable dataset key and a new dataset version;
  • source name, publisher and an absolute source URL;
  • the source publication date, when available;
  • when the source was retrieved and when it was last verified;
  • the usage-rights basis: official public record, open licence, public domain or specific permission granted;
  • the completed access/terms review: approved for republication or approved for a factual index.

Do not upload or republish restricted or paywalled material without permission. An accessible URL is not by itself a licence. Keep the underlying review or permission evidence in the approved records system; do not put passwords, tokens or unrelated personal information in catalogue data.

An import can contain up to 2,000 records. Every record needs a type, stable key, name, status, confidence, verification time and a reference that identifies the source row or section.

Supported review statuses are Verified, Provisional and Rejected. Supported confidence levels are High, Medium and Low. Verified does not mean permanent: it means the evidence passed the current review. A low-confidence record can generate a warning and needs special scrutiny before publication.

School records also require an official code and county. Add sub-county, ownership, gender, accommodation, pathway codes and subject combinations only when the cited source supports them. Pathway records must use one of the supported ShuleOne pathway codes. Subject-combination records must identify their pathway. Occupation and requirement records can be retained in the versioned foundation, but the first public endpoints expose only published schools and pathway references.

Use a record-specific verification time when it differs from the dataset. Do not turn a missing value into a negative claim. For example, a missing subject combination means “not published in this snapshot”, not “the school does not offer it”.

Submitting a snapshot creates an immutable import record and calculates its SHA-256 payload hash.

  • Valid data becomes a Draft. Review the counts, source, hash, validation results and every warning before publication.
  • Invalid data becomes Validation failed. Its errors remain in the audit record, and none of its catalogue rows are saved or made public.
  • Submitting the same dataset key, version and payload again is an idempotent replay. Reusing that key and version with different content is rejected; correct the data and issue a new version.
  • Only a valid draft containing at least one Verified record can be published.

Publishing makes that snapshot current and retires the previously published snapshot with the same dataset key. Provisional and Rejected records remain out of the public projection. There is no general delete or manual-retire API in this increment; use a reviewed successor version for ordinary replacement and follow the incident process if unsafe data must be withdrawn urgently.

After publication, verify the Careers compatibility response and the public school and pathway endpoints without relying on an administrator session. The school directory capability becomes available only when at least one Verified school record exists in a Published snapshot.

For a sample of records, compare the public official code, location, pathways, subject combinations, publisher, source URL, dataset version, verification date and confidence with the reviewed snapshot. Check that Provisional and Rejected records are absent, filters do not cross datasets incorrectly, and a replaced snapshot no longer contributes results.

The public explorer is not a complete national directory unless the published, reviewed datasets actually provide that coverage. Record coverage and source-date limits with each release; never infer completeness from a successful import.