The Skills England occupational maps, as an ontology

The occupational maps, as a graph

Skills England publishes the national occupational maps through a public API: occupational standards arranged into routes, pathways and clusters, each carrying a Standard Occupational Classification mapping, apprenticeship and technical education products, and a green-jobs classification, with progression relationships between occupations. That is a graph. Delivered as JSON over REST endpoints it reads as documents, and consumed the usual way a slice of it ends up flattened into a spreadsheet where the relationships that make it valuable are lost.

We turned the whole thing into what it already is: a single, connected, machine-readable graph. Every occupational standard is a node; its route, pathway, cluster, technical level, status, SOC codes, products, green themes and progression targets are typed edges. The result is 51,355 statements that validate against a formal SHACL schema with zero violations, published openly as Turtle and JSON-LD.

The challenge

Occupational standards sit at the join between three systems that rarely share a data model: the education system (apprenticeships, T Levels, Higher Technical Qualifications), the labour market (ONS Standard Occupational Classification, used for every official statistic on jobs and pay), and skills policy (routes, clusters, green-jobs themes). The occupational maps are the object that connects them, which is exactly why they are hard to use well. The data is relational but the delivery is document-shaped; there is no published schema for the connected object, so every consumer re-invents one; and the SOC crosswalk, the single most useful bridge to official labour-market statistics, is buried inside each record rather than exposed as a first-class mapping.

How we built it

We harvested a complete static snapshot of the Public API under the Open Government Licence and modelled it as an ontology. Three design choices make it trustworthy and reusable. First, the map is modelled as it is defined: the Route to Pathway to Cluster hierarchy is a SKOS concept scheme, so it can be browsed and reasoned over like any controlled vocabulary. Second, the SOC crosswalk is a first-class citizen: every distinct SOC 2010 and 2020 code becomes a concept in its own scheme, and each occupation carries an explicit mapping edge, so the bridge from standards to ONS statistics is a single query rather than a bespoke join. Third, the graph is constraint-checked, not just serialised: a SHACL shapes file states what a well-formed occupation, product and SOC concept must look like, and the graph is validated against it, and against referential-integrity checks, on every build.

Assurance

The graph conforms to the SHACL shapes with zero violations across all 51,355 triples, and passes four referential-integrity checks: no occupation without a route, no progression edge to a non-occupation, no green-theme link to a non-theme, and no SOC reference without a notation. Every one of the 1,269 standards is fully placed in the map (route, pathway, cluster and technical level all present); 1,079 (85%) carry a SOC 2020 mapping. The whole pipeline is reproducible: one script rebuilds the graph from the snapshot, another re-runs validation and the coverage report.

Why an open ontology matters for the skills system

The direction of travel in national skills policy is a single, authoritative, data-driven view of what the country needs, one that can map the routes from courses and standards through to occupations, and reason about skills demand sector by sector. Priority occupations for the industrial strategy are already identified by SOC code; a new standardised skills classification links skills to occupations through the most detailed level of SOC 2020; and the occupational maps themselves are published as free data behind a public API. The missing layer is not more data. It is the connective tissue that makes the data computable: a typed, validated graph in which routes, occupations, SOC codes, products and progression are first-class, queryable objects rather than fields inside separate documents.

That is what this ontology provides, and it is why we built it as an open artifact rather than an internal tool. Because the SOC crosswalk is a first-class citizen, the maps snap directly onto the classification that underpins every official statistic on jobs, pay and vacancies, and onto the emerging skills classification layered on top of it. Questions that used to need bespoke code become single queries: the full progression neighbourhood of any occupation, every standard that feeds a given SOC group, the apprenticeship and higher technical offer under any green theme, or the occupations that sit across more than one route. It lets employers, providers and analysts compute pathways, gaps and green-transition demand at scale instead of reading pages one at a time.

Open source, and free to reuse

Everything here is open. The SEOM vocabulary, the full instance graph in Turtle and JSON-LD, the ONS SOC 2010 and 2020 crosswalk as standalone SKOS concept schemes, the SHACL shapes and the reproducible build and validation scripts are all published on GitHub under an open licence, alongside the static snapshot they are built from. Anyone can query it, extend it, align it to their own vocabularies, or regenerate it from a fresh snapshot. We built it to be reused, and to show what the occupational maps become when they are treated as the connected, standards-based, machine-readable evidence base the skills system is asking for.

Ontology, data and reproducible build on GitHub