Construction: IFC, Uniclass, COBie and BOT

Construction has more data standards per pound of built asset than almost any other sector, and still loses most of its information at handover. The models are rich, the classification systems are thorough, the exchange formats are internationally standardised, and yet the facilities manager who inherits the building typically receives a folder of PDFs and a spreadsheet with half the asset tags missing. The reason is that the four things people call “BIM standards” do completely different jobs, and mixing them up guarantees loss. This lesson separates them: what each is, who runs it, what it costs to get, and where it is genuinely deployed.

IFC: the exchange schema, now including infrastructure

Industry Foundation Classes is the open, vendor-neutral schema for exchanging built-asset models. Its current edition, IFC 4.3, was approved as a final international standard in January 2024 and published as ISO 16739-1:2024. The headline change is scope: for the first time the schema properly covers infrastructure, adding rail, road, bridge, port and waterway domains to a model that had been building-centric since the 1990s. IFC defines entities such as walls, ducts, alignments and spaces, their property sets, their relationships, and the geometry attached to them. It is maintained by buildingSMART International, and the full schema documentation is openly published at the IFC 4.3 documentation site, so unlike most ISO standards you can read the normative content without paying.

Two practical caveats. First, tool support lags the standard by one to three years, so a project can require IFC 4.3 deliverables that the design team’s software exports imperfectly. Second, IFC is a schema in EXPRESS with STEP file serialisation, plus XML and an OWL rendering, and the OWL version is a translation of a schema rather than a purpose-built ontology. It inherits the schema’s shape, which is why lightweight ontologies exist alongside it rather than being replaced by it.

IFC5 is the announced re-architecture: away from monolithic EXPRESS and STEP files towards granular, componentised, web-native data, with early implementations demonstrated at buildingSMART events from 2025 onwards. Treat it as a direction of travel rather than a delivery format for a project starting today, and plan for a long period in which 4.3 and 5 coexist.

Uniclass: classification, not exchange

Uniclass is a unified classification system for the UK construction industry, maintained by NBS. It is compliant with ISO 12006-2 and organised into tables covering complexes, entities, spaces and locations, activities, elements and functions, systems, and products, among others. Its job is to give every object in a project a code from a controlled hierarchy so that costing, specification, scheduling and asset registers all speak of the same kind of thing. Uniclass tables are published by NBS and are freely downloadable, and they are also exposed through the buildingSMART Data Dictionary, which as of May 2025 hosted over three hundred dictionaries and offers an API so that classification lookups can be automated instead of pasted.

Classification is not a data model. A Uniclass code tells you what category a thing belongs to; it says nothing about the thing’s parts, its connections, its states or its history. Teams that try to use classification as their semantic layer end up encoding meaning in code strings and comments, which is exactly the failure mode formal modelling exists to prevent.

COBie: the handover payload

COBie, Construction Operations Building information exchange, is the format for delivering the asset information a facilities team actually needs: spaces, types, components, systems, spares, resources, jobs and documents, structured as linked worksheets. Technically it is a model view definition of IFC, which means COBie is a filtered, non-geometric projection of the same underlying schema. UK usage was defined by BS 1192-4:2014, which was withdrawn in 2022 and superseded by BS EN ISO 19650-4:2022 on information exchange, part of the wider migration from UK-specific BIM standards to the ISO 19650 series.

The deployment record is genuine and instructive. COBie deliverables were required on centrally procured UK government projects under the Level 2 BIM programme, which is why every UK contractor knows the acronym. It is also where the sector learned that a mandated spreadsheet does not equal usable data: COBie files arrive complete against the schema and empty of meaning, with placeholder manufacturers and duplicated types, because the schema constrains structure and nobody constrained content. That distinction between structural validity and content validity is the same one that separates OWL from SHACL, and construction learned it the expensive way.

BOT: the lightweight ontology layer

The Building Topology Ontology is a minimal OWL DL ontology describing the core topology of a building: sites, buildings, storeys, spaces, zones, the elements they contain and the interfaces between them. It was produced by the W3C Linked Building Data Community Group and published in the Semantic Web journal in 2020 by Rasmussen, Lefrançois, Schneider and Pauwels. The design brief was explicit: capture the topological backbone without inheriting IFC’s size and complexity, then let domain ontologies for products, properties, sensors and flow systems plug into it. The specification is public at the W3C LBD Community Group site.

Status matters here. BOT is a Community Group report, not a W3C Recommendation, so it carries community consensus rather than standards-track authority. In exchange you get something you can actually reason over: a handful of classes, clean transitive containment relations, and a graph that answers questions such as which elements adjoin which spaces on which storey in a single SPARQL query. In practice BOT is the bridge that lets a building model join a wider knowledge graph containing sensors, occupancy, energy and maintenance data, none of which IFC was designed to carry.

Machine-checkable requirements and the crosswalk

The missing piece for years was a way to state what a model must contain and have a computer check it. buildingSMART’s Information Delivery Specification fills that gap: requirements are written as machine-readable specifications against IFC entities, classifications and property sets, and a checker reports conformance instead of a human opening the model and looking. Paired with the buildingSMART Data Dictionary for controlled vocabulary lookups, IDS gives construction the equivalent of what SHACL gives an RDF pipeline: a gate that fails the delivery, not a comment on it.

Doing the crosswalk properly

A working construction knowledge graph needs all four layers wired together: an IFC entity carries the geometry and property sets, a Uniclass code places it in the classification hierarchy, a COBie row expresses the handover subset, and a BOT class puts it in a topological graph you can query. The joins are made through identifiers and through bSDD lookups, and every join is a mapping claim that should be recorded with its predicate, its source and a test. Expect the same losses as every other domain: temporal qualification, so the fact that a component was installed, replaced and reinstated collapses into a current state; roles, so a space serving as a plant room during construction becomes a plant room permanently; and units, so a value without an explicit measurement pattern becomes a number that someone will misread.

The UK is now pushing this further. The National Digital Twin Programme is extending the government’s Information Exchange Standard with IES:Building, a 4D extension for the built environment, which puts construction data on the same 4D ontological footing as defence and national infrastructure. If you work in this sector, the combination of IFC for exchange, Uniclass for classification, IDS for validation and a 4D ontology for the long-lived asset record is where the market is heading. Our construction data standards and AI course works through these crosswalks with real files rather than diagrams.

Apply it to your work