Theory
The n-squared problem of meaning
Every organisation models the same nouns — customer, product, account, patient, well — slightly differently. When two systems must exchange data, someone writes a mapping. With N systems each inventing their own model, you trend toward N×N point-to-point mappings, every one a place for meaning to leak.
An industry-standard ontology breaks that curve. Instead of N×N private models, everyone aligns to one shared, governed model in the middle — so integration becomes N mappings to the hub, not N² between peers. A real standard gives you three things a home-grown model can't:
- Shared semantics — pre-agreed classes and properties whose meaning a whole industry already accepts, so 'LegalEntity' or 'Observation' means the same thing across vendors and regulators.
- Stable identifiers — durable IRIs / codes (a FIBO class IRI, a SNOMED SCTID, a GS1 GTIN) that survive across systems and decades, so you can join data you didn't produce.
- Governance — a maintaining body (a consortium, an ISO committee, a foundation) that versions the model, arbitrates disputes and keeps it alive after its first authors move on.
Use Case Example: Two banks reporting to the same regulator both describe a counterparty's legal form. If each invents its own codes, the regulator must reconcile them by hand. If both reference the FIBO class for that legal form, the data lines up automatically — no bilateral mapping, no interpretation drift.