Glossary Term

Origin & evolution

The business glossary grew out of data governance practice, where it sits beside the data dictionary. The two answer different questions. A data dictionary documents technical metadata: what a field is and how to work with it. A business glossary defines terms in business language so teams share one understanding of a concept, independent of which table it lives in.

The semantic layer is the older relative. Business Objects patented the idea in the early 1990s, and the first implementations shipped inside BI products such as MicroStrategy, Business Objects, and Cognos. These early semantic layers let business users build reports in their own vocabulary without writing SQL, by mapping technical structures to business concepts like Revenue and Customer.

Recent practice split the semantic layer into a narrower metrics🔢MetricStrategyA unified metric that measures progress, health, or behaviour across the productView reference → layer. As dbt describes it, the metrics layer centralises metric calculation logic so that "revenue" computes the same way everywhere it is queried. The glossary term and the metric definition converge here: the glossary holds the human meaning, and the metrics layer holds the executable version of that meaning. Modern data catalogues knit the two together, so a term links to its source tables, lineage, and the reports that use it.

How it works in practice

A growth team and a finance team both report "active users". Growth counts anyone who opened the app in the last 28 days. Finance counts anyone with a paid, non-cancelled subscription🔂SubscriptionSales & RevenueA recurring subscriptionView reference →. Both are reasonable; they are not the same number, and the gap is roughly 40%. Every cross-team meeting wastes ten minutes reconciling which figure is real.

The fix is a glossary term. The organisation defines active_user as "a user with at least one qualifying session in the trailing 28 days", documents the two earlier readings as paid_active_user and engaged_user, and links each to the metric and the source tables that compute it. The 40% gap does not disappear, but it stops being an argument. Each number now has a name, and the meeting moves on.

Glossary term vs. its neighbours

• Metric is a measurement; a glossary term is a definition. "Active users this week: 18,420" is a metric. "What counts as an active user" is the glossary term that makes the metric legible. A metric without its term is a number whose meaning is folklore.
• Data domain🏷️Data DomainData & AnalyticsA logical grouping of related data assetsView reference → is the area of the business a term belongs to (marketing, finance, product). The domain owns and defines its terms, which is why the same word can carry different agreed meanings across domains, each documented rather than assumed.
• Data product🎁Data ProductData & AnalyticsA curated, reusable data assetView reference → is a packaged, consumable dataset. It uses glossary terms to describe its fields, so a consumer reading the product knows that "revenue" there matches the organisation's agreed definition.

In the graph

In the Unified Product Graph, a glossary term lives in the data and analytics region as the shared-vocabulary node. A product_defined_by_glossary_termProductdefined byGlossary Termhierarchy edge attaches the definitions a product depends on, and data_domain_defines_glossary_termData DomaindefinesGlossary Termhierarchy records which domain owns the meaning. Placing the term as its own node, rather than a comment on a metric, lets many metrics and data products reference one canonical definition; redefine the term in one place and every consumer inherits the change, which is the whole point of a shared vocabulary.