In depth
Think of metadata as the label and instructions that travel with your data. It removes guesswork by answering who created the data, what it measures, where it came from, when it was last updated, and how to use it.
Common types you'll see in analytics:
- Descriptive metadata: names, titles, summaries, tags, keywords.
- Structural metadata: how parts relate, like table joins, hierarchies, and dimensions.
- Administrative and technical metadata: owners, permissions, retention rules, data types, formats, and refresh schedules.
- Lineage and quality metadata: data source, transformation steps, version history, freshness, and quality checks.
- Semantic metadata: business definitions and rules, like the approved definition of "Active Customer."
Strong metadata helps both people and software apply business meaning without repeated clarifications or custom code.
Pro tip
Treat metadata like product requirements. Set mandatory fields for every metric and dataset: clear name, one-sentence description, owner, unit or currency, aggregation rule, time grain, and tags. Make these part of your workflow so nothing ships without context.
Why metadata matters
- Faster self-serve: people can search, filter, and trust results without pinging a data teammate.
- Consistent answers: shared definitions reduce conflicting numbers in meetings.
- Governance and security: owners, access levels, and retention notes prevent misuse.
- Auditability: lineage shows where numbers come from and which steps transformed them.
- AI-readiness: clear structure and definitions let assistants and NLP features answer questions accurately.
Metadata in practice
- Supply chain: an "On-Time Delivery" metric with definition, source tables, filters, owner, and last updated timestamp helps sales and shipping reach the same count.
- Finance: revenue metrics with currency, accrual vs cash basis, and period alignment stop reconciliation churn.
- Product: usage metrics with event names, user scopes, and exclusion rules prevent double counting.
Metadata in a metric catalog
A Metric Catalog is where metric metadata lives in one place, so people can discover, trust, and share the same numbers. Each entry typically records:
- Name, description, owner, and certification status
- Tags and categories for search
- Data source, aggregation method, and calculation logic
- Dimensions and hierarchies, like region, plan, or channel
- Unit, currency, time grain, and time zone
- Refresh schedule and last updated timestamp
- Access controls and lineage
This metadata is also what a Knowledge Graph draws on to connect metrics, dimensions, and definitions. Before adding more metrics, check that each existing one has an owner and a definition your team agrees on.
Related terms
What is a semantic layer?
A semantic layer is the shared business vocabulary and rules that translate raw tables into consistent, human-readable metrics and dimensions. It turns questions like "What do we mean by revenue?" into reusable definitions every chart and query uses.
What is data lineage?
Data lineage maps the journey of your data from origin to destination. It shows where data comes from, how it is transformed, and where it is used, so your team can trace any number in a dashboard back to its source and trust the answer before acting on it.
What is data governance?
Data governance is a formal framework of people, policies, and technology that keeps an organization's data accurate, secure, and usable. Think of it as the "Librarian" of a massive digital library: every piece of data is cataloged, protected, and accessible only to those with the right permissions. For a business leader, it sets the rules for who owns each number, so decisions rest on information you can trust and stay compliant with privacy regulations.
What is data quality?
Data quality measures the reliability of your data for making decisions. High-quality data is accurate, complete, timely, consistent across systems, valid, and free of duplication. When leaders trust the numbers, meetings move from debating whose number is right to deciding what to do next.
What is a context layer?
A context layer is a live, governed system that sits between your raw data and AI. It adds ownership, lineage, business rules, and metadata to every metric, so leaders and AI tools work from the same trusted numbers. It differs from a semantic layer, which standardizes what a metric means. A context layer also records who owns it, when it was updated, and how it connects to other data. The result is fewer conflicting numbers and answers you can act on without double-checking.