Metadata Management
Three types, one discipline
Metadata is data about data, and exam questions almost always hinge on classifying it:
- Business metadata - business names, definitions, business rules, glossaries, ownership, stewards, KPIs.
- Technical metadata - schemas, tables, columns, data types, keys, models, source-to-target mappings, ETL logic, file layouts.
- Operational (process) metadata - job logs, run statistics, error logs, batch timings, access and audit logs, refresh dates, record counts.
What metadata management delivers
Strategy and architecture (a metadata repository, often with integration layers), ingestion and integration from tools, control (quality of the metadata itself), and delivery: business glossary, data dictionary/catalog, lineage, impact analysis, search, and self-service BI.
Lineage traces data backward (where did it come from) and forward (where does it go). Impact analysis is forward lineage applied to change: "what breaks if I rename this column?"
Maturity typically progresses from metadata as documentation (manuals, spreadsheets) → metadata as a service (searchable, integrated catalog) → metadata as an enabler (automated policy enforcement, active lineage in pipelines).
Traps
- Confusing a business glossary (business terms and definitions) with a data dictionary (technical structures: columns, types, lengths).
- Treating lineage as purely technical; business lineage mapping definitions and reports also matters.
- Assuming buying a repository equals managing metadata - without stewardship it becomes stale.
- Overlooking operational metadata in audit and troubleshooting questions.
Rule of thumb
Ask what the question is about: business meaning → business metadata; storage, structure, or transformation → technical metadata; what happened when something ran or who touched it → operational metadata.