Lesson 7 — Integration, Delivery, and Product Thinking
Metadata integration is data integration
Source Metadata rarely uses one vocabulary or structure. A practical flow is:
HARVEST / EXTRACT → STAGE / AUDIT → STANDARDIZE → MAP EQUIVALENTS → RESOLVE OVERLAP / CONFLICT → MERGE / LOAD → AUDIT / HISTORY
Imagine one tool calls an entity Customer, another Client, and another Party. Similar labels do not prove semantic equivalence. The integration process must compare definitions, scope, relationships, and source authority.
If two sources disagree on ownership, “last load wins” is usually the wrong governance rule. A Steward or Governance decision may be required.
Why staging helps
A staging layer can preserve: - source evidence before changes; - auditability; - rollback/recovery; - investigation of bad imports; - controlled standardization before merge.
Metadata quality becomes visible during integration
Common problems include stale repositories, missing definitions, duplicate concepts, conflicting owners, broken mappings, and inconsistent names. Treat those as Metadata quality problems, not merely quirks of source tools.
Manage Metadata as a product
Good Metadata is not an accidental by-product of harvesting. It has: - consumers; - accountable source processes; - standards; - quality expectations; - change control; - service expectations; - feedback loops; - lifecycle management.
Integrate vs Distribute/Deliver
Integration answers: How do we reconcile and assemble source Metadata?
Delivery answers: How do people and systems get useful access to the managed Metadata?
Delivery mechanisms may include: - portals/intranets; - reports, glossaries, and documents; - DW/BI environments; - CRM/ERP/application layers; - flat/XML/JSON files; - web services/APIs.
A technically complete repository can still fail if users must email a specialist to discover a definition or lineage path.
Usage is feedback
Search logs and repeated questions can reveal missing content, poor terminology, confusing navigation, low trust, or weak socialization. Access and adoption are part of Metadata Management, not afterthoughts.
Source anchor: pp. 414–417.