Lesson 9 — Metadata, Lineage & Trust
Metadata is part of the architecture
A large DW/BI estate can contain multiple integration tools, databases, modeling tools, semantic layers, BI platforms, and versions. An integrated Metadata approach connects those perspectives so users and maintainers can understand data end to end.
Dictionary / glossary / Metadata
These answer questions such as: - What does this measure mean? - What is its business definition? - Which datatype/structure stores it? - Who owns/stewards it? - Which rules or restrictions apply? - Which source/target mappings are associated with it?
A dictionary definition can explain “Net Revenue,” but it does not by itself prove which ERP fields and transformations produced today’s dashboard value.
Lineage
Lineage answers:
Where did this come from? → what happened to it? → where did it go?
It connects sources, mappings, transformations, targets, semantic layers, reports, and other consumers as appropriate.
Source-to-target mappings are important lineage evidence, which is why they should be created during development rather than reconstructed after production.
Impact analysis
Impact analysis asks a different direction of question:
If this field/rule/process changes, what will be affected?
Lineage and dependency Metadata enable the answer, but lineage is not the same question as impact analysis.
Memory hook: - Metadata tells; - lineage traces; - impact predicts.
Keep models synchronized
If the logical model says Customer Status has one governed meaning while the physical warehouse and BI semantic layer implement another, trust breaks. Synchronizing logical and physical models helps Metadata, lineage, maintenance, and impact analysis remain useful.
Queryable audit data
Users/support teams should be able to inspect sufficiently granular evidence such as when data arrived, which load processed it, what was rejected, whether transformations completed, and whether freshness/quality expectations were met.
This does not replace business lineage; it supplies operational evidence that complements it.
Trust chain
The current visual artifact compresses trust into:
Source-to-Target Map → Business Rules → DQ/Remediation → Metadata → Lineage → Business UAT
Each link answers a different concern: - What maps to what? - Which rule changed it? - Were defects handled? - What does it mean? - Where did it originate/go? - Did the business verify the result?
If users cannot explain or reproduce a KPI, identify which link is missing rather than merely creating another dashboard.
Stop and check
Auditor asks which ERP fields and transformations produced dashboard revenue: lineage plus supporting mappings/Metadata.
User asks what “Net Revenue” means: dictionary/glossary/business Metadata.
Architect asks what breaks if Product hierarchy changes: impact analysis.
Source anchor: pp. 363–364, 376, 381–391.