15.1 Governance Foundations

1. Governance is a decision system

Data governance is not a committee that approves every table.

A useful enterprise definition is:

Governance is the system of policies, ownership, decision rights, standards, and controls that makes data trustworthy and usable at scale.

A governance model answers five questions:

1. Who decides?
2. What standard applies?
3. How is compliance measured?
4. What happens when the standard is violated?
5. Who owns remediation?

2. Governance dimensions

Dimension	Question	Typical artifact
Ownership	Who is accountable?	Data owner / steward
Definition	What does it mean?	Glossary term
Classification	How should it be handled?	Tag / classification
Quality	Is it fit for purpose?	Quality rule / SLO
Lineage	Where did it come from?	Lineage graph
Access	Who may use it?	Role / policy
Lifecycle	How long should it exist?	Retention / retirement rule
Change	What happens when it changes?	Change policy
Consumption	How should consumers use it?	Data product contract

3. Centralized versus federated governance

Centralized

One enterprise team owns most standards and decisions.

Advantages

• consistent policies,
• simpler escalation,
• easier initial implementation.

Weaknesses

• bottleneck risk,
• weak domain context,
• slow decision making.

Federated

A central governance function defines enterprise standards while domains own their data.

Advantages

• domain expertise,
• scalable stewardship,
• ownership closer to the source.

Weaknesses

• inconsistent execution if standards are weak,
• duplicated processes,
• harder cross-domain reconciliation.

Recommended enterprise pattern

Use federated execution with centralized standards.

The central function defines:

• minimum metadata,
• classification taxonomy,
• quality expectations,
• naming conventions,
• critical data rules,
• lifecycle principles.

Domains own:

• definitions,
• business context,
• quality remediation,
• product roadmaps,
• consumer relationships.

4. Data owner versus data steward

These roles are frequently confused.

Data owner

Accountable for business meaning, risk, and major decisions.

Data steward

Responsible for operational governance activities such as:

• maintaining metadata,
• coordinating quality remediation,
• validating definitions,
• managing classifications,
• supporting consumers.

The steward performs governance work; the owner remains accountable.

5. Governance policy hierarchy

A scalable model looks like:

Enterprise Policy
      |
      +-- Data Classification Standard
      |
      +-- Metadata Standard
      |
      +-- Data Quality Standard
      |
      +-- Access Standard
      |
      +-- Lifecycle Standard
      |
      +-- Domain Policies
              |
              +-- Finance
              +-- Procurement
              +-- Sales
              +-- Supply Chain

6. Minimum governed asset contract

Before an important asset becomes production-facing, require at least:

• business name,
• technical name,
• owner,
• steward,
• business description,
• source,
• domain,
• classification,
• refresh expectation,
• quality status,
• lineage,
• intended consumers,
• access requirements,
• lifecycle status.

Do not make every metadata field mandatory for every object. Instead define criticality tiers.

7. Criticality tiers

Tier 1 — Enterprise critical

Examples:

• financial reporting,
• regulatory reporting,
• executive KPIs,
• core planning datasets.

Require strong ownership, quality monitoring, lineage, and change management.

Tier 2 — Domain critical

Important to one business domain but not enterprise-wide.

Tier 3 — Team or exploratory

Useful for analysis but not a formal enterprise contract.

This prevents governance from becoming a bureaucracy.

8. Governance anti-patterns

Anti-pattern 1: govern everything equally

Result: governance effort grows faster than business value.

Anti-pattern 2: catalog without ownership

A beautiful catalog with no accountable owner becomes a searchable graveyard.

Anti-pattern 3: glossary without adoption

Definitions are useful only when models, reports, and decisions actually use them.

Anti-pattern 4: quality without remediation

A score is not governance. Someone must own the action.

Anti-pattern 5: security as governance

Security controls access. Governance additionally establishes meaning, ownership, quality, lifecycle, and accountability.

Architect exercise

Design a governance model for:

• 20 domains,
• 5000 assets,
• 300 critical KPIs,
• 2000 consumers,
• separate DEV, TEST, and PROD tenants.

Define:

1. central responsibilities,
2. domain responsibilities,
3. minimum metadata,
4. criticality tiers,
5. escalation model.

Your answer should explain how governance scales without requiring a central team to manually approve every object.