Data Governance
Also: Data Governance Framework, Enterprise Data Governance, DG
Data governance is the set of policies, roles, and processes that ensure data is accurate, secure, well-defined, and used responsibly across an organization.
What It Is
Data governance is the discipline of managing the availability, usability, integrity, and security of data across an organization. It defines who can take what action, on which data, in what situations, and using which methods. Rather than a single tool, it is a framework that combines policies, standards, roles, and processes supported by technology.
Governance answers practical questions: Who owns the customer table? What does "active customer" actually mean? Who is allowed to see salary fields? How long do we keep transaction records? When data changes, who approves it?
Why it matters
Without governance, organizations accumulate conflicting definitions, duplicated datasets, and unclear ownership. This leads to poor decisions, compliance risk, and wasted effort reconciling numbers.
- Trust: Decision makers can rely on reports when definitions and lineage are clear.
- Compliance: Regulations like GDPR, CCPA, and HIPAA require demonstrable control over personal data.
- Efficiency: Teams stop rebuilding the same metrics and arguing over whose number is correct.
- Risk reduction: Access controls and audit trails limit breaches and misuse.
For a Chief Data Officer, governance is the operating model that turns scattered data into a managed, accountable asset.
How it is used in practice
Governance programs typically establish:
- Roles: Data owners (accountable for a domain), data stewards (day to day quality), and a governance council that sets policy.
- Standards: Naming conventions, business glossaries, and agreed metric definitions.
- Catalogs and lineage: Tools that document where data comes from and how it transforms.
- Quality rules: Automated checks for completeness, validity, and uniqueness.
- Access policies: Classification of sensitive data and rules for who can view or edit it.
Concrete Example
A bank notices that "monthly active users" differs between the marketing dashboard and the finance report. A governance effort assigns the marketing analytics lead as data owner for the metric, publishes a single definition in the business glossary, documents the source tables and transformations as lineage, and adds quality checks to flag missing login events. After rollout, both teams report the same figure, audits become traceable, and disputes drop sharply.
Governance is ongoing, not a one time project: definitions, regulations, and systems keep evolving.
Frequently asked questions
What does data governance actually cover?
Data governance defines who can take what action, on which data, in which situations, and by which method. It combines policies, standards, roles and processes, supported by tools like catalogs and quality checks. It is not a product you buy but an operating model you maintain, since definitions, regulations and systems keep changing.
Who needs to care about data governance in a company?
Primarily the Chief Data Officer or whoever owns the data function, because governance is the operating model that turns scattered data into a managed, accountable asset. Business domain leaders are involved too: they act as data owners, accountable for definitions and access in their area. Finance, legal and compliance teams have a direct stake through GDPR, CCPA or HIPAA obligations.
What is the difference between a data owner and a data steward?
The data owner is accountable for a data domain: definitions, access rules, retention decisions. The data steward handles day to day quality within that domain, applying the rules and fixing issues. Above both sits a governance council that sets policy across domains.
How do you fix a metric that shows two different numbers in two reports?
Assign one data owner for the metric, publish a single definition in the business glossary, then document the source tables and transformations as lineage. Add quality checks on the underlying events so gaps get flagged instead of silently distorting the figure. A bank facing diverging "monthly active users" between its marketing dashboard and its finance report resolved it exactly this way, and both teams then reported the same number.
Where should you start a governance program if nothing exists yet?
Start with the metrics and datasets that already cause visible disputes or compliance exposure, not with an enterprise-wide catalog. Name an owner, agree one definition, document the lineage, then add automated checks for completeness, validity and uniqueness. Classification of sensitive fields and access rules follow, since they are what auditors ask for first.