Data Governance
Data governance is the set of policies, roles, and controls that define who may use which data, for what purpose, and to what standard.
Governance as accountability
Data governance answers organizational questions that technology alone cannot: who owns a dataset, who may access it, how long it is kept, what it may be used for, and how its quality is assured. It is less a product than a framework of policies backed by enforceable controls. Without it, data sprawls into ungoverned copies of uncertain provenance and permission.
Ownership and stewardship
Every important dataset needs an accountable owner and often a steward who curates its meaning and quality day to day. The owner sets policy; the steward maintains definitions, resolves quality issues, and answers questions from consumers. Clear ownership turns "nobody knows what this column means" into a person who does.
Classification and policy
- Classify data by sensitivity: public, internal, confidential, restricted
- Attach handling rules to each class: encryption, access, retention
- Tag purpose limits so data collected for one use is not repurposed silently
- Record legal and regulatory obligations that apply to each dataset
Enforcement, not just intention
A policy that only lives in a document is not governance. Effective governance binds policy to mechanism: access classifications drive access-control rules, retention policies drive automated deletion, and purpose limits drive audits of query logs. The catalog that describes data and the access system that guards it are where governance becomes real.
Auditability
Governance requires the ability to answer, after the fact, who accessed what and when. Immutable audit logs of access and change make this possible and are frequently a compliance requirement. In a research organization handling sensitive design and simulation data, a firm boundary between what is public and what is confidential is a governance control: classification determines which repository a file may enter, and the enforcement is a publication check, not a hope. See access control and data catalogs.