Data Governance and Stewardship
This domain tests the operating model that decides how data is defined, protected, and improved. CDP questions focus on accountability, policy artifacts, stewardship, metadata, quality remediation, and regulatory controls.
Core rules and concepts
- Governance versus management: governance sets direction, priorities, and accountability; management executes projects and operations. A data council or governance board resolves cross-domain issues.
- Roles: a data owner is accountable for a data domain and approves definitions and access. A data steward performs tactical stewardship: quality monitoring, issue triage, and definition maintenance. IT or a custodian operates platforms. Use a RACI to make decision rights explicit.
- Policy artifacts: a policy states intent and mandatory direction; a standard is a mandatory rule such as naming or classification; a procedure gives step-by-step execution; a guideline is recommended practice.
- Metadata and glossary: a business glossary defines terms; a data catalog inventories assets; lineage traces data from source to report. These artifacts make stewardship and impact analysis possible.
- Quality and issue management: define dimensions, thresholds, owners, and escalation paths. Remediation should address root cause, not only clean a one-time extract.
- Protection controls: classification, access control, retention, privacy obligations, and audit evidence must align with policy and regulation.
Worked example
Customer email completeness falls below threshold. The steward triages the issue, traces affected reports, and proposes a validation rule. The data owner approves remediation priority and definition changes; IT fixes the source capture process; the council updates the standard if the rule applies enterprisewide. The steward tracks completeness after the fix, so the outcome is measured rather than assumed.
Traps that cost marks
Saying the steward owns the data; confusing data ownership with system ownership; treating governance as a software purchase; mixing policy, standard, and procedure; cleaning data without fixing the source; and creating a glossary with no accountable owner or approval process.