Architecture Portal

Data governance

Decisions, responsibility and care through the data lifecycle.

Adrian Sutherland · Version 1.0 · · © 2005–2026

A correction travels

In the fictional cancellation service, a learner’s agreement was recorded against the wrong alternative. Staff correct the cancellation case, but yesterday’s report still counts it as resolved. Someone needs to decide who corrects that report, who is told and whether the mistake exposes a wider problem with the definition or process.

Data governance establishes who may make these decisions, the rules they apply and how unresolved issues are handled. It connects the meaning of information to its continuing care. In A Simple Architectural Framework (ASAF), it crosses Information, Management, Structure, Security and Metrics, through all four process areas.

Responsibility

Agree responsibility for each useful group of information, such as cancellation cases or bookings. The following division is a starting point for tailoring. It draws on the UK government’s data ownership model, published in April 2026, which distinguishes accountability, stewardship and technical custody. Here, ownership means decision responsibility.

Responsibility In the proposed service
Data owner The service owner approves case meaning, permitted uses, quality expectations and significant changes.
Data steward A nominated member of the service team maintains definitions, investigates recurring errors and coordinates corrections.
Technical custodian The register administrator, or later platform operator, implements access, recovery and disposal decisions.
Record owner The current case owner keeps an individual cancellation moving and its record current.
Consumer The reporting lead states what a report needs and applies correction or withdrawal decisions to its outputs.

One person may perform several roles. Record the authority attached to each decision, including who resolves disagreements between teams. A managed platform can enforce an access rule; the service still needs someone to decide what that rule should permit.

Decisions to retain

Concern A useful record
Meaning and authority Definition, scope, authoritative source, owner and known disagreements.
Quality Intended use, required checks, thresholds, exceptions and who may accept a limitation.
Access and use Permitted people, services and purposes; approval, review and withdrawal arrangements.
Lineage Source version, transformation and output, so a correction can find affected uses.
Lifecycle Collection purpose, retention trigger and period, disposal, holds and treatment of copies.
Change Proposed difference, affected consumers, acceptance, migration and rollback or correction.

Lineage describes where information came from and how it was derived. Include exports, reports, caches and search indexes where they affect a decision. If an assistant uses retrieved case material, specify which version and access rules apply, and how corrected or withdrawn material stops being offered. An offline copy needs a freshness rule and a way to report a failed or incomplete synchronisation.

Preserving a correction history is useful, but does not justify indefinite retention of personal information. Set an applicable retention policy before using real records. Define how authorised deletion reaches derived copies and how expired data is handled if a backup is restored.

Compare contributions

These sources answer different questions. The comparison uses their public guidance, with the scope of reading shown below.

Source Useful contribution Application here
DAMA International’s Data Management Body of Knowledge (DMBOK) overview A broad reference for data-management responsibilities and practices, adapted to context. Check coverage across the architecture. This comparison uses the overview, not a review of the full guide.
Government data ownership model Accountable owners, operational stewards and technical custodians. Map decision rights to actual roles; the government role structure is one organisational setting.
Government Data Quality Framework Assess quality against use, prioritise problems and improve through the lifecycle. Agree checks and a response to failure. Complete fields alone do not establish correct meaning.
W3C PROV overview Represent origins, derivation and responsible parties. Retain enough lineage to follow a correction. Our simple record does not implement the PROV specification.

A central team can maintain shared definitions and controls, but may become a queue for local decisions. Teams can own their information nearer its use, with shared rules for exchanges and a route for settling disputes. These are design choices to examine through decision time, unresolved disagreements and correction effort, rather than a ranking of the sources above.

Work through the cycles

Process area Continuing work
Visioning Agree which information and uses matter, and what quality is needed.
Transforming Change definitions, access and representations with the affected producers and consumers.
Operating Monitor quality, resolve issues, review access and carry out retention decisions.
Governing Set authority and policy, review exceptions and settle conflicts across uses.

A correction within an agreed definition can stay with the case owner. A repeated ambiguity may return to the steward and data owner. A new purpose or external recipient reopens the relevant use and sharing decisions. Retain the sound definition and controls while changing what the finding affects.

For a medallion design, specify the same responsibilities across bronze, silver and gold data. A gold report needs its own accepted meaning, quality and correction route; its layer name does not establish fitness for a decision. The processing guide develops that example.

A worked decision

The case example follows the incorrect agreement into a derived report. Use the blank and completed governance records to record responsibility, affected copies, decisions and proposed checks.

Methodology configuration

Your method may use information asset owner, domain owner or data product owner. Check the scope and authority behind each title. An existing service review, data contract or change record may already hold the decisions above. Link those records through Methodology configuration.

Dials

Choose how often quality and access are reviewed, which changes need wider agreement and what may be checked automatically. Frequent automated checks can expose errors sooner; someone still needs authority and time to act on them. Match the review effort to the consequence of using wrong or stale data.

About this edition

Refreshed for the September 2026 website update. Develops earlier Portal and ASAF material; the fictional governance example was added in 2026.

Scope, limitations and next checks

Status and accountability

Read this page with its boundaries visible

Status

In development

Last reviewed

Intended users

  • Architects connecting information decisions to delivery and operation

Non-goals

  • A prescribed organisation chart, platform or compliance assessment

Limitations

  • The fictional arrangements and checks need application in a real context.

Next evidence sought

  • Try the governance record with a data owner and the people maintaining and using the data.