Learn AI /Data and Infrastructure /Data governance
Data governance
- Starter
- Data and Infrastructure
- Policy officers
- Managers & front offices
- Programme & data teams
Also called: data stewardship, data responsibility.
Back to the milestone you came from
In five seconds
What it is
Data governance is the set of rules, roles and checks that control how data is collected, stored, used, shared and corrected over time.
In plain words
Data governance is the organised way an institution makes sure data is handled properly: clear rules, clear responsibilities, careful access and regular checks.
An everyday example
Several relatives share a list of emergency contacts. They agree which details belong in it, who may edit them and how an outdated number should be corrected. One person checks duplicate entries, records when details were last confirmed and removes information that is no longer needed. Everyone can use the list, but not everyone can overwrite it. The value comes less from the spreadsheet than from the rules that keep it accurate, current and appropriately shared.
An exam file is simpler than most real data environments. In organisations, data often comes from many systems at once, changes continuously, and may cross legal or geographic boundaries. Governance also covers retention, sharing, deletion and documentation over time, not just one review meeting.
Remember thisData governance makes clear who may collect, change, share and delete data, under which rules, with enough traceability to resolve the inevitable query about which version was cleared.
Understand it
Understand it
Why it matters
Non-specialists make decisions from data all the time: budgets, programme changes, risk reviews and public reporting. If the data is poorly governed, numbers may look precise while actually meaning different things in different files, coming from unknown sources, or being used beyond their original purpose. Good data governance does not guarantee truth, but it makes data more traceable, consistent and accountable. That helps leaders ask better questions, spot weak evidence earlier, reduce legal and operational risk, and avoid building AI or analytics on a shaky foundation.
How data governance turns messy data handling into controlled use
After viewing, the reader should be able to describe data governance as an ordered process of setting rules, assigning roles, checking quality and controlling use over time.
01
1. Set purpose and rules
Define why the data is being kept or used, what standards apply, and what is not allowed.
Show a box with a document icon and short labels: purpose, definitions, retention, sharing rules.
02
2. Assign roles
Name who owns the dataset, who maintains it, who may use it, and who approves changes.
Show a box with four person icons labelled owner, steward, user, approver.
03
3. Check and document
Record source, meaning and dates, then test for missing, inconsistent or outdated entries.
Show a box with a table icon and side tags reading source, date, definition, quality check.
04
4. Control access and use
Allow only appropriate access, apply safeguards, and make sure use matches the agreed purpose.
Show a box with a lock over a dataset and a small badge saying approved use only.
05
5. Monitor and improve
Track changes, log issues, review compliance, and update rules when data or needs change.
Show a box with a checklist and circular arrow, plus a small issue log panel.
A left-to-right step diagram with five connected boxes. Each box shows one stage in governing data, from setting purpose and rules to monitoring use and correcting problems. Read in order to see how raw data handling becomes documented, accountable and reusable.
Clicking a step highlights it and reveals a short example of that stage, while keeping the full sequence visible.
Describe this without the diagram
The diagram presents five ordered stages of data governance. First, an organisation sets the purpose for a dataset and writes rules on definitions, retention and sharing. Second, it assigns roles such as owner, steward, user and approver. Third, it documents where the data came from, what fields mean and when records were updated, then checks for missing or inconsistent entries. Fourth, it controls who may access or use the data and for what approved purpose. Fifth, it monitors changes, logs issues and improves the rules over time. The main message is that trustworthy data use comes from repeated management, not from collecting data alone.
With reduced motion: A static five-box horizontal diagram with arrows between boxes. The same labels and examples appear without animation or highlighting.
Data governance is about making sure data is handled in a controlled, accountable way. It covers who may collect it, who may change it, what quality standard it should meet, how long it should be kept, and what records should exist about it. Those records are often called metadata, meaning information about the data, such as source, date and meaning.
What goes in? Many kinds of data, from spreadsheets and survey answers to documents and sensor records. Also important are the rules, roles and processes around that data. For example, one team may own a dataset, another may use it, and a named steward may check whether definitions stay consistent.
What happens? The organisation sets policies, assigns responsibilities, defines common terms, checks quality, manages access, and tracks changes. It also decides how data may be shared, archived or deleted. If problems appear, such as missing values or unclear labels, governance provides a process for fixing them and recording what changed.
What comes out is not new data by itself, but a more reliable data environment. People can see where data came from, whether it is suitable for a task, and what limits apply. That makes analysis, reporting and AI work safer and more consistent.
Data governance is often confused with privacy or AI governance. Privacy concerns how information about identifiable people is collected, used, shared and kept. AI governance covers the design, use and oversight of AI systems. Data governance supports both, but it also applies to non-AI uses of data.
How “Maintaining a shared household contact list” maps onto the real thing
| In the story | In the AI system |
|---|---|
| one marking guide | common data definitions and standards |
| rule for recording absent students | agreed handling of missing data |
| coordinator checking names and logged changes | data stewardship, quality control and audit trail |
| only authorised staff can open the full file | access control and responsible data use |
A worked example
A reporting officer at a regional bureau must clear a weekly situation report, but the translation queue contains multiple versions of tables with inconsistent labels and no reliable record of which figures were corrected.
Starting from: Country office submissions, translated tables, email corrections and clearance records stored under inconsistent filenames.
- The reporting officer agrees a single intake template with required fields for source, reporting period, language, version and clearance status.
- A named focal point checks each submission, records its location and returns incomplete entries before they enter the translation queue.
- Edit rights to the master dataset are limited, while translators and reviewers receive only the access needed for their tasks.
- When a country office confirms a correction, the focal point updates the master record and logs what changed, why and on whose authority.
- Superseded working files are retained or deleted according to the applicable records schedule, rather than accumulating indefinitely in shared folders.
Result: The cleared report uses traceable figures from the correct source versions. Translators no longer have to infer which attachment is final, and any later query can be followed through the correction and clearance record.
In a United Nations workplace
At a regional bureau, the reporting officer finds that inputs for the weekly situation report are waiting for translation because filenames, indicator labels and clearance status are recorded differently by each country office.
The officer sets one intake standard, requires metadata on source, reporting period, language and clearance, assigns responsibility for the master dataset, and introduces access controls and a correction log. Translation can then proceed from the right version, while later changes remain traceable.
Before you try thisSituation report inputs may contain personal or security-sensitive information. They must not be entered into an external translation or AI tool unless it is approved for that material. Remove data that is not needed, restrict access to authorised colleagues, and preserve the cleared source and correction history.
Where you may hear it
- A regional bureau review of a delayed situation report
- An ICT governance board reviewing access to a new data platform
- A monitoring review before figures are cleared for a briefing note
- A procurement meeting examining where a vendor will store and process data
- An internal audit interview on who changed the master dataset
Avoid confusion
Avoid confusion
Often confused with
AI governance
The rules, institutions and processes used to guide how AI is built, used and overseen so its benefits and risks are handled in the public interest.
Privacy
Privacy means controlling how information about identifiable people is collected, used, shared and kept, so they are not exposed, tracked or harmed unfairly.
Common misconceptions
Common misconception: Data governance only matters for large technology companies.
Correction: Any organisation that relies on data can face confusion about definitions, access, quality or retention. A small programme team with three spreadsheets can still make poor decisions if one file counts households and another counts individuals without saying so.
Common misconception: Data governance means locking data away so nobody can use it.
Correction: The aim is controlled and appropriate use, not blanket restriction. Good governance often makes data easier to use because people know what it means, who approves access and what limits apply.
Common misconception: If data is accurate today, governance is finished.
Correction: Data changes over time. Staff move, collection methods change, systems are updated and new uses appear. Governance includes ongoing monitoring, change logs and review, not a one-time clean-up.
Common misconception: Data governance is the same as privacy.
Correction: Privacy is one important part when personal data is involved, but governance is broader. It also covers quality, documentation, ownership, retention, access, interoperability and appropriate use, including for non-personal data.
Quick check
Which example best shows data governance rather than just data collection?
Data governance is about the rules, responsibilities and controls around data. Standard definitions, named stewardship, change logs and role-based access are all governance mechanisms. The other options may involve data, but they do not show how its handling is being governed.
Go deeper
Go deeper
The simple version
Data governance is the organised way an institution makes sure data is handled properly: clear rules, clear responsibilities, careful access and regular checks.
The technical version
Data governance is the framework of policies, roles, standards, controls and review processes used to manage data across its lifecycle. It covers data ownership, stewardship, metadata, quality management, access control, retention, lineage or traceability, issue resolution and change management. Its goal is to make data usable, consistent, compliant and accountable across collection, storage, sharing, analysis and deletion. In practice, it links business rules with operational processes so people can tell what a dataset represents, where it came from, what constraints apply, and whether it is fit for a particular use.
Reviewed external resources
Primary sourceAI Risk Management Framework (opens in a new tab)NIST
This link explains risk management concepts and suits readers who want to see how data-related controls fit into wider AI risk management practice.
Intermediatelink checked 2026-07-31
Primary sourceOECD AI Principles (opens in a new tab)OECD
This link sets out high-level principles for trustworthy AI and suits leadership or policy readers looking for an international governance frame.
Intermediatelink checked 2026-07-31
EncyclopaediaData governance (opens in a new tab)Wikipedia
This link gives a broad introductory overview of the term and suits readers who want a quick starting definition and background.
Intermediatelink checked 2026-07-31
VideoVideo explanations: Data governance (opens in a new tab)YouTube search
This is a search results page for videos on the topic and suits readers who prefer to compare several short visual explanations before choosing one.
search results, not one pagelink checked 2026-07-31
DiscussionCommunity discussion: Data governance (opens in a new tab)Reddit search
This is a search results page for community discussion on the topic and suits readers who want to see common questions, confusions and practitioner viewpoints.
search results, not one pagelink checked 2026-07-31
Where this came from
Drafted with a language model through an offline pipeline, checked against this site's schema, and every external link above fetched to see whether it answers. Treat it as a starting point and check anything that matters against the primary sources. How this site is built and what that means.
The one thing to remember
Data governance makes clear who may collect, change, share and delete data, under which rules, with enough traceability to resolve the inevitable query about which version was cleared.