Data governance is the set of decisions about who is accountable for data in an organisation, how its quality is measured, and on what terms it is shared. A career in the field is not about operating a tool but about holding a specific decision-making role — and that distinction is where the questions we hear most from clients begin.
Quick Overview
What you’ll learn:
- Which roles an organisation actually fills, and what each of them decides
- How governance work differs from data engineering and data analysis
- Which regulatory competences are unavoidable on this path
- What a sensible first assignment looks like for someone entering the field
Who this article is for:
- Analysts and data engineers considering a change of track
- Managers who need to fill a data owner role in their area
- Compliance specialists taking over responsibility for data
Reading time: 8 minutes
”What does data governance actually do?”
The shortest answer: it settles disputes about data before they become disputes about a report. Two departments show a different customer count because each treats an inactive customer differently — and that is not a technical problem. Settling it requires naming the person with a mandate to define the term, and a place where the definition is recorded and visible to both departments.
The working scope follows from that: a business glossary, a data catalogue with named owners, quality rules with explicit thresholds and an exception procedure, sharing rules, and a register of decisions. Tools only service this layer. The documentation at learn.microsoft.com states it plainly — catalogue, glossary, quality rules and reporting on the state of data are platform features, while the owner and the definition have to come from the organisation.
”Which roles can be filled here?”
Data owner. A business role rather than a technical one. The owner is accountable for the definition of a term, permitted uses and access decisions within their area — usually the head of a function, not a database administrator.
Data steward. Works at the seam: translating business rules into technical ones, maintaining quality in specific datasets and keeping the register of exceptions. This is the most common entry point for someone with an analytics background.
Data protection officer. Accountable for the lawfulness of personal data processing, and a role distinct from the data owner, with its own independence. Its presence in the structure does not relieve owners of accountability for definitions and access.
Chief data officer. Accountable for data policy as a whole and for keeping areas from deciding in contradictory ways. The role appears once the number of owners is large enough that disputes no longer settle between them.
A mature organisation fills these roles in parallel rather than in sequence — a policy written without owners produces a document nobody applies.
”How is this different from data engineering and analysis?”
By the object of the decision. A data engineer is accountable for data arriving and being processable. An analyst is accountable for the conclusion drawn from it. Someone working in governance is accountable for whether the data may be used for a given purpose, what a term means and who gets to decide. That is why the daily work of the role is largely agreeing and recording decisions rather than writing queries.
The move is most natural from analytics: an analyst already knows the data, knows which terms are contested and which reports diverge. The missing part is usually the craft of running an agreement process and familiarity with regulatory requirements. If you are choosing between tracks, it is worth reading the two neighbouring descriptions: the data engineer path and the data analyst path.
”How much regulation is involved?”
More than the job title suggests. The GDPR, published at eur-lex.europa.eu, establishes the data protection officer role, the obligations of controllers and processors, and the requirement to identify a lawful basis for processing — which in practice means a decision to share a dataset needs a recorded justification rather than an assumed one. The Regulation on European data governance, also available at eur-lex.europa.eu, adds a framework for the re-use of protected public sector data together with rules for data intermediation services and data altruism.
The implication for a career path is straightforward: regulatory competence here is not an optional extra reserved for lawyers. Without it the data owner role collapses into approving access requests with no criterion, which is simply passing risk downstream.
”Which platform should I learn?”
Learn the concepts across platforms rather than the buttons of a single one. The catalogue documentation at cloud.google.com and the governance documentation at learn.microsoft.com describe the same objects under different names: a catalogue of assets, a business glossary, lineage, classification of sensitive data, quality rules and access policies. Knowing that mapping makes a migration an exercise in translation rather than relearning.
What transfers between tools is the discipline: a named owner for every asset that feeds a decision, a definition stored where both sides of a dispute can see it, and an exception route that is documented instead of improvised.
Read Also
- Why effective data management is essential for an organisation — the business case for the whole area
- How to become a data engineer — complete career path — the neighbouring, more technical track
- How to become a data analyst — complete career path — the most common starting point for governance work
Build Your Skills
This topic is developed in the Data Governance and Data Quality Management and Data Governance and Data Quality — Framework courses. The first works on quality rules and roles; the second arranges them into a coherent framework for the whole organisation. Contact us to match the scope to the stage your company is at.
Frequently Asked Questions (FAQ)
What does data governance look like in an organisation starting from scratch?
It starts with a catalogue and owners, not with a tool. The practical order: inventory the datasets used in decisions, assign an owner to each, define the contested terms, set quality rules for the highest-impact datasets, and only then choose a platform. The reverse order ends with a deployed catalogue whose owner fields are empty.
What is master data management?
It is the governance of datasets used across many processes at once — customer, product, supplier, location. Their defining feature is that an error propagates into all those processes simultaneously, which is exactly where quality rules and an unambiguous owner carry the most value. Master data is usually the first area in which an organisation names roles.
How is data quality measured?
Through rules with explicit thresholds: completeness of required fields, uniqueness of identifiers, conformance to a value list, freshness against the source, and consistency across systems. Each rule needs an owner and an exception route, because a rule without an exception route gets switched off during the first urgent release. Reporting on rule state is a platform feature, as the documentation at learn.microsoft.com describes.
Do you need personal data expertise to work in data governance?
Yes, to the extent of recognising when a matter requires the data protection officer. A data owner does not replace that role but does make access decisions, so they have to know when a decision exceeds their mandate. The full set of obligations is in the GDPR text published at eur-lex.europa.eu.