Data Governance in a multinational: global owners, local stewards and who decides what

The problem: a global owner far from the context

This is a recurring debate among data leaders in companies with a presence in several countries. A global Data Owner brings consistency: one definition of "customer", "active supplier" or "revenue" for group reporting. But if that owner is far from the business, they miss the local context, decide slowly, and end up approving without judgement or blocking reasonable changes. A fully local model has the opposite problem: each country defines its own and the group cannot consolidate anything.

The usual way out is a federated model: a small global layer that fixes the essentials and regional or local layers that run day to day. It is the same principle DAMA-DMBOK uses for data governance operating models: centralised, decentralised or federated, depending on how much autonomy each unit has.

What should be global and what local

A practical test is to ask whether the data is consolidated or compared across countries. If so, the definition must be single. If not, it can be local.

Five questions to assign a Data Owner

Before naming owners, answer for each domain and critical data element:

• Who defines what this data means?
• Who defines the rules under which it must be created and maintained?
• Who can decide to change those rules?
• Who is accountable for the business impact if the data is wrong?
• If two areas or countries disagree, who decides?

If any answer is "nobody" or "it depends", the owner role does not exist yet, even if someone has the title.

Why stewards almost always work better locally

The Data Steward does the work: spots inconsistencies, maintains the glossary, applies quality rules. That requires knowing the business processes and systems of their country, which are often different. They also know local data quirks: not every country uses postal codes, some identifiers start with zero and get lost when loaded as numbers, and regulation changes which customer data can be updated and how.

Even so, an executive or global steward is useful to maintain group definitions and coordinate local ones. Their job is not to approve every change, but to make sure local changes do not break what is consolidated.

How to resolve global vs local conflicts

1. Classify the change. Does it affect consolidated data or only local operations?
2. Assess the impact. The steward proposes and documents which reports, dashboards and processes are affected.
3. Decide at the right level. Local if the impact is local; the data governance body if several countries are affected.
4. Record the decision. A definition without a date, owner and reason gets argued again in six months.

A useful signal: if the global lead needs extra local context to understand a dashboard, they are probably seeing more detail than needed and the data products delivered to them should be reviewed.

Common mistakes

• Global owners with no other function. If being an owner is their only job, they have no contact with the data or its users.
• Copying the headquarters framework without adapting it. Keep the high-level principles and protocols, and adapt the operating model to each country.
• Not documenting what is global. Without that list, each country interprets it its own way.
• Forgetting regulation. Cross-border transfers of personal data and local requirements should be reviewed with legal advice; this article is for guidance only.

Link to the AI Act

If the group uses high-risk AI systems, Article 10 of the AI Act requires data governance practices for training, validation and testing data. With the Digital Omnibus postponement, those obligations apply from 2 December 2027 for Annex III systems. Knowing who is responsible for each dataset in each entity is a useful base for when that time comes.

Frequently asked questions

Should data owners be global or local in a multinational?

It depends on the data. Data consolidated in group reports needs one definition and therefore an owner with global scope. Data with mostly local impact can have a local owner. What matters is writing down which data is which.

Where should data stewards sit?

Usually in each country or business unit, because they work with the data daily and know the local processes, systems and regulation. They can coordinate with a global steward who maintains group definitions.

What if a country wants to change a definition that affects the group?

The local steward proposes it, the domain owner assesses the impact and, if several countries are affected, a data governance body with local representation decides. If the impact is local only, the local owner approves it.