Data owner, data steward: who does what?

Two roles, one common confusion

In most governance projects, both titles appear in the very first meeting, and then blur. A data owner is appointed who ends up fixing product records themselves, or a data steward is asked to settle a dispute between two departments that disagree on what an active customer is. The result: decisions are not taken at the right level, and the groundwork is done by no one.

The distinction is simple, though: the data owner is accountable for the data, the data steward takes care of it. One decides, the other brings the decision to life, and each needs the other for governance to move beyond the slide deck.

The data owner: accountable for the data

The data owner is a business leader, not an IT specialist. They own a data domain (customers, products, suppliers, accounting entries) because their activity produces that data or depends on it the most. They often sit on their department’s management committee, or very close to it.

• They approve definitions: what an active customer, an obsolete product or a delivered order is
• They set the expected level of quality and the rules that measure it
• They decide who may view or change the data of their domain
• They settle disagreements between teams, and take issues beyond their scope to the governance committee

They spend only a few hours a month on it, but those hours carry weight: without their approval, a quality rule remains a suggestion, and a definition remains one opinion among many.

The data steward: bringing the rules to life

The data steward is the data owner’s operational relay. They know the data closely, often because they enter or use it every day, and they have time genuinely set aside for the role, written into their objectives.

• They document the domain in the data catalog and the business glossary
• They track quality indicators and analyse anomalies
• They fix discrepancies, or have them fixed at the source by the right teams
• They propose new rules to the data owner and report what is blocking

They are often the person users call when a figure looks wrong. A good data steward can explain where a piece of data comes from, how it is calculated and whom to contact to correct it.

Data owner and data steward side by side

Assigning them from one team to the whole group

The right unit is not the application but the data domain. A single domain spans several tools (the ERP, the CRM, the planning tool), and it is precisely between those tools that inconsistencies arise. So you appoint one data owner per domain, and one or more data stewards depending on data volume and geography.

• Finance: management control often owns the chart of accounts, the entities and the steering indicators
• Supply chain: operations own products, bills of materials and flows
• Production: the industrial department owns technical data, equipment and performance indicators
• Sales: the sales department owns customers, prices and terms of sale

In a group present in several countries, a group data owner, who sets the shared definitions, is often combined with local data stewards, who apply them close to the sites and ERPs. A data office coordinates the whole: it trains the new roles, equips the catalog and prepares the governance committees.

The most common mistakes

• Making IT the data owner: IT hosts and secures the data, but cannot decide its business meaning
• Appointing data stewards without giving them time: the role stays on paper and quality does not improve
• Creating too many roles at once: three well-run domains beat fifteen half-run ones
• Not writing down who decides what: without a responsibility matrix, arbitration goes up to general management, or never happens

Where to start

1. Pick two or three critical domains, those that feed your key indicators or your upcoming AI project
2. Appoint a data owner and at least one data steward for each, with a short mission letter
3. Write the definitions of the terms under debate, and publish them in a shared glossary
4. Measure a few quality rules and track how they evolve every month
5. Set up a light governance committee that arbitrates and widens the scope step by step