Data owner vs data steward: who does what

When a report shows the wrong number, the question "who is responsible?" often has no answer. IT says the system works correctly. Sales says the data came from finance. Finance says nobody told them about the change.

Two roles fix this: the data owner and the data steward. They are at the heart of any data governance program, and they are often confused. This article explains how they differ and how to introduce them without hiring anyone.

The data owner: makes the decisions

A data owner is a business leader accountable for a particular set of data, such as customer, product or financial data. The owner:

• approves definitions and quality rules;
• decides who can access the data and for what purpose;
• resolves disputes between departments;
• decides where to invest in fixing problems;
• answers to management for the quality of the data.

The owner should be someone who feels the consequences of bad data and has the authority to change how people work. That usually means a senior manager in the business, not a system administrator. The commercial director is a more natural owner of customer data than the CRM administrator.

The data steward: does the daily work

A data steward looks after the data day to day. The steward:

• monitors quality checks and investigates errors;
• maintains reference data and the business glossary;
• answers colleagues' questions about what the data means;
• proposes new rules to the owner;
• reports recurring problems and how they are progressing.

A steward is usually an experienced specialist from the business team who knows the data and how it is used: a senior analyst, an operations specialist or an accountant.

And what about IT?

IT teams keep the systems running: they store the data, secure it, back it up and build pipelines. This role is often called the data custodian. IT can implement a rule, but it should not decide what the rule says.

Here is how the three roles work in a simple example. Customer records contain duplicates:

1. The steward finds the problem through a quality check and estimates its size.
2. The owner decides how duplicates are identified and which record is kept.
3. IT implements the rule in the system and adds an automatic check.
4. The steward confirms that the check works and tracks the number of duplicates over time.

How to introduce the roles

Start with one data area

Choose the area where problems hurt most, such as customers or products. Do not assign owners for all company data at once: the roles stay on paper. A 90-day governance plan describes this approach step by step.

Use people who already do the work

In most companies, someone already corrects the reference data and answers questions about it. Make that work explicit, and allow time for it. These roles rarely need new hires, but the extra responsibility must be reflected in the person's priorities.

Write down the responsibilities

One page is enough: which data the role covers, what the person decides or does, and how issues are escalated. Publish it where everyone can find it, along with the names.

Set up a simple rhythm

A short monthly meeting between owner and steward works well: quality figures, open problems and decisions needed. When several areas are running, the owners can meet quarterly to settle questions that cross departments.

Measure the result

Choose a few numbers for each area: share of empty key fields, number of duplicates, time to fix an error. If the numbers improve, the roles are working. If nothing changes after several months, find out why before expanding.

Common mistakes

• Making IT the owner. IT then has to make business decisions, and business teams feel no responsibility.
• Appointing a very senior owner who has no time. Better a manager who actually takes part than a top executive whose name is just on the list.
• Giving the role without the authority. A steward who cannot get a decision from anyone will soon stop raising problems.
• Assigning stewards with no time set aside. If the role is simply added to a full workload, it will always lose out to urgent tasks.
• Having several owners for one set of data. When several people are responsible, nobody is. Other departments can take part, but one person decides.

Summary

The owner decides, the steward does the daily work and IT provides the technology. This simple division ends most arguments about whose data it is. Start with one area, name real people and give them time and authority, and you will have a working process to repeat in other areas.

If you need help setting this up, see our Data Governance service: we assign roles together with your team, set up the rules and leave you with working checks.