Governance and administration
Data governance framework: a practical guide
A data governance framework only works if it lives in Power BI: owners, one certified model, and row-level security. Here's how to build one.
Austin Levine
·
Updated
A data governance framework is the written set of rules for who owns each dataset, how numbers are defined, and who can see what. In Power BI it shows up as workspace roles, certified semantic models, row-level security, and sensitivity labels. A good one makes a report trustworthy enough to act on.
Most teams build a data governance framework, save it as a PDF, and never open it again. Then two reports disagree in a leadership review and everyone remembers why the rules mattered. A data governance framework only matters if it runs inside the tools people use every day, which for most companies means Power BI.
This guide skips the 40-page template. It shows you the four questions a framework has to answer, the five steps to set one up in Power BI, and how to tell where you sit on a data governance maturity model.
Key takeaways
A data governance framework writes down who owns each dataset, how numbers are defined, and who sees what.
In Power BI it shows up as workspace roles, certified models, row-level security, and sensitivity labels.
A working framework answers four questions, so you do not need the 40-page document.
Keep it alive after go-live with a few habits that cost about an hour a month.
Why does your data governance framework stall at the first leadership review?
Take a normal Monday leadership review. Finance shows revenue of 4.2 million. Sales shows 4.5 million for the same month. Both numbers came out of Power BI. Now the meeting is about whose figure is right instead of what to do next.
This happens for a plain reason. Each report was built on its own copy of the data, with its own version of the revenue measure. No single model is marked as trusted, so five people built five slightly different ones. 67% of organizations say they don't fully trust the data they use to make decisions (Drexel LeBow and Precisely, 2025), and a leadership review where two reports disagree is what that distrust looks like in a room.
A framework stops the argument because it names one owner and one definition per number. Once that exists, the meeting can move on.
What does a data governance framework actually cover in Power BI?
You don't need the 40-page document. A working framework answers four questions, and everything else is detail that sits under them:
Who owns this dataset? One person's name, not a team alias.
What does each number mean? Revenue has one definition, written in one place.
Who's allowed to see it? Access by role, enforced by row-level security.
How do changes get checked before they reach a report people trust?
If you can answer those four for your top three reports, you already have a framework. The written document just records the answers so the next person doesn't have to guess.
How do you build a data governance framework in five steps?
Here are the five steps we use with clients, in order. Each one maps to a real setting in Power BI, so you can do them this week.
Give every dataset one named owner. Open the workspace, set the contact in the semantic model settings, and put that person's name in the model description. When nobody owns a dataset, nobody fixes it when a number looks off.
Write each measure down once. Revenue, margin, and churn should each have one DAX definition inside a shared semantic model. Mark that model as a certified semantic model so people see the badge and stop building private copies.
Set access by role. Use workspace roles (Admin, Member, Contributor, Viewer) to decide who can edit, and row-level security to decide who sees which rows. Test it with View As before you publish, so a regional manager only sees their own region.
Control how changes ship. Put development, test, and production stages in Power BI deployment pipelines (also written Power BI pipelines) so a change gets reviewed before it reaches the management report. A number that changes overnight with no review is how trust gets lost.
Turn on the audit log and sensitivity labels. The Microsoft 365 audit log records who viewed or exported each report. Sensitivity labels like "Confidential" travel with the file even after someone exports it to Excel.
None of these five needs a consultant to switch on. They need someone with the authority to say "this is the owner, and this is the definition," and then the discipline to keep it in place.
What's a data governance maturity model, and where do you sit?
A data governance maturity model is a simple way to rate how far along you are, usually across five levels. The point of a maturity model data governance rating isn't a grade for its own sake. It tells you the next thing to work on. If you're stuck at level 2, turning on sensitivity labels is wasted effort until you have one owner per dataset.
Level | What it looks like day to day | In Power BI terms |
|---|---|---|
1. Ad hoc | Everyone exports to Excel and emails it around | Personal workspaces, no shared model |
2. Repeatable | A few shared reports, each with its own version of revenue | Several semantic models, none certified |
3. Defined | One owner per dataset, measures written down | Certified semantic model, workspace roles set |
4. Managed | Access is by role, changes get reviewed | Row-level security, deployment pipelines, audit log on |
5. Optimized | Definitions rarely argued, trust is high | Sensitivity labels, monitored refresh, lineage checked |
Most companies we meet sit at level 2. The jump to level 3 is the one that ends leadership-meeting argument, because it puts a single certified model behind the numbers. Getting from level 3 to level 5 is worth doing, but it earns you less than that first jump.
How do you keep the data governance framework alive after go-live?
A framework stops working if no one keeps it updated. Three habits keep it honest, and together they cost about an hour a month:
Review the certified model owners every quarter. People change teams, and a dataset with no owner is where wrong numbers start.
Read the audit log once a month. If a sensitive report is being exported to Excel and emailed around, your row-level security is being worked around, and you want to know.
Keep every change flowing through deployment pipelines so nothing reaches production untested.
Poor data quality costs organizations an average of 12.9 million dollars a year (Gartner), so an hour a month spent on these three habits is easily worth it. The habits matter more than the document. A one-page framework that people follow beats a 40-page one that sits in a folder.
If you want to see where you stand before you start, the free Power BI Governance and Cost Scorecard rates your setup in about five minutes and shows your top three risks. If you'd rather walk through your own reports with someone, book a governance scoping call and we'll look at your certified models, row-level security, and deployment pipelines together.
Frequently asked questions
Is a data governance framework only for big companies? No. A 20-person company with three Power BI reports still needs one owner per dataset and one definition of revenue. The framework is smaller for a small team, not absent.
How long does it take to set up? The first useful version takes a week or two: assign owners, certify one model, turn on row-level security. Reaching a higher maturity level is a matter of months, done step by step.
Do we need Power BI Premium for this? Most of it works on Power BI Pro. Certified models, workspace roles, row-level security, and deployment pipelines are all available without Premium. A few features, like very large models and some sensitivity-label controls, need Premium or Fabric capacity.
Who should own the framework? Someone with authority over the numbers, usually in finance or a data team lead. Governance without authority gets ignored, because the owner can't make the definition stick.
What's the difference between a framework and a maturity model? The framework is the set of rules you run. The maturity model is how you measure whether those rules are working, so you know what to improve next.
Read next: Power BI row level security, explained and How Power BI incremental refresh really works.
