Who sees what: row-level security in Power BI, explained
Row-level security (RLS) lets one Power BI report show every person only their own region, team or customer. The rule is usually one line of DAX. Most problems come from everything around it. Here is the whole picture on one page, with a free guide and checklist to download at the end. Not sure what you need to read? Take the 2-minute RLS check.
What RLS does
Without RLS, you either build one report per audience or everyone sees everything. With RLS, you build one model and one report, and Power BI filters the rows for each person when they open it.
You define a role in the semantic model. The role holds a DAX rule that keeps some rows and hides the rest. Then you add people, ideally security groups, to that role in the Power BI service.
Static and dynamic RLS
| Static | Dynamic | |
|---|---|---|
| Rule | [Region] = "North" | [Email] = USERPRINCIPALNAME() |
| Who sees what lives in | The roles (one role per group) | A security table (one row per person and region) |
| New region or person | New role, republish | Add a row |
| Best for | A few fixed audiences | Many users, or access that changes often |
How the filter travels
In dynamic RLS, the rule filters only the security table. Every other table is filtered by following relationships from there, from the "one" side to the "many" side.
The first hop goes "uphill", from the security table to the dimension. On that one relationship, set the cross filter direction to Both and tick Apply security filter in both directions. Keep the others single-direction.
This leads to the most common leak: a table the filter never reaches, such as a targets table linked only to the date table, is shown in full to everyone in the role. Every table is either on the filter path, blocked with FALSE(), or a leak.
Four rules worth remembering
- RLS only filters Viewers. Workspace Admins, Members and Contributors, and anyone with Write on the model, see every row. Give report users Viewer access, and Build permission if they need to build.
- Roles add up. Someone in two roles sees the rows of both. Two filters in one role mean both must match. And a user in an RLS-only role plus an OLS-only role gets visual errors.
- The storage mode matters. DirectQuery can't use PATH or CUSTOMDATA in role rules. Direct Lake needs a decision about which identity reaches the data. Always test in the mode you'll publish.
- Deployments move roles, not people. On the first deploy to Prod, roles have no members and users see blank visuals. Add security groups once in each stage.
Not covered by RLS
RLS filters rows in the model. It doesn't hide saved slicer selections, typed titles, hidden columns (use OLS), files people forward after export, or reports shared with publish to web. And Copilot answers with exactly the data the user can see, gaps included.
Testing it properly
- Use View as with Other user and the role ticked together, for every role a person holds.
- Test with a real Viewer account, and a real guest if you share outside your organisation.
- Time the main pages with and without the role (Performance Analyzer or DAX Studio).
- Check who has access every month: workspace roles, model permissions, role rules, role members and the security table.
The full series
This page sums up a 20-part series. Each part takes one problem and its fix, and every part has its own page:
Check your RLS: what should you read?
Fourteen quick questions, about two minutes. Each one is a common RLS mistake. At the end you get a reading list with only the chapters you need, linked straight to the right page of the free guide. Nothing you answer is saved or sent anywhere.
Download the free guide
Who Sees What: A practical guide to Row-Level Security in Power BI
- All 20 parts as chapters, each with DAX or steps to copy
- A "check before go-live" box in every chapter
- The RLS go-live checklist on two pages
- How to check every item, click by click, in Power BI Desktop and the Power BI service
- A free Fabric notebook that audits every model in a workspace: who skips RLS, every role, rule and member, leaky tables, and tests as real users
Thanks. Your downloads are ready.
Download the guide (PDF) Audit notebook (Fabric)
Not you? Use a different email
Most checks need no code: the guide shows the clicks. The notebook is for monthly audits in Fabric; it only reads and changes nothing.
One line of DAX. A lot around it.
Get the model right, give report users Viewer access, test as real people in the stage you publish to, and check who has access every month.
Want it checked on your own model? The RLS review runs this checklist for you, at a fixed price.
Questions, or a part you'd add? Send me a message or find me on LinkedIn.
All examples use made-up data. Power BI and Microsoft Fabric change often; check Microsoft Learn's row-level security pages for the latest details.