Power BI March 2026 update: what actually matters

March is packed. Translytical exits preview, PBIR becomes the default Desktop format, and DAX gets improvements we’ve been asking for years. Here’s what matters.

---

Translytical task flows — GA

The headline feature of this update. Translytical task flows exit preview and reach general availability.

What does it mean? Your users can now write data directly from a Power BI report. Update records, add rows, trigger workflows — without leaving the report.

Practical use cases:

• Modify a discount in a table and have it reflect in the database
• Annotate data directly on report data points (monthly sales, incidents, etc.)
• Trigger automated actions — for example, changing an offer status to “Accepted” automatically sends an email to the partner contact

Write-back supports Fabric SQL databases, Fabric warehouses, and Fabric lakehouses (for files).

This is a paradigm shift. Power BI stops being a read-only tool and becomes an operational platform. If anyone ever asked you why nobody looks at your dashboard, part of the answer was this: reports couldn’t do things. Now they can.

PBIR is now the default Desktop format

Starting March 2026, Power BI Desktop uses PBIR (Power BI Enhanced Report Format) as the default format for new reports.

What changes? Instead of saving everything in a PBIX binary blob, PBIR saves reports as human-readable JSON folders. This enables:

• Real version control — you can put your reports in Git and get meaningful diffs
• Programmatic editing — scripts that modify reports without opening Desktop
• Collaboration — merge conflicts you can actually resolve, not binaries that overwrite

The service already had this as default since January. Existing reports convert automatically when edited and saved.

If you’re on PBIR-Legacy and want to stay there a bit longer, you can disable the preview in settings. But when PBIR reaches GA, it’ll be the only supported format. Better to migrate now.

Custom totals — no DAX required

One of those improvements that seems small but eliminates daily friction.

Until now, if you wanted a table visual column total to show a Sum instead of the measure’s default behavior, you had to write DAX. Now you can do it with a right-click.

Available options: Sum, Min, Max, Count, Count (Distinct). Without writing a single line of code.

For analysts working with complex measures where the default total makes no sense, this saves time every day.

DAX: JSDoc, 256 parameters and auto-rename

Several DAX improvements that add up:

• JSDoc for UDFs — you can now document user-defined functions with /// (triple slash). Description, parameters, types. Smart autocomplete at last.
• 256 UDF parameters — the previous limit was 12. If you needed more, you had to get creative. Problem solved.
• INFO.USERDEFINEDFUNCTIONS() — new function that returns information about UDFs in your model. Useful for audits and documentation.
• Auto-rename in dependencies — when renaming tables, columns, or measures, Power BI automatically updates references in DAX code. Includes dependency tracking for UDFs.

Auto-rename is particularly important. Before, renaming a table broke every measure that referenced it. Now the engine maintains coherence. Fewer debugging sessions from broken references.

Copilot: more integrated, less decorative

Incremental but relevant improvements:

• Unified UI — the Copilot pane in reports and apps now matches the standalone experience
• Auto-refresh in AI Narrative — the AI narrative visual updates automatically when slicers change, no more hitting Refresh manually
• Improved feedback — you can attach diagnostics with your rating, helping the product team understand context

Is Copilot worth it? We already shared our take — useful for quick exploration, but with serious limitations. These improvements are incremental. They don’t change the fundamental equation, but they make it less frustrating.

Q&A is being deprecated

Microsoft has announced the deprecation of Power BI Q&A, the natural language query feature that existed before Copilot. If you used it in your reports, time to migrate to Copilot.

Not surprising. Q&A never worked well for complex queries and Copilot covers the same use case with better technology. But if you have production reports with Q&A visuals, put it on your backlog.

Direct Lake GA

Direct Lake on OneLake reaches general availability. For those working with Fabric, this matters: import mode performance directly against the lakehouse, without moving data.

If you’re evaluating whether VertiPaq compression is still relevant with Direct Lake — yes, it is. Direct Lake uses VertiPaq internally, but data model optimization is still your responsibility.

My take

This March update is one of the good ones. Not because of a single revolutionary feature, but because everything points in the same direction: Power BI wants to be an operational platform, not just an analytical one.

Translytical turns reports into applications. PBIR turns reports into versionable code. Custom totals and DAX improvements reduce daily friction. Direct Lake GA closes the Fabric loop.

If I had to prioritize one action: migrate to PBIR now. Don’t wait until it’s mandatory. Put your reports in Git, set up a pipeline, and stop treating PBIX files as magic artifacts nobody can review.

Keep exploring

• What is DAX in Power BI and why it matters — DAX fundamentals for beginners
• VertiPaq: how Power BI compresses your data — Relevant now with Direct Lake GA
• Why nobody looks at your dashboard — Translytical solves part of this problem