OpenLineage lets you trace a large number of ETL jobs that mutates tables, letting you trace changes to tables significantly easier. Databricks Unity Catalog has support for Column Level Lineage and it's highly useful: View data lineage using Unity Catalog - Azure Databricks | Microsoft Learn Databricks Demo   AWS also supports OpenLineage:  Amazon DataZone introduces OpenLineage-compatible data lineage visualization in preview | AWS Big Data Blog It's actually quite easy to get OpenLineage working in Fabric Spark since you already have the JAR pre-installed: Column level lineage in Fabric Spark with OpenLineage and stashing the lineage in Delta Lake | Raki Rahman   All we need is a way to store and query historical state (e.g. in Delta Lake), and a UI component that hooks into the Lakehouse catalog and offers an API on top of OpenLineage schema. The OneLake Catalog UI could look like this:   To prove this can be fairly easily achieved since the OpenLineage events are robust with a stable schema, I vibe-coded this little static site that works with the default Fabric Spark OpenLineage events: Marquito — OpenLineage Visualizer       ... View more

Category color in the new Card visual resets when repointing to a different data source. Even when using the same field, changing the data source causes the category/header color to revert to default. This breaks formatting consistency and requires manual rework during dataset switching (e.g., Dev → Prod). Formatting should persist when the same field is used. ... View more

We have been observing that Power BI semantic models are automatically triggering a refresh immediately after being republished to the Power BI Service. In the refresh history, this is tagged as "On-demand", even though no manual action was performed. Upon reaching out to Microsoft, we received the following confirmation: “This is the expected behavior from the Power BI Service once a semantic model is republished. The behavior is by design and applies to all semantic models in Import mode (or those using auto-aggregations). This behavior has been consistent for several years, and there are currently no plans to change it in the product roadmap.”   While this behavior can be beneficial in certain scenarios—particularly for smaller models that complete refreshes within seconds—it introduces challenges in the following cases: Incremental Refresh Models:  In some situations, the refresh may process more data than expected, leading to longer execution times. Large or Complex Models: Models with sizes exceeding 5–10 GB or with high complexity may require several minutes to refresh, increasing resource consumption.   This behavior can significantly impact capacity resources, especially in Fabric capacity-enabled workspaces, where such unintended refresh activity may lead to unnecessary resource utilization and potential performance issues.   It would be highly beneficial if this behavior could be controlled at the user level or at least at the tenant level, allowing organizations to prevent unintended refreshes and better manage capacity consumption. ... View more

Paginated reports - please make it available in Pro as well as Premium .. thank you ... View more

This feels like a genuine design bug related to how Fabric and Power BI fit together. I'm working in a workspace backed by an F64 Fabric capacity. Using the same user account (Fabric Free), I can edit and refresh semantic models just fine through notebooks, pipelines, the Fabric CLI, and Git integration (PBIP/TMDL). All of those are documented Fabric paths, and they work as expected. If I try to do the same thing from the Power BI Service UI (for example, manually refreshing the semantic model) I get hit with an "Upgrade to a paid Power BI license" message. Same model, same workspace, same capacity, same user, different outcome. The only difference is whether I clicked a button in the UI or initiated the action through a Fabric surface. The issue is about inconsistent enforcement of the same capability. Whether an operation is allowed shouldn't depend on whether it's triggered via the UI versus Fabric orchestration. That makes behavior feel non‑deterministic and forces operators to avoid the UI entirely for tasks that are otherwise allowed and already running on paid capacity. It also creates artificial licensing pressure. The capacity is paid for, the operation is supported and allowed via multiple paths, but you still need Pro just to press the UI button. That's hard to justify, and it actively worsens governance and operability. There are already community threads where people hit parts of this problem (for example, being blocked from UI refresh with Fabric Free licenses on capacity), but none really address the bigger issue: Fabric allows the operation everywhere except the Power BI UI. At minimum, this should be reconciled. Either the UI entitlement checks should align with what Fabric already allows, or there should be a clear, capacity‑scoped "operator" capability for management of Power BI items. Even clear in‑product guidance explaining why automation is allowed but UI actions are blocked would be better than the current experience. Right now it just feels fragmented. Fabric presents itself as a unified, capacity‑first platform, but the Power BI UI still behaves as if it's living under a separate, legacy licensing model. ... View more

Matrix visuals could significantly enhance the user experience by emulating Excel's intuitive sorting, searching, and filtering capabilities, even for hierarchical columns. This idea would the current limitation in Power BI's matrix visuals, which lack a comparable experience to Excel's established standard for data manipulation. ... View more

Create a warning when importing data that does use query folding in Power BI Desktop. ... View more

We don't currently use Fabric Copilot because of the CU risk or our Fabric capacity, and as a result have it turned off on our tenant.   Power BI report builders and consumers in my organization resort to copying and pasting code and issues to M365 Copilot for help, which requires additional time and loses context of the object you are working on, and is not agentic. We are losing productivity gains and potentially better results.    Pro code developers are starting to use GitHub Copilot in VS Code but this is not accessible for everyone.   I can see that the Fabric and Power BI Copilot features released to date and data agents would be beneficial for both types of users, however due to the cost it will remain turned off for the foreseeable future unless the costing changes or we can bring our own user licence for AI features.   Most of the Power BI builders/developers in my org have a M365 Copilot or GitHub Copilot licence, I happen to have both at the moment. It seems strange that we can't use these Microsoft licences that we already pay for to use Copilot in Fabric and Power BI. Can we please have the ability to bring our own user AI licences.   The uptake of Fabric copilot features will increase and we will require more user licences as a result. GitHub Copilot and M365 Copilot also have better governance features for organization admins to track queries and usage. ... View more

Anyone who uses PowerBI long enough has had a PBIX with 50-60 measures and it has become time consuming and tedious to mantain. In Excel you can do a quick find/replace to edit several formulas - in PowerBI you need to select each one individually. An "excel-like" interface for editing measures would save a lot of time! This would take PowerBI to the next level regarding productivity. I've prepared a mockup for this as well as a DAX Editor. Let me know what you think. Mockup: https://i.imgur.com/z6TBOQb.png?1 ... View more

  Description / Problem Our organization has fully enforced phishing‑resistant authentication methods for all employees and administrative accounts, in line with Microsoft security best practices. These include passkeys and FIDO2 hardware security keys (for example, YubiKey). However, we currently need to maintain Power BI on‑premises data gateway (February 2026 version) as an exception to these policies. The reason is that the embedded web authentication engine used by the gateway does not support FIDO2 authentication. Because of this limitation, we are unable to enforce phishing‑resistant authentication methods for gateway sign‑in and administration. Current Behavior The Power BI on‑premises data gateway currently supports: Password‑based authentication, or Passwordless authentication using number matching While number matching improves security, it does not meet the definition of phishing‑resistant authentication required by many enterprise security and compliance standards. Impact Forces organizations to maintain authentication exceptions Creates a security gap in otherwise fully phishing‑resistant environments Prevents full alignment with Zero Trust and passwordless‑first strategies Affects enterprises with strict regulatory, compliance, or security mandates Requested Enhancement Please add FIDO2 authentication support (passkeys and hardware security keys) to the Power BI on‑premises data gateway in a future update. This would allow organizations to: Fully enforce phishing‑resistant authentication across all Fabric and Power BI components Eliminate password‑based authentication for gateway administration Align the gateway with Microsoft’s broader identity and security roadmap ... View more

Secondary axes are automatically managed by the system in some cases, making it problematic for report creators with no control over intervals, ticks, and other crucial components, limiting effectiveness of various types of analysis. It would be great if the secondary axis included comprehensive format settings so that greater control over displayed data is provided. ... View more

When visiting the monitor page, not all refresh events appear. Monitor page does not show all semantic model refresh failures. The failed refresh events shows in refresh history of semantic models. Even the details are available in onelake catalog when searched for the semantic model -> monitor -> we could see the failures events but in monitor hub the details are not available. The Engineering team confirmed it by design but however the monitor hub should provide all the details so that it provides a centralized view of job execution health, progress, and outcomes to quickly identify problems and take action. ... View more

In need to programmatically retrieve the status of all gateways in a cluster, not just the primary. Specifically, require status per gateway, detection of states like 'Partially offline', and alerting when any gateway becomes unavailable. Currently, APIs seem to return only the primary gateway status, with no visibility into secondary nodes and also alerting options, such as triggering alerts if a threshold of refresh failures is exceeded, for example, 20% failures within the last hour to ensure high availability and monitoring. ... View more

Power BI treats text values in the semantic model (VertiPaq / DAX engine) as case‑insensitive. While Power Query (M) is case‑sensitive, once data is loaded into the model: ... View more

While documentation for Dataflows Gen1 exists, clearer guidance on roadmap, support lifecycle, and migration expectations is needed, as Gen1 is now a legacy artifact. Although migration guidance to Gen2 is available, more detailed scenarios—such as migrating Dataflows Gen1 with linked entities (Linked dataflow gen1's) to Gen2—along with improved clarity, additional context and consolidation across documentation, would help plan and transition more effectively. ... View more

Summary As of Power BI Desktop March 2026 (v2.152.x), dropdown slicers visually accept keyboard input but no longer perform any filtering. Typing is silently ignored with no error, toggle, or UI indication. This behavior previously worked and appears to have been removed without notice. Description Dropdown slicers still present a visual affordance that suggests typing will filter values (cursor focus, keystroke acceptance), but no filtering occurs. This creates a false affordance: the control appears interactive, but user input has no effect. The only way to enable search is to change the slicer style to List and enable Header → Search. There is no in‑product guidance indicating that search is now List‑only, and no indication in the dropdown slicer UI that typing is unsupported. This behavior disproportionately impacts enterprise datasets with high‑cardinality fields such as Work Item IDs, Ticket Numbers, or Account IDs, where dropdown slicers were historically usable via typing. Steps to Reproduce Open Power BI Desktop March 2026 (v2.152.x) Add a slicer visual Set Style = Dropdown Bind the slicer to a text field (e.g., formatted Work Item ID) Click the dropdown and type any characters Result: No filtering occurs. Typed input is ignored. Expected Behavior One of the following: Typing filters dropdown slicer values (previous behavior), or Dropdown slicer disables text input entirely, or UI guidance indicates that search is supported only in List slicers Actual Behavior Keystrokes are accepted visually No filtering occurs No warning, tooltip, or setting explains the behavior Impact Significant analyst time wasted troubleshooting data types, sorting, and slicer configuration Erosion of trust in Power BI UI consistency Repeated rediscovery of the same regression across teams Increased support burden in large enterprise environments Suggested Fix Restore typing‑to‑filter behavior for dropdown slicers, or Remove the typing affordance from dropdown slicers, or Add explicit UI messaging (e.g., “Search available only in List slicers”) Classification UX regression / breaking change without discoverability Final Thoughts If this behavior is intentional, could the dropdown slicer disable keyboard input or surface guidance that search is List‑only? ... View more

Summary In a Data Factory Copy activity, the source and destination Connection fields support Add dynamic content, so they can be bound to a pipeline / library variable and parameterized per environment. The session-log location connection (logSettings.logLocationSettings.externalReferences.connection, set under the Settings tab > Enable logging) does not expose that affordance: it accepts only a fixed connection picker and serializes as a hard-coded connection GUID in the pipeline definition. In the same logLocationSettings block, path accepts dynamic content, but connection does not. The connection reference variable type (Variable Library, preview) — designed precisely to parameterize external connections across environments — currently lists supported items as Notebooks and User Data Functions only, not Data Factory pipelines, so it does not cover this scenario either. Microsoft Support confirmed this is by design (Support Case 2605270040006335, May 2026) and recommended I submit it here for the engineering team. Business impact Because the connection cannot be parameterized, it must be re-pointed manually in every higher environment after every deployment. In a clone-based deployment model (e.g., fabric-cicd publish_all_items), every deployment re-pushes the source-environment connection GUID into prod and overwrites the manual fix. This is not theoretical — it caused a production incident in our environment when a routine redeployment reset the prod log connection to a non-production storage account. Requested enhancement (either resolves it) Add "Add dynamic content" to the Copy activity session-log location connection field, consistent with the source/destination Connection fields — so it can bind to a pipeline / library variable. OR extend connection reference variable (Variable Library) support to Data Factory pipelines, including the Copy activity log-location connection. Current workaround (insufficient as a product answer) We substitute the connection GUID per environment at deploy time using fabric-cicd's parameter.yml find_replace. It works but requires teams to maintain external find/replace rules for a value the product should parameterize natively, and it does not help teams using Deployment Pipelines or Git-only promotion. References Copy activity logging: https://learn.microsoft.com/azure/data-factory/copy-activity-log Connection reference variable type (preview): https://learn.microsoft.com/fabric/cicd/variable-library/connection-reference-variable-type Microsoft Support Case 2605270040006335 (closed by-design; redirected here) ... View more

Customer is seeking a solution to retrieve gateway roles assigned to all users across all gateway connections within their tenant. However, based on our current assessment, there is no out-of-the-box method or available API that supports retrieving this information at a tenant-wide level. Could you please confirm if there are any alternative approaches or solutions to achieve this requirement? Raising this on behalf of our customer. ... View more

If it was possible to offer a robust and customizable style management system for the matrix, it could be the key to enhancing the professional appearance of tables and matrices. Further, giving report creators the ability to adjust gridlines, color, style, and the addition of custom icons, bars, sparklines, and other elements, along with a variety of pre-designed styles, would significantly improve style management options for tabular elements. ... View more