Currently, to allow business users to access Fabric Warehouse or Lakehouse data through Excel, we need to provide workspace-level access. This exposes additional artifacts and workspace items that users do not actually need or donot need to view or have access to. It would be great if Microsoft Fabric supported more granular access specifically for Excel connectivity, allowing users to: Access only Warehouse/Lakehouse SQL Endpoints Query authorized tables/views from Excel Avoid visibility into unrelated workspace artifacts, reports, notebooks, pipelines, etc.   Many business users still rely heavily on Excel for analytics, so enabling secure and limited Excel-based access would greatly improve governance while supporting real business workflows. This would help maintain the principle of least privilege while improving the overall user experience in Fabric and analytics. ... View more

In our multi-tenant solution we use partitions in Apache Kafka to separate data for each tenant. To process events from a specific partition (aka tenant) and stream them into a tenant specific destination (KQL database in this case), it would be nice if it is possible to subscribe to a specific partition when configuring an Apache Kafka datasource in an eventstream.   Today it is mandatory to do 'Event processing before ingestion' and use transformations to filter tenant specific events and route them to the tenant specific KQL database. When you can subscribe to a specific partition, it becomes possible to use 'Direct ingestion' instead. ... View more

I recently raised a ticket to MS because I could not "choose" an account in the salesforce.com connector (with Oauth2 authentication) Vs the one used in Fabric Dataflow itself. This because the account used to devlop reports does not have direct access to SFDC (as per complnay policy the account in SFDC must be a "Personal account", where in Fabric I am using a functional account). By desing curretly this is not possible insted it is POSSBILE in PowerBI Desktop data models for example (diffeent login). Request/Suggestion would be to "align" the behavior in DATAFLOWs to behave as in Desktop Datamodel    Below the outcome of the ticket I raised TrackingID#2606030050003758 Call summary: ->You are using Functional account to login to Power BI service and then trying to use the personal account for Salesforce source login in the semantic model settings as the Functional account does not have access to Salesforce. ->However, Salesforce only supports Oauth2 authentication currently and uses the account that is signed in to Power BI to access the source. Due to this, you could not successfully sign in to the Salesforce. ->We confirmed that this behavior is expected as the only authentication method we have for Salesforce is Oauth2. ->With this we concluded to use the personal account which has access to the Salesforce data source as of now. For any feedback regarding the product,  I recommend you to please visit "Fabric Ideas - Microsoft Fabric Community"  as this is the only place where product feedback is looked into. ... View more

I'm trying to create a relationship between two tables, but the "New Relationship" window hides the "Save/OK" button at the bottom of the screen. I can't scroll within the window, resize it, or even resize the Power BI window itself because it's in full-screen mode. The screen zoom is set to 100%. Nothing works. This is a fundamental user interface problem that renders one of the core features unusable. It's unacceptable for a dialog box to hide its own confirmation button. And one last thing: this program of yours is Microsoft gar6age It's useless. Everything gives an error; every equation gives an error. I'd rather use a program from the Stone Age than use this one again.     ... View more

Currently when configuring Copy activity/Dataflow in microsoft Fabric, changing the read method from Full copy to incremental copy returns. "incremental copy is not supported for the selected source store."   This limits implementation of efficient medallion architecture patterns and increases capacity consumption because full refreshes are required.   Please add native incremental support for dataverse, including support for: -> Change tracking/ CDC -> ModifiedDate or watermark columns -> Insert/Update/delete detection -> Delta processing for lower CU consumption  -> Pipeline integration   Business impact : Without incremental capability, large datasets require repeated full refreshes, increasing processing time and Fabric CU consumption. ... View more

When saving a PBIX file, the Windows "Date modified" timestamp does not update even though the file was saved. Excel updates the timestamp correctly. This makes it difficult to track whether the report has been updated. ... View more

Would have been nice to have a functionality for 'exclude' in a Tree map visual (and other visuals)   It would have been nice to slice and dice the figures by not only explicitly including what to see, but to exclude e.g. the largest customer / biggest product etc. ... View more

The current Groups dialog window is fixed in size, which makes it difficult to work efficiently when grouping fields with long values (e.g., bilingual labels such as English/French). Users must constantly scroll horizontally between “Ungrouped values” and “Groups and members” to add and rename groups, which significantly slows down the workflow. Allowing the dialog window to be resized or expanded would greatly improve usability and productivity, particularly for enterprise and bilingual reporting scenarios. ... View more

Current behavior: When editing a Direct Lake semantic model in Microsoft Fabric, users can select tables at the schema level (e.g., gold). However, this selection is a point-in-time snapshot, any new tables added to that schema afterward are not automatically included in the semantic model. Users must return to "Edit semantic model" and manually tick each new table every time one is added. Expected / requested behavior: When a schema is selected in the semantic model table picker, Fabric should treat that selection as schema-scoped, meaning all current and future tables within that schema are automatically included upon the next refresh or model sync, without requiring manual intervention. Business impact: In environments with active development where new tables are regularly added to a Gold/serving layer schema, this creates unnecessary manual overhead and a risk of reports missing data from newly landed tables. It breaks the principle of a self-maintaining semantic layer. Proposed solution options: Add a "Include all future tables from this schema" toggle when selecting at the schema level. Treat a schema-level checkbox as a dynamic binding, auto-syncing new tables on model refresh. Expose a setting at the workspace or semantic model level to opt into schema-aware auto-discovery. Workaround today: None via UI. Partial workaround via XMLA/TMSL scripting, which requires technical expertise and manual triggering.   Fetch new tables in direct lake  @ahmedshalabyy12  ... View more

In Power BI visuals that support the zoom slider (e.g., line and stacked column charts), there is currently no option to control the default position of the zoom window on the X-axis. I would like to suggest adding a setting that allows authors to define the initial zoom position—specifically the ability to default the view to the most recent data (right side of the axis), while preserving the natural left-to-right chronological order. This would significantly improve usability for time-series analysis, where users are typically most interested in recent events, without forcing a reversed axis that conflicts with standard reading direction and data interpretation practices. ... View more

Current behaviour When using the Invoke Pipeline activity with Wait on completion enabled, the duration reported for the Invoke Pipeline activity in the parent pipeline's monitoring is significantly longer than the child pipeline's own run duration (as shown in the child pipeline's run history). The gap is invisible orchestration overhead — queueing, connection/token acquisition (the Preview activity relies on a linked service/connection), metadata persistence, status polling, and checkpoint writes between parent and child. In nested patterns this overhead compounds with each level. Why this matters Misleading monitoring: the parent activity looks slow when the child actually finished quickly, making it hard to attribute time and diagnose real performance bottlenecks. Penalises recommended patterns: modular, reusable, parameterised child pipelines (a pattern Microsoft itself promotes) are the worst affected, because every hop adds overhead. Cost transparency: since Fabric CU consumption is duration-based, it is unclear how much of this orchestration wait time is billed versus actual child execution. Customers can't currently see or reason about this. Requested improvement In the activity output and monitoring view, break the Invoke Pipeline activity duration into its components (e.g. queued / orchestration overhead / child execution time) so the actual child run time is clearly distinguishable from orchestration wait. Ideally, also reduce the orchestration overhead of the Preview activity (particularly connection/token acquisition) and confirm that CU billing reflects actual execution rather than idle orchestration time. Repro Create a child pipeline with a single short activity (e.g. a 5-second Wait). Create a parent pipeline with an Invoke Pipeline (Preview) activity calling it, Wait on completion = true. Run the parent and compare the Invoke Pipeline activity duration against the child pipeline's own reported run duration. Observe the parent activity reports a substantially longer duration than the child's actual run. ... View more

The item soft-delete and recovery feature, currently in Preview, support several item types, as shown at https://learn.microsoft.com/en-us/fabric/admin/retention-recovery#supported-item-types. However, semantic models and reports are probably the most heavily used item types across Fabric estates. Yet they are notably absent from the list of supported item types.    The Fabric roadmap https://roadmap.fabric.microsoft.com/?product=administration%2Cgovernanceandsecurity#plan-d838b85d-e1b2-f011-bbd2-0022480a2ecf does not provide any information regarding if or when those item types will be added. ... View more

When we build validator/conditional columns, we want to test all possible conditions to ensure that the code for said column works properly. Sometimes the live data does not have all the conditions we want to test. I propose (if it doesn't already exist) a data test mode that users can temporarily enable for directly changing data at the cell level. So if you need to test for blanks and your data has no blanks, you can enable test mode (name to be workshopped later) and convert one of your values to a blank. Once the test is done, you disable test mode, and your data goes back to normal. If possible, this could be used in both Power Query and in Table View. ... View more

Currently Power BI doesn't have the option to move totals from the bottom of a matrix or from the right of the matrix. Would be great to be able to move totals to the top of a report so users don't have to scroll all the way to the bottom or right to see it. ... View more

Right now there's no way to run a custom containerized job on Fabric. The execution has to happen on a remote endpoint you stand up and maintain yourself (Azure Function, ACI, etc.) that Fabric just calls out to. Notebooks and User Data Functions cover Python, but not "bring your own image." Would love a "container job" item, point it at an image in ACR, give it a command + env vars + resource limits, run it on Fabric-managed compute, trigger it on-demand or on a schedule, and have the logs/run history show up in the monitoring hub like any other Fabric job. Today I run things like dbt and custom data ingestion jobs in Azure Container Apps Jobs, orchestrated from ADF. I'd happily consolidate all of it into Fabric if I could run the containers natively instead of keeping a separate compute plane around just for the "run my image" cases that notebooks/Spark don't cover. Anyone else running into this? Is there a roadmap item I've missed? ... View more

Currently, there is no tenant-level control available in Power BI to completely disable the “Explore” (Explorer) feature across the environment. The feature’s availability is governed only through dataset-level permissions rather than a centralized administrative setting. Provide Tenant-Level Setting to Disable “Explore” (Explorer) Feature Globally in Power BI ... View more

Need an audit trail / diagnostic event that explains why a semantic model’s compatibility level was auto-upgraded and to track or identify which user or action triggered the use of specific features along with timestamp (such as DAX Query View Desktop or Web Modeling Edit) that led to the automatic conversion of the semantic model version and the option to disable the auto upgrade of version. ... View more

We would like to propose an enhancement to improve efficiency when working with SharePoint data sources. Currently, when connecting to a SharePoint site, all available files are loaded, and users must then apply filters to identify and work with the required files. This process can be time-consuming and resource-intensive, especially in environments with a large number of files. We are requesting the ability to directly select or specify the required files from the SharePoint site at the time of connection, without the need to load and filter the entire dataset. Benefits of this enhancement: Improved performance by reducing unnecessary data loading Enhanced user experience with more streamlined file selection Reduced resource consumption for large-scale SharePoint environments This capability would significantly improve usability and efficiency for enterprise scenarios leveraging SharePoint with Fabric/Power BI. ... View more

When using shortcuts that reference Materialized Lake Views, if I need to drop and recreate the MLV (to add a field for example), then the SQL Analytics endpoint metadata gets out of sync and files can then be missing.  I created a support case (2605260030001505 for any MS Internal folks looking at this idea) as there appears to be no way to force the metadata to resync other than to drop and recreate the shortcut. I worried about this due to the issues we have found with the same in Semantic Model extracts where the underlying object seemingly gets dropped and recreated and even models that were on extract then have corrupted objects which cannot be found.   Any easy way to rebuild the metadata would be great and when a semantic model is on extract, the files should not be deleted referencing an object that may be dropped to ensure extracts continue to work until the next build. ... View more