Power BI Weekly

Issue 374

11th September 2026

Welcome to the 374th edition of Power BI Weekly! Just the one official post from the team this week, which follows up on the Q&A deprecation we covered back in December: the Power BI Q&A retirement has been pushed back to February 2027, after which any report still containing a Q&A visual will show an error in its place (so it's worth taking an inventory of where you use it now, particularly if that's in embedded PaaS solutions or a sovereign cloud, where there's currently no Copilot alternative to move to).

Over in the community, Alberto Ferrari has written a framework for writing DAX at the correct granularity, setting out the four grains to identify before you write a measure so you can avoid the usual "incorrect totals" issue. Mim is also back with a follow-up to last week's post, this time on writing Databricks Parquet that VertiPaq likes, testing some lesser-known Spark settings to see how well Direct Lake handles files written by Databricks (a nice companion to the Azure Databricks storage mode article I highlighted last week).

Finally, Kerry Kolosko has published A matter of scatter, reminding us that the core scatter visual samples your data down to 3,500 points by default (which can hide its density, or even whole regions of it), along with showing some alternatives like Deneb, Fabric notebooks and Datashader for when you need the full picture.

Data Prep

  • Writing Databricks Parquet That VertiPaq Likes
    A follow-up to Microsoft's Azure Databricks storage mode white paper: Mim tests little-known Databricks Spark settings for row-group size and dictionary encoding, plus a date clustering key, to see how quickly Direct Lake transcodes the resulting Parquet at TPC-DS SF100 and SF1000. With clustering, the Databricks layout matches the paper's V-Order layout at SF100 but still trails delta-rs and the paper's layout at SF1000. The takeaway: VertiPaq cares about physical layout, not which engine wrote the files. Personal opinion, with the configuration and a Claude Code skill in a GitHub repo.
  • Dataflow Gen1 to Gen2 Migration: 3 Easy Methods (Power BI / Fabric Tutorial)
    Learn three straightforward methods-Save As, Upgrade, and using Power Query templates-to migrate your Dataflows from Gen 1 to Gen 2 in Power BI, with step-by-step guidance for each approach.
  • The Warning Icon Every Power BI Pro Dataflow User Needs to See....
    Ned looks at the warning Power BI Pro users will see on Dataflows Gen1 and explains why he thinks Dataflows Gen2 are the better option.
  • Power Query desde cero: guía para principiantes en Power BI [Spanish]
    A Spanish-language beginner's guide to Power Query in Power BI, walking through the interface, applied steps and M, column profiling, handling nulls, replacing values, column from examples, and filtering in Power Query vs in Power BI, using a downloadable Excel dataset.

Data Modeling

  • Writing DAX at the correct granularity
    Alberto introduces four grains to identify before writing a measure (source table, business rule, iteration/evaluation, and requested output) and the principle of evaluating a business rule at the lowest grain where all its inputs are unambiguous, aggregating only afterwards. A lemonade production model shows why averaging rates breaks totals, the same thinking is applied to budgets, and the article ends with a checklist for grain-aware DAX.
  • Everything about Date - Calendar Table in Power BI - Three methods explained with Demos
    This Power BI video explains why accurate Year-to-Date and fiscal calculations depend on a correctly set up date table, demonstrating three methods-including the default auto-generated table, custom tables with "Mark as Date Table," and the most flexible "Calendar Options"-and showing live demos of each approach in action.
  • You Can Create Calculated DAX Columns in Direct Lake — But Should You?
    Amit looks at DAX calculated columns in Direct Lake models: they're now possible and convenient to add, but he questions whether you should use them.
  • Building and Enriching a Microsoft Fabric Data Agent on a Power BI Semantic Model
    How to put a Fabric Data Agent on top of a Power BI semantic model and make its answers trustworthy: prepare the model with Prep for AI (clear names, descriptions, synonyms, AI data schema, verified answers, RLS/OLS), keep calculation rules in the semantic model instructions (Data Agent instructions aren't used for DAX generation), then benchmark and iterate using the Data Agent SDK.
  • v3.6.1 Release
    DAX Studio 3.6.1 is a patch release fixing issues found after 3.6.0, including ribbon theming and spacing, OneLake path handling and authentication token scopes, Server Timings labels, stale names in ERD diagrams, query trace completion and query plan capture, plus improved Excel add-in launch.

Report Authoring and Interactivity

  • Power BI Q&A retirement reminder: February 2027 timeline update
    Microsoft has pushed the Power BI Q&A retirement back from December 2026 to February 2027. From then Q&A stops working, and existing Q&A visuals will show an error. The post maps each Q&A experience to its Copilot alternative, notes that Copilot (and Fabric Copilot Capacity) is now available from F2 capacities, and confirms there are no plans for a Copilot replacement for embedded (PaaS) Q&A, and no Copilot in sovereign clouds yet.
  • My button & slicer disappeared in Power BI Service. Here's why.
    In the Power BI Service, selecting a row in a table brings that table to the front, hiding a slicer and button layered on top of it. It's documented behaviour rather than a bug, and the fix is to switch on 'Maintain layer order' (Properties > Advanced options) for the visual that jumps forward.
  • I Built an Entire Power BI Dashboard without Power BI
    Chandeep builds a complete Power BI dashboard with Claude Code using the PBIP format, from Power Query and a date table to relationships, DAX measures and report pages, testing and fixing what Claude generates along the way. He finishes on the skills that matter more with AI: auditing output, data modelling, and understanding the business.
  • Fabric Apps Tutorial: Build a Dashboard with Write-Back Using Claude Code
    Valerie uses Claude Code to build a Fabric app on top of a Power BI semantic model (her dive log), including write-back so new records can be logged directly in the app. She covers the prerequisites, the prompt that gets a usable first version, where the written-back data lands, and iterating on feedback.
  • Power BI Button Slicer
    Part 3 of a slicer series, covering the button slicer: how it differs from list slicers, tile, image and chip styles, adding a Select All option, formatting and selected-state styling, single vs multi-select, and cross-filtering, shown across eight button slicer variations on one dashboard.
  • A matter of scatter
    Power BI core scatterplots default to 3,500 points using high-density sampling, but developers can bypass these limits using Deneb or Datashader for large datasets.

General