POWER BI NEWS NOVEMBER 2025
Edited by Riccardo Dominici
MICROSOFT 365 APPLICATIONS FOR WORK
COURSES - Tell your friends about them, they're very cheap
SEARCH ENGINE AND COURSE CARDS ON
MICROSOFT 365 APPLICATIONS
ORGANIZE YOUR STUDY WITH THIS GANTT CHART
WIKI CARDS EXPLORER WITH AUTOMATIC
SUMMARY ON CLICK
1.
Introduction and overview of the update.
3
2. Events and
Announcements Fabric Data Days and FabCon 2026.
4
Fabric Data Days: Live
training and community action.
4
FabCon 2026: The Power BI
and Fabric Community Conference.
4
3. General News
Deprecation of Visual R and Python.
5
What is changing and in
what context
5
Timing of the deprecation.
6
What users and
organizations need to do.
7
Practical implications
and considerations.
8
4. Copilot and AI.
Artificial Intelligence at the Service of Data Analysis.
9
Standalone Copilot in the
Mobile App Ask Anything, Anywhere
.
9
Standalone Copilot Web
Updates.
10
Improved Copilot Report
12
Improvements to Verified
Answers.
13
Remote Power BI Model
Context Protocol (MCP) Server.
15
Automatic column
expansion in matrices ( Grow to fit )
17
New Card View..
18
Enhanced image
visualization.
20
OneLake Catalog User
Data Functions in Translytic Streams.
21
6. Modeling News.
23
Versioning of the
semantic model
25
TMDL in Visual Studio
Code (GA) Advanced Tabular Model Editor.
26
7. Data Connectivity.
Next-generation Spark and Impala connectors.
27
8. What's New in
Views. Part One.
29
KPI monitoring range
coverage.
30
Decomposition Tree. All Expanding mode.
31
Dynamic Legends in Zebra BI Charts.
Legends that change with the filter.
33
Drill Down Bubble PRO by ZoomCharts.
Interactive multi-level bubble charts.
34
9. What's New in Views
(Part 2)
35
Power BI Theme Generator:
Create custom themes with AI and best practices.
35
Power Gantt Chart by Nova Silva.
Managing task dependencies.
37
10. Conclusions.
40
What's New in Power BI
November 2025
Welcome to this eBook dedicated to the new
features of Microsoft Power BI introduced in the November 2025 update. This
update is packed with features and improvements, ranging from AI integration
into analysis, to new reporting and visualization options, to updates to data
connectivity and modeling. Each chapter of this eBook corresponds to a
slide from the original presentation and delves into the topics covered in
detail, ensuring a complete technical and functional overview. The language is
professional
yet accessible, designed to be useful to both Business Intelligence analysts
and developers as well as less technical business users who want to
understand the impact of these new features.
Before we get into the details, let s briefly
outline the highlights of the November 2025 Update.
In the following chapters, we'll explore each
new feature in detail, providing in-depth explanations, technical/functional
context, practical implications for different types of users
(analysts, developers, business users), concrete use cases, connections
to other Power BI features or related Microsoft tools, as well as any advantages
and limitations to consider. Additionally, where appropriate, we'll include
references to official Microsoft sources for those who want to delve
deeper.
The first chapter introduces the Power BI
November 2025 Update, providing a high-level overview of the major innovations
and changes. This monthly update is particularly significant as it touches
on nearly every aspect of the Power BI platform, from AI capabilities to
visualizations, from core modeling to data connectors.
Why is this update important?
Microsoft updates Power BI on a monthly basis, introducing new features
and deprecating those less aligned with future strategies, in order to keep the
platform cutting-edge. November 2025 is no exception: among the most
significant new features are substantial improvements to Copilot (the
integrated AI assistant), strategic changes such as the deprecation of R
and Python-based visuals in certain embedding scenarios, and numerous reporting
and visualization features requested by the community.
technical perspective, these new developments indicate
a strengthening of Power BI's integration into the Microsoft Fabric
ecosystem and a growing focus on generative AI applied to BI. For
example, the presence of a standalone Copilot in the mobile app and
improvements to Verified Answers signals a desire to make conversational
data analysis more intuitive and pervasive. At the same time, the
deprecation of R/Python visuals in some contexts underscores the commitment to
greater
security and performance in embedded solutions.
functional standpoint,
business users will notice benefits such as more interactive and engaging
reports (thanks to new visuals, hero images in cards, etc.) and greater
immediacy in data querying (thanks to improved Copilot). Analysts will
have enhanced tools to create reports faster and with less manual effort (Report
Copilot, auto- resize in matrices), while developers will have access to
advanced features for managing models and integrations (local MCP server, TMDL
extension for Visual Studio Code, updated connectors).
November 2025 update improves the Power BI
experience on multiple fronts: events and training, stability and
security, AI and automation, reporting and modeling features,
data integration and custom visualizations.
References: For an
official summary, the Microsoft Power BI blog from November 2025 lists the key
points: deprecation of R/Python visuals in Embed
for your customers , enhancement of Copilot/AI (Copilot mobile, Verified
Answers ), new reporting, modeling, and connectivity
options, and visualization enhancements. This overview serves as a guide to
exploring each aspect in more detail in the following chapters.
In the second chapter, we delve into the events
and announcements highlighted in the November 2025 update. These relate to
community initiatives and conferences that enrich the Power BI and Microsoft
Fabric ecosystem. In particular, the corresponding slide highlights two key
events: Fabric Data Days and FabCon
2026.
Fabric Data Days is
billed as a two-month learning and community event around Microsoft
Fabric and Power BI. Kicking off on November 4, 2025, it features a series of live
sessions, contests, and networking opportunities focused on data culture.
Users can participate in interactive lessons to improve their skills on
Power BI and Fabric services, try their hand at QuickViz challenges (likely
rapid data visualization contests), and earn discount vouchers for
certifications (exams PL-300 for Power BI and DP-600 for Fabric, as
mentioned).
Practical implications:
For analysts and developers, Fabric Data Days is an excellent
opportunity to update themselves on the latest Power BI features and
improve their analytical skills through hands-on sessions. IT professionals
and managers can benefit from networking with the community, discovering
use cases and experiences from other companies. Even less technical users can
participate in the introductory webinars, gaining greater familiarity with the
platform in a guided context.
Tip: During these
events, it's a good idea to leverage social channels and dedicated forums (such
as the Microsoft Fabric Community and local groups) to ask questions and
share learnings. Active engagement offers added value: speakers (including
members of the Microsoft team ) often provide additional
materials and demos that enrich the training.
FabCon 2026: The Power BI and Fabric Community Conference
FabCon 2026 is another big announcement: it's the
third edition of the FabCon Americas
conference, scheduled for March 16 20, 2026, in Atlanta, GA (USA).
FabCon is described as the ultimate community-led event
for Power BI, Microsoft Fabric SQL, Real-Time Intelligence, AI, and Databases.
In essence, it's a community-organized but Microsoft-supported event,
packed with technical sessions, keynotes, workshops, and entertainment:
Sessions
and Keynotes: Presentations by Microsoft experts and
community speakers on advanced topics in BI, AI applied to data, real-time
analytics, etc. (e.g., there may be a keynote on the future directions of Power
BI).
Expo Hall
and Partner Day: An exhibition area featuring
Microsoft partners (solution providers, integrators) and a dedicated partner
day where you can connect with vendors and discover complementary products.
Community
lounge & networking: Informal spaces to meet other
attendees, exchange ideas, and perhaps ask experts 1:1 questions ( Ask the
Experts sessions mentioned).
Power Hour
& Data Viz Championship: Special activities such
as Power Hour (a live session showcasing creative solutions in one hour)
and a global Data Visualization competition great opportunities to see
visualization best practices in action.
Final
party: last but not least, a moment of fun and
networking, in this case at the famous Georgia Aquarium, to seal the end of the
event in a convivial atmosphere.
Practical implications: FabCon is a must-attend for Power BI
superusers and developers who want to delve deeper: it offers advanced
training and visibility into preview features or innovative ways to use the
platform. Business analysts and BI team leaders can also gain valuable
insights into adoption strategies and industry trends. Last
but not least, attending (or virtually, if online sessions are
available) demonstrates the company's commitment to data culture,
allowing them to then bring the acquired knowledge back internally.
Tip: Registration
for FabCon 2026 is now open (with the discount code FABCOMM,
which offers a $200 discount on the registration fee, as advertised). If you
can't make it to Atlanta in person, check to see if any recorded or
livestreamed sessions will be available the community often shares key content
post-event.
In summary, this chapter highlights how
Microsoft offers not only software tools, but also training events and
community conferences to support the growth of Power BI users and
practitioners. Investing time in these initiatives can accelerate the effective
adoption of new features introduced later.
Let's now delve into a major change
announced in the November 2025 Update: the deprecation of support for visuals
created with R or Python in reports embedded for external users through
the " Embed for your customers" solution. This chapter
explains what this is, why Microsoft made this decision, and what the practical
implications are for those who use these types of visualizations.
Embed for your customers
(also known as the "app owns data" scenario) is a Power
BI embedding mode in which a custom application displays Power BI
reports to its end users without them having a Power BI license of their
own. In other words, the app handles authentication and "owns" the
data, providing users with an integrated experience. A typical example: a
company creates a portal for its customers that embeds interactive reports
created with Power BI; customers navigate the report without directly logging
in to Power BI.
Microsoft has announced that starting in May
2026, it will end support for embedding reports and dashboards that contain
visuals built in R or Python in two specific scenarios:
Embed for your customers (app owns data), above;
Publish to
web, the other public embedding method (when
generating a non-authenticated public link to the report).
After the deprecation date (May 1, 2026),
reports embedded in these ways will still load, but any visuals based on R
or Python code will appear blank. Important: This change does not
impact embedding scenarios. internal:
Embed for
your organization ( user owns data ), i.e. embedding reports where the end user logs in with their own
Power BI credentials (for example, embedding a report in internal SharePoint),
will
continue to support R/Python visuals.
Secure embeds
in SharePoint, portals, or similar internal sites are not affected by
this deprecation.
In short, the restriction only affects
unauthenticated or external app-centric scenarios. The following table
summarizes the change:
|
Embedding
Scenario |
Description
|
Visual
R/Python support post-May 2026 |
|
Embed for your customers (App owns data)
|
Custom app reporting for external users (no license)
|
Not supported
(R/ Py visuals shown blank)
|
|
Publish to web ( public link )
|
Report published publicly |
Not supported
(R/ Py visuals shown blank)
|
|
Embed for your organization (User owns data)
|
Embed report with authenticated user (internal)
|
No impact (visual R/ Py still works)
|
|
Secure
embed (SharePoint, Portal) |
Embed in secure authenticated context
|
No impact (visual R/ Py still works)
|
Why the deprecation?
According to Microsoft, the decision is part of its ongoing commitment
to providing a secure, scalable, and robust analytics platform. R and
Python-based visuals, while powerful, involve running external code (R/Python)
within reports. In open contexts such as public embeds or for external clients,
this can pose security risks and require complex supporting
infrastructure. Furthermore, the company refine the feature set based on market
needs and performance: the use of such visuals in external scenarios is likely
limited by maintenance costs. By focusing on alternative solutions (for
example, custom DAX-based visuals or the use of other integrations such as Fabric
Notebooks
),
Microsoft can ensure better performance and a smaller surface area for
potential vulnerabilities.
Microsoft has provided a clear timeline for
this transition, which we summarize:
This timeline gives organizations approximately
6 months from the announcement to
take action
on their reports.
The announcement also includes practical
recommendations on how to address deprecation:
Mapping
and Assessment: Administrators or BI teams should identify
all reports and embedded dashboards targeting external or public customers
that
use R or Python visualizations. A comprehensive inventory helps understand
the extent of the impact.
Plan the
migration: For each identified report, you need to find
alternative solutions for R/Python visualizations. This may mean:
o
Replace the visual with
a Power BI standard (e.g. a native chart or a certified custom visual available
on AppSource ) that offers similar functionality
without running R/ Py scripts.
o
Py logic to DAX or Power Query
if possible, then move the calculation into the Power BI data model.
o
Using Notebooks in Microsoft Fabric: If R/Python visualization was used for advanced analysis that couldn't
be easily replicated in DAX (for example, machine learning or specific
statistical graphs), Microsoft suggests considering
using Fabric Notebooks. In Fabric (the platform that integrates various
services including Power BI), Notebooks allow you to run Python/R code in a
controlled context and then optionally integrate the results into a report.
This strategy maintains advanced analytical capabilities but moves execution
outside of the report embed, into a dedicated environment.
Update
documentation and inform stakeholders: The application
development teams performing the embedding, as well as
customers/end users if necessary, should be notified. For example, if
a certain public dashboard will no longer display a specific analysis, this
should be communicated and perhaps replaced with a link to an internal report
or alternative document.
Testing
and Quality Assurance: After updating your reports, it
is essential to test them in the relevant embed scenarios to ensure that all
visuals function correctly and that the user experience is not degraded.
For Power BI developers and BI
managers with embedded solutions in production, this deprecation is a
reminder of the need to actively monitor cloud product roadmaps. Power
BI Embedded Analytics is an important subset, and support changes like this
are likely to occur as the service evolves. Fortunately, Microsoft has provided
adequate advance notice.
For end-user business users using embedded
reports, the direct impact should be minimal if report owners take timely
action. In some cases, they may notice a change in visual type (e.g., a
different chart replacing the previous R-based one), but ideally, the
information value will remain the same.
A long-term benefit of this change is
potentially improved stability and security for embedded reports: by
removing the need to run R/Python code on Power BI servers for public
scenarios, potential security vulnerabilities and computational resources are
reduced. Additionally, this change could incentivize Microsoft to further
enhance Power BI's native visuals to address any gaps, or improve
integration with solutions like Python Scripts in Power Query or Notebooks.
Limitations to consider:
Obviously, the downside is the loss of flexibility. Some very specific
visualizations obtained via R/Python libraries (for example, advanced
statistical graphs like genetic heatmaps or custom maps) may not have an
immediate counterpart in Power BI. In such cases, one consideration is whether
Embed
for your customers is still the right scenario: for example, an app could
load static images generated elsewhere, or the raw data could be exposed via
API and processed by the client. These are more complex solutions, so it's best
to evaluate the importance of the visual in question on a case-by-case basis.
In conclusion, the deprecation of R/Python
visuals in external embed contexts is a strategic change aimed at
strengthening the platform in terms of security and performance.
Organizations should take this opportunity to renegotiate their advanced
reporting practices, making the most of standard visuals or the new
integrated analytics tools offered by Microsoft Fabric. As additional
resources, Microsoft encourages consulting the official documentation on Embed
for your customers and secure embeddings, as well as reaching out to
the community forums or your Microsoft representative for support during the
transition.
One of the cornerstones of the November 2025
Update is the extensive set of improvements and new features in the Copilot
and AI area. Microsoft is investing in making artificial intelligence a
key element of the Power BI experience, both to facilitate content creation
(reports, dashboards) and to enhance data consumption with natural
interactions. In this section, we'll explore:
Introducing standalone
Copilot to the mobile app (new preview).
Updates to
the Copilot chat with your data experience in the web/desktop version.
Improvements
to Report Copilot (AI that helps you create and edit reports).
Feature
Extensions for Verified Answers (verified answers).
Announcing a remote
MCP (Model Context) server Protocol ) for
chat with your data.
Each of these elements contributes to a common
goal: making interacting with data in Power BI more natural, immediate, and
intelligent.
The first new feature is the arrival of a dedicated
Copilot in the Power BI mobile app (for both smartphones and tablets).
Essentially, the mobile app (for both iOS and Android) will have a button on
the homepage that opens a chat interface with Copilot, allowing users to
ask questions in natural language about their data directly from their phone.
Key Features:
Integrated
in-app chat: Users can ask text-based questions (by
typing them) or even dictate them by voice (voice dictation is supported
on iOS, so iPhone/iPad, as specified). For example, a traveling executive could
ask, "Show me total sales this quarter by product," simply by
speaking into the phone.
Responses
with graphs and text: Copilot doesn't just respond
with text, but when possible, it directly generates
a relevant visualization. For example, if you ask for a KPI, Copilot might
automatically create a small bar graph or a card with the number. The user can
tap the generated graph to expand it, analyze it further, or interact with it
(for example, to see details).
Citations
and further information: An important aspect is that
Copilot provides citations or references to the sources of the data used
for its response. This means that, for example, if Copilot returns the number
of sales, it will display a reference to the report or dataset from which it
drew the data. By tapping the citation, the user can open the original
report for further information or to check the context. This feature
increases confidence in the AI's responses, as it allows the user to
trace the source data (which is crucial for making decisions based on those
responses).
Instant
insight sharing: If a user discovers something
interesting through Copilot (say, a graph showing a decline in sales in a
region), they can share it right away: the app lets you share the generated
visual (presumably via screenshot or as a copy of the graph in the message)
or copy the text of the response, making it easy to send via email,
chat, or insert into presentations on the fly.
Availability:
This feature is listed as Preview and will be available
in the weeks following November 2025 via an app update on the stores.
Therefore, it requires no special configuration other than updating the app
when the update is released.
Context of use: Imagine
a business user say, a sales manager who's out of the office.
Traditionally, they'd have to open their laptop, connect to Power BI Service,
and search for the right report to find a certain number. Now, they can simply
open the app on their phone and ask Copilot directly. Within seconds,
they'll get the answer, perhaps with a small chart. This makes Power BI much
more accessible on the go. For an analyst or data scientist,
the feature is convenient for quickly cross-checking data or gathering insights
without having to manually build queries on the fly.
Technical considerations:
It is mentioned that using Copilot requires that the data be on Fabric
capabilities (i.e., in an AI-enabled environment). Additionally, semantic
models must be AI-ready with techniques such as enabling Verified
Answers and providing context (e.g., table descriptions, synonyms, and
well-curated Q&As). If the model is not AI-ready, the app should alert the
user with clear messages. This means that to take full advantage of Copilot
Mobile, the tenant or organization must have enabled AI previews and
prepared the datasets according to the guidelines (an IntendedUse
field in the dataset, Q&As enabled, etc.).
Current limitations:
Since it's a preview, some features may still be missing. For example,
dictation is iOS-only at launch. Furthermore, it's likely that Copilot Mobile
will primarily support reading already modeled and published data,
and not editing or creating content (you can't create a report
from your mobile phone with Copilot; that's more of a function of the Copilot
Reports desktop/service side). It's a consumption and query tool, not an
authoring tool.
Alongside the mobile launch, Microsoft
announced improvements to the Copilot chat with your data experience in
the web version (the Copilot launched in Power BI Service and Desktop a few
months earlier). These improvements aim to make using Copilot simpler and
more intuitive. Some noteworthy new features include:
Automatic
data source selection: Previously, when asking Copilot
a question online, if there were multiple possible relevant datasets or
reports, the system would ask the user to choose one from a list before
proceeding. This step has now been streamlined: Copilot automatically tries
to identify the best data source and selects it in the background. Only if
the question is too general or ambiguous will Copilot ask for clarification
(for example, if you ask "Show me sales," and you have separate
datasets for different businesses, it might ask "Do you want Sales Europe
or Sales USA data?"). This automation makes the experience more seamless:
in many cases, you get the answer immediately without additional clicks,
at the cost of waiting an extra second or two for processing (the extra time
allows Copilot to reason about the source choice, but it's an acceptable
trade-off to avoid interruptions).
Streamlined
search results: When you ask a query like "Find
2023 sales reports," Copilot could return a list of relevant items
(reports, dashboards, workspaces). This list is now presented in a more
streamlined, text-based format, making it easier to read, rather than
cluttered graphical tabs. This helps users quickly locate what they're looking
for.
Direct
access from the Home page: By the end of November,
Copilot will be accessible directly from the Power BI Service home page with
immediately available question input. This means that as soon as you log in to
the Power BI portal, you'll find a prominent "Ask your data a
question..." box. If you prefer the classic home interface, you can always
switch back (there appears to be a toggle for the home page style).
Smart
suggested prompts: Copilot will start showing question
suggestions based on items you've recently opened. For example, if you
recently opened a sales report, Copilot might display a prompt like "Ask:
What was the best product in the Q3 sales report?" This feature guides the
user through the data without even having to fully formulate the question,
reducing the blank page problem.
Improved
item attachment to questions: If you want to specify a
specific report or dashboard as the context for your question (rather than
relying on automatic selection), the item attachment interface is now more
convenient. Items (reports, datasets, etc.) are sorted by most recent and
filterable by type, with a design consistent with other parts of Copilot and
Microsoft 365. This makes it easier to manually set the context when needed.
Practical implications:
For an analyst using Copilot in Power BI Service, these
improvements mean less friction. They can query data more quickly, even
for simple tasks like finding the location of a piece of information, thanks to
the built-in search function. For a new user or a business user, the
presence of question suggestions is an invitation to use Copilot as a starting
point for exploration: it makes the experience more guided. A concrete example:
Mario, a manager, opens Power BI Service to see sales trends. Instead of
navigating through dozens of reports, he notices the Copilot box with a
suggestion, Want to know the trend by region over the last month? He clicks
and receives the answer with a graph and a link to the main report, which he
can open only if he needs to dig deeper. He has saved time and gained immediate
insight.
technical/administrative
standpoint, enabling these features requires the Power BI administrator
to activate Copilot (which is still in preview in 2025 and must be
explicitly enabled at the tenant level). Furthermore, search and suggestions
require the service to properly index content. Security-conscious companies can
rest easy because Copilot respects permissions: a user can ask "find sales
reports," but will only see those to which they have access.
Report Copilot is
the feature that helps automatically create or modify reports using AI,
based on user requests. Introduced a few months earlier in preview, it will be
further enhanced with the November 2025 update, making it more capable of
understanding user intent and producing visuals that meet expectations.
Three specific improvements are mentioned:
Smarter
visual recommendations: Copilot now automatically
chooses the most appropriate visual type for the data the user wants to
display. For example, if I ask "Create a chart of monthly sales by product
category," Copilot will decide whether to use a histogram, a timeline, a
clustered column chart, etc., based on best practices. Previously, it could use
more generic choices; now it's smarter in presenting the information in the
most effective way.
Expanded
visual library: Copilot supports more visual types
than
before. This means it can also use additional visualizations (perhaps it now
includes advanced maps, gauges, etc., where previously it was limited to
bars/lines/pies). This gives the AI more options to compose reports that
best meet needs; for example, if the data is geographical, it could directly
propose a map, which it couldn't do before.
Better
context understanding: Copilot can interpret more
complex commands and nuanced requests. You can describe in more detail
what you'd like to see, and it can grasp the details. For example, "Create
a report with total sales, average margin, and a chart showing monthly trends,
filtered by the last year" Copilot should now understand all the
components (metrics, visual type, time frame) and generate a report page with
these characteristics. In the past, it might have been confused or ignored some
of the instructions.
The claimed result is that Copilot can
generate entire pages of reports in seconds with more accurate and useful
results than the initial version. And it's available on both the Power BI
service (in the browser) and Power BI Desktop, so both traditional
report authors and those working on the web service can use it.
Practical example: A
data analyst needs to quickly prepare an initial report for Q4 sales. Instead
of starting from scratch, he opens Power BI Desktop, activates Copilot, and
writes: "Create a page with: a card with Q4
sales totals, a bar chart of sales by region, and a table with the top 10
products by revenue." In a few seconds, Copilot inserts these
elements, already configured in the open dataset. The analyst can then refine
the details (format, additional filters), but has saved substantial time in the
basic creation. This allows for faster iteration with business
colleagues: you can generate a draft report, show it, and then adjust or deepen
it based on feedback.
functional standpoint,
Report Copilot further lowers the barrier to creating complex reports. A business
user with no DAX knowledge might be able to generate meaningful analyses
using text descriptions alone. However, it's still important to understand what
to ask for: AI doesn't replace business logic. Furthermore, for Copilot to be
effective, datasets must be well-modeled (relationship coverage, pre-built
measures, etc.), otherwise the AI will have little to use.
Considerations: Report
Copilot is currently evolving; it's always advisable to carefully review the
generated reports. AI is powerful, but it can lack
the critical sense to understand whether a graph is relevant or showing
misleading correlations. This leaves the analyst in the loop as a
supervisor. Over time, however, these improvements could lead to a future where
drafting a report is a matter of description, and the analyst spends most of
their time interpreting and narrating, rather than manually constructing
graphs.
The Verified Answers (verified answers)
is a feature related to Copilot/Q&A in Power BI that allows you to pre
-set validated answers to specific data questions. Essentially, a report
creator can "certify" that certain questions correspond to specific
visuals and numbers, so that when a user asks that question, they get a
reliable and validated result. The November update introduces several
improvements to this feature, making it more flexible and powerful:
Full
visual state inheritance: Previously, saving a verified
answer would capture any filters currently applied to that visual, but not
necessarily any other state such as slicers or broader contexts. Now, a
verified answer retains the entire visual state, including filters
applied via slicers, cross-sections, drill- throughs, etc. For example, if I
have a page with year and region slicers and generate a verified answer to What
is the turnover in Europe in 2024? with year=2024 and region=Europe
selected, then that answer will retain those selections. In the future, when
the user asks Turnover in Europe 2024, Copilot will return the appropriate
visual already contextualized to Europe and 2024, without the need to
reset filters.
Support
for additional visual states: The additional states supported
in the verified answers are explicitly mentioned:
a)
Slicer selections (already mentioned),
b)
Field
parameter choices ( i.e. if the user has selected for
example Views for [Month/Quarter] with a parameter, it is now taken into
account),
c)
Filters and
cross-highlighting (when you click on an element in a chart that filters
another),
d)
Drill -
through filters (when you drill to a detail page).
This means that almost any condition that
influences a visual can be part of a verified response. The result is that AI better
understands what to show with the right context when recalling that
response.
More
configurable filter combinations (3-10): When
configuring a verified response, the report creator can define a few filter
variations for which that response is valid. For example, they might say: This
visual is the answer if the user asks for sales filtered for any combination
of (Year=2022 or 2023) and (Region=Europe or USA). Previously, there was a
limit of 3 predefined filter permutations; now, this goes up to 10
permutations. This allows you to cover more complex use cases, such as
multiple years x multiple regions x multiple products in various combinations,
if you know those questions are common. The observation was that most
real-world questions have 1-3 filters, but some have more, so this update aims
to cover those needs as well.
Improved
filter reliability: Behind-the-scenes improvements
have been made to the mechanism by which Copilot matches the user's question
to the correct verified answer and applies filters. This has essentially
reduced the number of cases where the AI misinterprets intent or forgets a
filter. This ensures that if an appropriate verified answer exists, it is used
correctly and with the correct filters applied, reducing discrepancies between
the question asked and the displayed answer.
Support
for new visual types (card and Azure Maps): With the
introduction of the new Card visual (which we will discuss in the next
chapter), it is now possible to use this in the Verified mechanism as well.
Answers. The same goes for Azure Maps. Previously, some visualizations weren't
supported in verified answers; now, the list is expanded, allowing for richer
answers (for example, you might want the question "show sales by
city" to return a map rather than a table, and this is now possible if you
configure it as a verified answer).
Why is this important?
These improvements ensure that verified answers are truly integrated
into the user experience, faithfully reflecting the state the user expects.
In the past, there were complaints: When I ask a question via Q&A, the
number I get doesn't take into account the filter I had in the report. Now,
however, thanks to state inheritance, the answer appears consistent with what
the user is seeing and with their interactions. This increases trust in using
Q&A/AI: the user sees that the AI doesn't give random numbers, but respects the filters and
intent.
Use case: A report
created by an analyst might have a built-in set of frequently asked questions.
Example: In the Customer Satisfaction 2025 report, the analyst sets up
verified answers for questions like What is the average CSAT score for
[Product] in [Region]? with corresponding visuals. A manager opens the report,
selects product X from the slicer, and then asks Copilot (or in the Q&A
box) What is the average CSAT? Copilot recognizes that there is a verified
answer for this question (covering the context of the product, selected region,
etc.) and directly displays the default visual. The manager gets the answer
immediately and knows it has been verified by the analyst, so they
consider it trustworthy and contextualized.
For report authors,
Verified management
Answers is now more laborious (you have to consider
more states), but also more powerful. It almost becomes a conversational
experience design job: you have to predict what
questions users might ask and ensure you offer prompt and appropriate answers.
It's a new field of BI design, straddling data analysis and conversational UX.
The latest news in the AI space is a feature
that is very much geared towards developers and integrators: the Remote
Power BI Model Context Protocol (MCP) Server. This is a server component
(coming soon, in preview) that allows you to enable the chat with your data
functionality outside of Power BI, via agents or custom applications,
all in a secure and controlled way.
Simply put, Microsoft is enabling developers to
build their own custom Copilots that interact with Power BI models. To
do this, the MCP server offers three key tools to external AI agents:
Get
Semantic Model Schema: An API that allows the agent to
retrieve the structure of the Power BI semantic model (dataset). This allows
the agent to understand the tables, columns, measures, and relationships in
short, the vocabulary and structure of the data.
Generate
Query: A service that uses Power BI Copilot logic to
generate a DAX query based on user input and best practices, starting with a
natural language question. So, if the user asks What
is the total sales per product this year? , the agent can call this tool, which
will return an appropriate DAX query (e.g.,
CALCULATE(SUM(Sales[Amount]),
YEAR(Date)=2025) in pseudocode ).
Execute
Query: A mechanism to actually run the DAX query
against the Power BI model and get the results.
By combining these three capabilities, an
external agent can converse with the data: it understands the question,
translates it into DAX, runs DAX on the dataset, and returns an answer (which
can then be further processed and presented as desired).
How and where can it be used?
Microsoft suggests that any application hosting an MCP client can
implement these tools. For example, Visual Studio Code could integrate
it (perhaps for developers testing conversational queries directly in
VSCode ). Furthermore, non-Microsoft clients can
also be supported, using Service Principal authentication over Entra ID (Azure
AD). This means that third-party companies could create custom applications,
bots, and chat services that connect to this MCP server to provide answers
about company data. A concrete example: a company could develop an internal bot
(on Teams, or on its own portal) where an employee asks "show me the
performance of KPI X this month" and the bot uses the Power BI MCP server
to return the value and perhaps an ASCII chart or image.
Security: Everything
is authenticated with Microsoft Access ID (Azure AD), so only legitimate
agents with the appropriate permissions can access data through the MCP server.
This keeps the data protected according to the roles and permissions already in
place in Power BI.
Implications for developers:
With this, Power BI developers/ architects have a new way to
integrate Q&A and Copilot capabilities into custom flows, extending the
analytics experience beyond Power BI itself. Until now, asking a natural
language question to the data required using the Power BI interface (visual
Q&A or Copilot). With the MCP server, you could potentially have Q&A
integrated into a corporate smart speaker, or within Microsoft Teams
through
a bot. This opens up analytics scenarios. conversational omnipresent.
Status and considerations:
At the moment (end of 2025) it is in preview and available "
soon " so it's probably in private testing or will be released soon for
Power BI Premium users. It should be noted that it will almost certainly
require Premium or Fabric capabilities, since we're talking about
queries on models (i.e. datasets that support multiple query concurrency,
etc.). Furthermore, using these tools presupposes that the models are well
described: the agent will generate sensible DAX only if the model is
organized, with intuitive names and perhaps with predefined measures.
Otherwise, it risks giving incorrect or null results.
In conclusion of this substantial chapter, we
can say that Copilot and Power BI's AI capabilities in November 2025
represent a significant leap forward. The aim is to:
Bring AI
everywhere (on your phone, inside and outside Power
BI),
Making the
experience simpler and more integrated (fewer clicks,
more context retained),
Give more
control to power users (customizable verified
responses, developer hookup via MCP).
To take advantage of these innovations,
companies must prepare their data and models (so that Copilot works
well) and train users on these new capabilities. An informed user who
knows they can ask Copilot questions on their phone, or that they can trust
verified answers, will benefit enormously in efficiency. On the other hand, expectations
must also be managed: AI does not replace data
governance; on the contrary, it strongly requires it (clean datasets, certified
measurements, correct permissions, etc. are prerequisites for AI to truly
shine).
Helpful references:
For further information, Microsoft provides documentation on how to
create reports with Copilot and prepare
data for AI and Verified Answers. Developers can read more technical
details about the MCP server in the dedicated documentation.
5. Improved reporting views and
functionality
Chapter 5 focuses on what's new in Reporting,
namely those features that improve the experience of creating and viewing
traditional reports in Power BI. The November 2025 update includes several
notable improvements:
The option to
automatically expand the columns of a matrix to fill the available
space.
The new Card
view is becoming generally available with advanced features like hero
images and collage layouts.
Image view
enhancements with new styles and interactive states.
An integrated
enhancement with Fabric: the OneLake Catalog now supports user feature
selection for Translytical workflows
(in preview).
Let's analyze each of these entries, explaining
the context and practical usefulness.
Power BI tables and matrices are essential
tools for displaying detailed or aggregated data. Often, a matrix (a
pivot table with rows and columns) doesn't completely fill the available
horizontal space in the report: perhaps there are only a few columns with
fairly narrow data, leaving a lot of empty space on the right. In the past, the
report designer had to manually resize the width of the columns or the
visualization to achieve an aesthetically pleasing result.
With the October 2025 update, Microsoft
introduced a feature called Grow to fit for tables, and now in November
2025 it is extending it to matrices as well.
What does
Grow to Fit do? In
short:
When enabled,
the view automatically expands columns to fill all available space. The
algorithm takes the extra space in the view and distributes it evenly across
the existing columns.
This only
happens if the total column width of the content is smaller than the width of
the visual. If, on the other hand, the content is too wide to fit, the default
behavior remains ( horizontal scrollbar if necessary).
It helps to avoid
large empty areas and improve readability, without having to manually
intervene on each column.
How do I enable it?
In the matrix formatting options, under "Column Headers" >
"Options," there's an auto-size toggle and a resizing behavior
setting, which now includes Grow to fit. Enable this option and the
matrix immediately adjusts. The announcement gives a concrete example: if you
have a matrix that's, say, 800px wide but the columns together only cover
500px, enabling Grow to fit will allow them to expand proportionally to
800px, eliminating any empty spaces and that unsightly feeling of a
"small" table inside a large box.
Previous bug fix: In
the initial October implementation, there was a bug noted: an unwanted
horizontal scrollbar would appear even when it wasn't needed. This was fixed in
November, so if Grow to fit is enabled and there's enough space, no
scrollbar will obstruct the view (the scrollbar will only appear if the
content actually exceeds the available space).
Practical implications:
This improvement, while small, is very useful for report designers. Analysts
and report developers can now ensure, with a simple toggle, that their
pivot tables appear clean and utilize space optimally, regardless of the
filtered or inserted data. This reduces the number of cases where a business
user sees a report and thinks something is missing because there's a large gap
on the right that gap will now be filled by the wider columns, improving
aesthetics and readability (slightly larger fonts spread across more space).
One detail: Grow to fit is particularly
useful in scenarios where the columns of a matrix are dynamically generated
(for example, columns per year, if I have few columns in certain filters and
many in others). Previously, the analyst had to choose: either size for the
maximum possible case (with the risk that in minor cases everything will be
compressed to the left), or size for the most common case and accept that when
there are many, scrolling will appear. Now, with Grow to fit, you can
leave management to the system, which adapts scenario by scenario. More
automation, less manual maintenance.
Note: This option
doesn't change the data, just the presentation; its effect is immediately
visible and doesn't affect exports (e.g., if you export to Excel, the
columns will have their base size, not their fictitious growth). This option is
for interactive visualization in Power BI only.
For more details, see the updated documentation
on tables and matrices in Power BI, which includes the Auto-size width and
Grow to fit options.
A Card is a widely used Power BI
visualization type for displaying a single key value (a KPI, a summary number).
Until now, there was the classic "Card" visualization that displayed
a value with perhaps a label. However, Microsoft has had a new, more flexible
and modern version of the Card in preview for some time. With the November 2025
update, this new Card becomes General Availability (stable and
production-ready), bringing with it all the advanced features developed.
The features of the new Card visual are quite
rich:
Hero
Images: The card now supports the addition of a hero
image. In design, a hero image is a visually striking image, typically used
as a focal point. Within a card, a hero image can be a logo, a product photo,
or an icon representing the KPI. It serves to provide immediate visual
context to the number. For example, if the card displays 95% Customer
Satisfaction, you could add a customer satisfaction icon or the company logo to
make the box more eye-catching. It complements the classic callout image (which
was a smaller image associated with the value). There are now two types: the hero
image (prominent) and the callout image (surrounding).
Flexible
image sources: The image (both hero and
callout ) can be uploaded manually, provided via URL,
or even taken from a column in the dataset. This last option is very
powerful: it allows for dynamic cards with images that vary depending on the
data. For example, if each product has an image URL in the dataset, a Card
showing sales for a certain product could automatically display that product's
photo as the hero image.
Control
image positioning and adaptation: You can define how
the image fits on the card (filling, alignment, cropping, etc.) and you have
styles for borders, effects, filters, therefore highly customizable rendering.
Dynamic
collage layout: The new Card introduces a layout
called Collage, where the three elements (main value, reference label,
and hero image ) can be arranged with different
priorities. Essentially, you can decide to highlight one of the three in a
larger space and the other two in smaller spaces within the card. For example,
you could give ample space to the number and image and make the label (e.g.,
Q4 Sales ) smaller. Or, if the focus is on the image (say, you're showing a
photo of a top seller), you can make the image take up the majority and the
value (e.g., revenue) and label are secondary. The collage layout offers
flexibility beyond the classic uniform block layout.
Customizable
component order: You can choose the display order
of
the three components (Value, Reference Label, Image). For example, you can
decide that the image appears on top, the value in the middle, and the label at
the bottom, or the value on top, with the image in the middle semi-transparent,
etc. This means that the report designer has complete control over how the card
appears, always putting what matters most to the audience at the center.
Unified
and modern formatting experience: The new Card has
been designed to have a formatting panel that is consistent with other
modern visuals, with clear categories and consistent naming (for example,
similar to the new button slicer or other visuals introduced
in 2024). This makes options easier to find and also
ensures consistency across visuals: paragraph styles (font, colors,
background ) work the same way on cards as on other
elements, making life easier for formatters. They've also updated the
default card styling, so they look nicer out-of-the-box (modern colors and
fonts).
Compatibility
with the old card: Please note that the old Visual
Card remains available if you prefer it or if it's needed for backwards
compatibility (perhaps there are reports that used non- migratable
customizations ). However, the push is clearly toward using the new one, which
offers much more.
Practical implications:
For report designers, this new card allows you to create much
more attractive and informative KPI dashboards. For example, a management
dashboard can have different cards: YTD revenue with the company logo as the
hero image, customer count with perhaps an icon of a group of people, customer
satisfaction with a smiley emoticon, and so on. Instead of just big numbers and
text, each key indicator can now be visually contextualized. This captures
attention and also helps business users immediately connect the number
to the entity (because a picture is often worth a thousand words in terms of
recognition).
Furthermore, the ability to take images from
the dataset opens up dynamic scenarios: imagine a Card showing Top Performer
of the Month with the photo of the number 1 seller and their turnover: there
is no need to manually update the image, if the dataset allows it, it will
change on its own based on the data.
Accessibility and consistency:
The effort to standardize controls and layout also means that new cards
meet accessibility standards (contrast, image alt text, etc.) and are
consistent with the overall themes. This helps organizations that must follow
brand guidelines: with theme generation (which we'll discuss in the
visualizations chapter), it's easier to ensure a card adheres to the corporate
colors and integrates well with other objects.
Practical use case:
A financial analyst is preparing a quarterly report. He decides
to include a summary page at the beginning with the key KPIs: revenue,
operating margin, costs, etc. Using the new Cards, he manages to include not
only the number but also a small trend graph (as an image? He could upload a
mini chart as a hero, but more typically, no) or, more usefully, the department
logo on each card, to create a pleasing visual effect. The margin, perhaps, is
illustrated with a growing graph icon, and costs with a piggy bank icon. These
touches ensure that when the manager looks at that page, he immediately
associates the symbol with the concept and then reads the number: the
information is absorbed more quickly than if it were just text.
technical side, the
visual being GA means it's considered stable, so it can be used in production
with the assurance that any major bugs have been fixed and that it will remain
forward-compatible. To take advantage of it, simply update
Power BI Desktop to the latest November 2025 release (or higher) and you'll
find the new Card in the visual panel (perhaps replacing the old one as the
default option).
Those interested in learning more can consult
the detailed announcement on the blog and the documentation. Microsoft
obviously encourages you to use the new Card and share feedback. The
documentation provides step-by-step examples on how to add hero images and
customize the layout.
Another improvement in the reporting section
concerns the Image visual. Until now, the image object in Power BI was
pretty basic: you inserted an image (physically or via URL) and displayed it.
There weren't many styling options, nor any interactivity beyond the hyperlink,
if set.
With the update:
Advanced
formatting options are added for the image,
similar to other elements: for example
you can set a background to the image, a border with color and
thickness, shapes (round corners or special shapes), and Effects like
drop shadows. This allows you to better integrate decorative images or logos
into the report design without having to pre -process them externally
(previously, you might have manually added a background to the image itself,
but now you can do it via Power BI).
Even more
interestingly, the image now supports different states: Default,
Hover, Pressed (and a general All state). This
means the image can behave like a button: you can, for example, have it change
its outline color or appear slightly brighter when hovered ( Hover ), and have
another effect when pressed (clicked). This effectively turns the Image visual
into a sort of custom button. In fact, many people were already using
images as buttons (to navigate between pages, to activate bookmarks, etc.), but had to settle for a single static state; now they can
make that experience richer by giving the user visual feedback when interacting
with the image.
Multiple
image sources: As with the Card, it seems the Image
visual now allows you to use URLs or data as the image source (not just
manual uploads). This means the visual image can be linked to a model field:
e.g., I have a table with image URLs, I can link it and have the visual image
dynamically display the image corresponding to the context (filters). It's not
entirely clear whether it's that extensive, but from the text " Expanded
image sources (URL, data column, upload)" it seems so.
Fit
Options: They've expanded the options for how the
image fits into its container: Fit, Fill, Center, Stretch. This lets you
decide whether the image should fill the entire space by cropping out any
excess ( Fill ), maintain the aspect ratio by centering ( Fit /Center), or
stretch to fill the entire space (Stretch). Again, it was very rigid before (I
think there was only fill vs. fit ).
Implications and Uses:
With these new features, a report designer can achieve many
things:
Build eye-catching
navigation menus using images as buttons that highlight feedback when
hovered over (imagine an icon menu for navigating report pages: now on hover
you can make a shadow appear or slightly change the icon to indicate it's
clickable).
Show dynamic
images, for example in a product report, selecting a product from the
filter displays the product photo in the Image visual (which gets the URL from
the filtered data).
Enhance your brand
identity in your reports by adding your company logo in a corner with a
well-defined style (perhaps a rounded edge or shadow), or by adding decorative
background images with transparency. All directly in PBI without external
editing.
For end users, this means more
interactive and polished reports. For example, if they click an image of an
arrow to move to the next page, the arrow now animates on click, giving the
feeling of an application rather than a static PDF. These are small details,
but they matter in the user experience, making Power BI feel more like a
professional app and not just a "sheet of charts."
Limitations: Despite
the improvements, the Image visual isn't a "data" visual in the
strict sense (it doesn't aggregate data, it only displays images). Furthermore,
using images from the model can impact the model's size if not managed properly
(in general, it's better to use URLs to images hosted on a CDN or SharePoint,
rather than importing them all into the.pbix ). Make sure images uploaded via URL are accessible (if the
report is running, the URL must be public or managed with authentication that
PBI can use).
OneLake Catalog User Data Functions in
Translytic Streams
This part is a little more technical and tied
to the Fabric ecosystem. OneLake is the
unified data lake within Microsoft Fabric, and OneLake
Catalog is the centralized data catalog where one can register various
assets, including user- defined data functions. functions ).
The update mentioned states that
OneLake is now Catalog supports the selection of
user-defined data features in translytic streams.
Let's try to understand this in simple terms:
Translytic
flows is a Fabric term for processes that combine
transactional and analytical operations (typically in the data engineering
space on Fabric). For example, in a Fabric data pipeline, you might want to
apply custom functions.
User data
functions in this context I imagine are definitions of
transformations or calculations (like functions in SQL, Spark, or other) that a
user has created and saved for reuse.
Before this update, finding and using these
features in the OneLake catalog was likely limited.
Now:
You can browse
and search for user functions in the OneLake
catalog when building a flow that requires one.
You can see
the details of the function and filter them (for example, filter by my
functions vs. company approved functions).
The context
is when you are selecting a function for an action of type Data function in a
data flow (presumably, for example in Data Factory inside Fabric, you integrate
the function into the pipeline ).
This feature is referred to as Preview.
Practical implications:
This is primarily aimed at data engineers or developers working
on Fabric and integrating the results into Power BI. The connection with Power
BI is that the platform is tightly integrating with OneLake
(the data lake behind the datasets). Having user functions at hand means reusing
logic defined elsewhere within your own processes. For example, if someone
on the team wrote a function to calculate a customer's tax age and published it
to the catalog, I can now search for and apply it in an ETL flow on Fabric
without having to rewrite it.
For a pure Power BI user, this update
might go unnoticed, as it affects data preparation more than the report
experience. However, on a strategic level, it signals Microsoft's intention to
unify the experience between Power BI and the rest of Fabric: using
functions from the catalog in a report isn't yet a concept, but who knows,
maybe you'll be able to do so in the future (for now, it remains within
pipelines and dataflows ).
Example: A company
has a data science team that created a function in Spark to classify customer
loyalty based on history. This function has been saved in OneLake.
Catalog. Now a data engineer preparing a fact table for Power BI can drag an
Apply function operation into their Fabric pipeline, choose the
customer_loyalty_score function from the catalog (they
can find it by searching the integrated catalog), and apply it to the customer
data in the pipeline, generating a field that then goes directly to the ready-to-use
Power BI dataset. Without that integration, they would have had to know where
the code resides, replicate it, etc.
Benefits: reuse,
consistency (everyone uses the same certified logic), governance (only approved
functions appear if marked, thanks to filters like Endorsed in your org ).
functional perspective,
for a BI manager, this potentially means fewer discrepancies between
measurements taken by different teams, because if there's common logic, it's
encapsulated in a function in the catalog and reused. This improves the
reliability of the data that ultimately appears in reports.
Note: This is a
very Fabric feature, which further suggests Microsoft is pushing Power BI
users toward Fabric adoption (which includes data engineering, data factories,
etc.). For companies still only using standalone Power BI, this might be a sign
that Fabric is worth exploring to get the most out of the integrated
tools.
In this chapter on new features in Reporting,
we've seen how even seemingly minor elements (like sizing columns or adding an
image) can make a difference in Power BI's day-to-day usability. Adaptive
matrices reduce manual work, the new Card transforms KPIs visually,
giving us more visual impact, the advanced Image view allows for richer
and more dynamic interfaces, and the OneLake integration highlights the
path to an increasingly unified and reusable data analysis environment.
Microsoft Sources: The
information presented here is taken from the Power BI November 2025 Feature
Summary and official documentation (e.g., the Learn table and matrix
page for the Grow to fit option ). For more information, we recommend reading
the original Power BI blog post for additional examples and screenshots of the
new options, as well as the updated documentation for the Card and Image
visuals for full details.
In this chapter, we address innovations related
to data modeling in Power BI. These features are primarily intended for
power
users, developers, and BI engineers who design and manage
datasets and data models on which reports are built. Specifically, the November
2025 update brings:
Power BI
Modeling MCP Server (Preview): a local server to
enable AI agents to interact with the model (related to the Copilot discussion,
but here from a modeling perspective ).
Semantic
Model Version History: Now generally available, allows
you to preserve and restore previous versions of a dataset.
TMDL
Extension for Visual Studio Code (General Availability )
- The tabular model definition editor in VS Code with new features.
Let s take a detailed look at how each one
improves the lives of those who work under the hood of Power BI.
Power BI Modeling MCP Server (Preview)
On-Premise AI for Your Model
This functionality, complementary to the one
discussed in the Copilot chapter, provides a Model Context server A locally
executable AI Protocol (MCP) that enables AI agents to interact with
Power BI models using natural language and commands.
In practice, if the remote MCP server part
(discussed
before) allows you to chat with data on a cloud server, the local Modeling
MCP Server seems to be a component designed for Power BI Desktop or
local development, where AI can help build/modify the model itself.
From the short summary:
This local
server allows you to agents (like Copilot) to interact with models like
those opened in Power BI Desktop, providing the ability to build and edit
models using natural language.
It allows you
to perform bulk operations on the model, apply best practices and
agent workflows (perhaps automated scripts).
It is
available as an extension for Visual Studio Code (which suggests that it
works in synergy with VS Code to let the AI talk to the model).
This sounds like: you could have an AI agent in
VS Code that, via the local MCP server, does things on the Power BI model. For
example, I could tell it, "Create a date hierarchy in the Calendar table
and add a Year -to-Date measure on sales," and the agent could perform
these actions in the open model.
Since this is in preview, they are
likely testing these capabilities with a limited audience or on specific
features.
Practical implications:
For a BI developer, this could be a game- changer in model
design: imagine being able to ask AI to create 50 standard measures (Margin,
Growth %, etc.), or rename fields using consistent notation, or implement
formatting rules. It could accelerate data modeling tasks that currently
require manual DAX writing or repetitive clicks.
While still immature, it suggests that
Microsoft envisions a future where AI- assisted modeling allows more
people to successfully build a model without knowing every detail of DAX or
Power Query.
Example: A BI
developer is working on a complex model with many tables. In VS Code (with the
appropriate extension), they launch the MCP server and open an AI chat for the
model. It asks: Add a date table to the model with Year, Month, and Quarter
columns and set up relationships with the Sales and Orders tables on the
DateKey columns. If the AI
is advanced enough, it could do exactly that: generate the DAX for the date
table (CALENDAR), create the columns, and establish the
relationships. The developer then checks the result, but if everything
is OK, they've saved considerable time.
Considerations: Obviously,
entrusting AI with model management requires utmost care: an error could
break existing measurements or introduce ambiguity. It's likely that initially
it will be used for tedious but simple tasks, while the critical parts will
always be verified by humans.
This is a long-awaited new feature in
Power BI dataset management: the ability to keep a history of model versions
and revert to a previous version when needed. It will become
generally
available in November 2025, meaning all datasets (likely in Premium
workspaces or on Fabric) will benefit from this feature without having to opt
into preview.
How does Version History work?
From the notes:
Up to 5
recent versions of the model are automatically
captured in certain events, such as:
o
When you open a template in edit mode in the
service.
o
When you publish or upload a. pbix
file to the service.
o
When you do a version restore (I believe this also
generates a checkpoint).
These
versions are viewable in a history panel similar to
the one in Office (those who have used version history in SharePoint/OneDrive
will recognize the concept).
You can restore
a previous version directly from this list.
Essentially, it provides a safety net:
if an update to the dataset causes problems (e.g., you accidentally delete an
important measure, or make a mistake in reporting), you can revert to a version
with one click instead of having to manually re -publish the old version (which
sometimes isn t even easy to preserve).
Practical implications: BI development teams often mimicked versioning by maintaining
different.pbix files (v1, v2, etc.) or using tools like ALM
Toolkit to save models. Now, some of this is integrated, at least for the
latest releases. This simplifies lifecycle management: if a bug is discovered
after a release, you can quickly roll back to the previous working version,
minimizing downtime for users.
For IT professionals and administrators,
this represents an improvement in governance and compliance: being able to
retrieve previous versions helps with audits or in situations where it is
necessary to understand "what has changed" even if diff is not
explicitly mentioned, having the historical version allows for manual
comparison.
Example: An analyst
changes the "Net Profit" calculation in a dataset and publishes the
update. A day later, it's discovered that the new calculation was incorrect,
causing confusion in the reports. With Version History, the analyst opens the
dataset in the service, goes to the Version History section, sees the version
from two days ago (before the change), and clicks Revert. Within
seconds, the dataset reverts to its previous state (with the correct metric),
and the reports display the correct value again. He can then correct the metric
offline and republish again. Without this feature, he would have had to find
the old.pbix file and republish it,
or rewrite the formula from memory.
Limitations: It
maintains up to 5 versions so it's not infinite versioning. A backup strategy
is still required if you want a longer history or if you want it to be
replicated in code. It probably only works on the Service (so for cloud
datasets); for those who only work locally on.pbix,
versioning remains manual, but you can always publish it to get the
functionality.
TMDL stands for Tabular Model Definition
Language, a scripting language (similar to
JSON/YAML in syntax) for defining Power BI tabular models (which share the same
Analysis Services engine ). Microsoft released an
extension for Visual Studio Code to work with.tmdl
files in the context of Power BI Projects.
With the November 2025 update, this TMDL
extension becomes generally available, and brings several new features:
Syntax DAX
highlighting in TMDL: When writing DAX expressions in the.tmdl file,
DAX syntax is now highlighted (colors for functions, fields, etc.). This makes
it easier to read and write measurements and calculations directly in the text
code.
Power
Query (M) support: The extension likely now also
handles the Power Query part of the model (the M queries), perhaps offering
intellisense or at least structuring those sections.
Breadcrumb,
code navigation: Improvements in navigability, to jump
between sections of the model easily, suggesting that if I have many
tables I can have an index or similar.
Code
actions and formatting: possible shortcuts (e.g.
generate a measurement stub, fix formatting) and automatic formatting of the
document, to make it consistently readable.
Localization:
Multi-language support, useful if your IDE and locale
names have variations.
Power BI Projects workflow,
where the entire Power BI workspace is treated as a project with source files
(a concept introduced to bring Power BI closer to Source Control and DevOps ).
In this scenario, being able to modify the model via code allows for git
version management, merges, etc.
Practical implications:
For BI developers who prefer a code-first approach or
have advanced ALM needs, this GA extension means the tool is ready for everyday
use. You can develop a model in the comfort of VS Code with all the tools (and
even the benefit of writing measurements in a robust text editor) and then deploy
it to the service. This also facilitates continuous integration: you can
create build scripts that validate the TMDL and publish it.
Not all BI teams will adopt TMDL, because it
requires coding skills and is outside the user-friendly environment of PBI
Desktop, but for complex enterprise projects with dozens of tables and hundreds
of measures, managing the model as code can reduce errors and foster
collaboration (multiple people can work on different parts and then merge the
changes).
Example: A team of
three developers is working on a large enterprise model. Instead of passing
the.pbix file around, they use Power BI Projects: each can
edit different parts (one adds a dimension table, another writes new financial
calculation measures) directly in the TMDL on separate branches in Git. Thanks
to the IntelliSense extension, the measure author sees tooltips for existing
columns or measures while writing DAX, and syntax errors are highlighted. They
can code review the changes as they would any other code (for example, checking
that the names follow the standard). Finally, they integrate the changes, and
the latest commit is published to the service as a new
dataset. This flow provides greater control over the changes and enables
scenarios like rollbacks and parallel branching (development vs. production)
without too much difficulty.
Considerations: TMDL
is conceptually similar to the JSON schema that defines a dataset, but with a
simpler syntax. With the GA extension, we can expect Microsoft to push Power BI
Projects more in the future as a natural progression for development teams
(especially when combined with Fabric, where a project can include pipelines,
report pages, etc.).
For those who prefer the traditional approach
(PBI Desktop), nothing directly changes: it's a plus for those who want to use
it.
In summary, this chapter on what's new in Modeling
highlights how Microsoft is simultaneously:
Paving the
way for AI even in the modeling phase (MCP agent
server, preview).
dataset management
and governance capabilities (ready-to-use historical versions).
Providing
professional development tools for those who want to
treat Power BI as code (TMDL extension in VS Code).
These are innovations that empower advanced
users without directly impacting the end user, but the benefits also
trickle down to the latter: more robust models and faster development translate
into reliable reports available in less time.
References: The
Power BI blog announcement for Modeling and documentation on TMDL and Version
History are sources for this information. For example, the section on Version
History explains how history is captured and managed (similar to Office). For
TMDL, Microsoft Docs and GitHub have demo resources, while the extension can be
downloaded from the VS Code marketplace.
Moving on, we come to the news in the field of
Data
connectivity. In November 2025, the main focus is
on two very important connectors in the Big Data ecosystem: Spark and Impala.
The update brings version 2.0 of these connectors into general
availability ( Generally Available). Available
), with a renewed architecture
based on Arrow Database Connectivity (ADBC).
Spark and Impala 2.0 Connectors Faster and
More Secure with Arrow (ADBC)
Apache Spark and Cloudera
Impala are query engines on big data clusters (Spark is general- purpose
Distributed computing (Impala is a SQL engine on Hadoop ). Many companies have
large datasets in data lakes or Hadoop clusters, and
connecting them to Power BI is essential for analyzing that up-to-date data.
The news is that Microsoft has released a new
implementation for these connectors:
Based on ADBC
(Arrow Database Connectivity): ADBC is an open-source driver (sponsored by
Apache Arrow) that aims to provide an efficient connection layer between
analytical applications and databases, leveraging the in-memory Arrow columnar
format. Essentially, it's a driver that optimizes the exchange of tabular data
between systems.
The
Spark/Impala 2.0 connector built on ADBC promises better performance and
increased security.
Specific
benefits mentioned:
o
Reduced overhead:
Arrow is known for minimizing data copying and enabling zero-copy transfers
between processes. This can make data extraction between the database and Power
BI more efficient.
o
Memory safety: Arrow
drivers are written with memory safety in mind, reducing the risk of crashes
due to buffer overflows etc. (issues that could affect native drivers).
o
Seamless integration with Fabric and PBI
Desktop: Being a common driver, it makes it easy to
align across environments (for example, using the same driver on Fabric Data
Factory and Power BI Desktop).
Spark and
Impala are part of this shared implementation, so it's possible they now have
similar connection parameters and better reliability.
Simply put, if previously Spark/Impala
connectors could be based on classic ODBC/JDBC, they are now moving to a more
modern system (Arrow) that should speed up import and DirectQuery times, and
reduce any compatibility issues, by leveraging open standards.
What changes for the user?
Ideally, nothing drastic in terms of how they connect: the user will
see the connector updated to the new version when they choose "Spark"
or "Impala" in the data sources in Power BI. They might notice:
Improved
performance: Queries that used to take 2 minutes now
maybe take 1 minute, or faster dataset loads.
Stability: Fewer connection errors or dropouts.
Maybe some new
connection options: for example, Arrow allows you to pass complex data
types; they could have added support for types that previously didn't map well.
This GA update means it's considered ready for
production use; those who were already using the preview of these connectors
(if available) can now count on official support.
Implications for different roles:
For data
analysts: if you're working in DirectQuery on a Spark cluster for
near-real-time reporting, you can expect lower latency. At the same time,
Arrow, being columnar, could compress data better in transit, so even importing
millions of rows from Impala could be more efficient.
For IT /database
admin: Arrow drivers are open and optimized, giving you more confidence
that you won't run out of resources. Additionally, Arrow is designed for
efficiency on modern architectures (multicore, vectorized). operations ). So
perhaps less load on clusters and gateways, with the same queries.
security front, a modern driver reduces the risk of exploits via outdated
libraries. And having an open-source project behind it can facilitate bug fixes
and transparency.
Use case: A company
has a data lake on Azure that runs Spark (Azure Databricks, for example) to
prepare data. With PBI November 2025, they upgraded the gateways and Desktop to
the new version and started using the Spark 2.0 connector. They noticed that a
dashboard that previously struggled with DirectQuery (visuals perhaps taking several
seconds to load) now runs smoother, perhaps because the Arrow driver exchanges
the data in better columnar blocks. Furthermore, where there were previously
issues with certain types (e.g., high DECIMALs from Impala), they are now
imported correctly.
Additional considerations:
Arrow is a rapidly evolving project, and ADBC is relatively new. By
embracing it, Microsoft is signaling its commitment to adhering to open
standards. This could also pave the way for other future connectors on Arrow
(e.g., connectors for cloud databases like BigQuery and Snowflake, if not
already present, could leverage Arrow for consistency).
From the perspective of the common Power BI
user, the difference is under the hood, but it will contribute to a smoother
experience when working with big data.
This chapter, while focused on a single topic,
emphasizes the importance of efficient data connectivity. A great model
and beautiful visuals are of little use if you can't connect well to your data
sources. The Spark and Impala updates reflect the continued focus on this
aspect: keeping connectors up to date and leveraging new technologies to
speed up access to heterogeneous data.
Anyone interested in learning more can check
out the detailed announcements and perhaps any benchmarks Microsoft or the
community may publish on pre- and post-Arrow performance. The official Power BI
Spark/Impala connector documentation will highlight any requirements (for
example, it may require installing new drivers on the on- prem gateway if
used).
The latest innovations to be covered concern custom
visualizations and visual enhancements in Power BI. The November 2025
update introduces several new visualizations available on AppSource or
built-in, and enhances some existing ones. For convenience, we'll split the
discussion into two parts (Chapters 8 and 9) given the number of items.
Activity
Gauge by Powerviz
Decomposition
Tree All Expanding (new feature of the breakdown tree view)
Dynamic legends in Zebra BI Charts
Drill Down Bubble PRO by ZoomCharts
These cover a range from multiple KPI tracking,
multidimensional analysis, adaptive financial visuals, to interactive bubble
charts.
Activity Gauge by Powerviz
Measure multiple progress at a glance
Activity Gauge is a
new custom visual (developed by Powerviz
) that allows you to view progress against targets for multiple categories
simultaneously. Think of it as a cross between a gauge (a semicircular
indicator like a speedometer) and a multi-category graph.
Characteristics:
It is able to display multiple gauges in one, one for
each dimensional category, allowing you to see progress towards a goal for
each.
It supports multiple
targets and custom data labels, with the ability to define thresholds (e.g.
different colors at 50%, 80%, 100% of the target).
It offers detailed
color customization, with 7 predefined schemes and over 30 available
palettes. This is useful for making it match corporate colors or emphasize
certain states (green/yellow/red).
Smart
labels: Labels that avoid overlapping and position
themselves intelligently even if there are many gauges (perhaps they rotate or
jump if space is tight).
Customizable
Center: The center of the gauge can be customized; for
example, in a classic gauge there is a central value; here you could have a
common text or icon that qualifies the whole.
Does it
support interaction and tooltips as expected in PBI visuals, and is it [
certified ]?? I don't know but probably yes if it appears in the summary.
Other
functions: Fill patterns, conditional formatting, ranking (maybe
show top N gauge if many), annotation, grid view (maybe it
switches from circular gauges to some sort of progress bar? I don't have the
exact details, but it sounds rich in configurations).
Practical use: A
typical scenario for the Activity Gauge might be a performance dashboard for
different departments or regions: for example, display a series of gauges, one
for each region, indicating the percentage of monthly sales target achievement.
This way, in a single view, you can immediately see who is above 100% (perhaps
in blue), who is at risk (in yellow), and who is below target (in red).
For a manager, this is very intuitive: a
quick glance and they understand the relative position of various elements.
Unlike a normal gauge, here you see them all together instead of just one;
unlike a bar chart, here you retain the "target" indication more immediately
(the middle of the gauge, the needle). And it also creates a strong visual
impact if designed well.
Implementation in Power BI:
Since it's a custom visual, it will need to be imported from the
marketplace ( AppSource ) into Power BI. Once imported, it appears as a visual
in the report. The report author must map:
A category
field (e.g. Region).
A progress
value (e.g. Current Sales).
A target
value (e.g., sales goal). And then configure the details (colors for different
ranges, etc.).
Does this visual offer drill-down
capabilities ? Unclear, but with "Activity" in the name, perhaps
it could represent hierarchies? Not mentioned, but consider: could
"Activity Gauge" also suggest timelines? However, the description
mentions multiple static categories.
Advantages and limitations:
Advantage:
allows comparison of attainment across multiple elements, better than many
separate, disordered gauges.
Limitation:
If there are too many categories (say, 50 products), the multiple gauge isn't
ideal; a bar chart is better. So, it's situational: useful for comparisons of a
small number of entities (say, up to 10?).
Another
limitation could be comprehensibility: make sure to label well which gauge
corresponds to what (smart labels help, but if they are too many or close
together, it can be confusing). The grid view perhaps transform the layout to
mitigate this.
Implications for users:
An analyst can use it to present performance results in a more
engaging way to the business. A business user might find it motivating
to view their department as a gauge, like a car dashboard, encouraging target
completion.
The Decomposition Tree is a native Power BI
visual that's very useful for drill-down analysis where the user
explores the contributions of different factors to a total value. The update
introduces a new feature called " All Expanding .
What it does: Allows
you to add a field as a legend or series that creates additional bar
charts next to each tree node showing a parallel breakdown on another
dimension.
In practice:
Previously,
the tree worked like this: you have a total value (e.g., total sales), you
choose a dimension (e.g., region), the tree splits into branches (Europe,
Americas, Asia, etc.) with proportional lengths. Then you choose a node and
another dimension (e.g., product), and that expands into sub-branches, and so
on. It's an iterative drilldown where at each step you choose the next
dimension to break down.
With All
Expanding , it seems that you can add an additional dimension ( legend
) and then the tree displays a small graph next to each main node representing
the breakdown according to that other dimension. For example, if the main tree
is by Region, you could add Market (Consumer vs.
Business) as a legend and the tree would look something like:
a node for Europe with a mini-column chart next to it (two columns, one for
Consumer and Business showing the share of European sales of those segments).
Simultaneously, a node for Americas, with its own mini-chart for
Consumer/Business.
Indeed, the note mentions "more column
charts beside each node showing breakdown from another dimension."
So exactly that: for each node, a small column chart appears that
expands "all" of that element with respect to another dimension.
also image
support in nodes (perhaps to put icons or photos related to size values, e.g.
country flags).
Why it's useful: It
removes a limitation: previously, you could only decompose one branch at a time
in depth; if you wanted to see the composition in two dimensions simultaneously
for all of them, that wasn't possible with a single view. Now you can at least
do that for one additional level. In practice, the decomposition tree can now
do a parallel split on two dimensions: the first (that of the nodes) and
the second (that of the legend) displayed alongside.
A use case: sales analysis where you want to
see the first level by geographic area and, within each area, the
division between product lines. Now you don't have to expand one area at
a time; you see them all side by side: Europe with the corresponding Product
A/B/C columns, Americas with A/B/C, etc., immediately comparable. If you then
want a third level, you return to the usual drilling mechanism.
This turns the Decomposition Tree into
something more like a visual mix between a tree and a comparative
cluster bar chart.
Usage considerations:
The user can still interact: perhaps they can
choose the as -legend dimension and activate it, and the tree remains static at
two levels. Or they can still expand a certain branch further.
Adding images to the nodes is an aesthetic
touch: now that there's room for a defined node, being able to add an image
(country logo, product icon) makes the tree prettier and more identifiable.
Practical implications:
For analysts
and business users, the decomposition tree was already a highly appreciated
tool for root cause analysis (including its root cause suggestion feature).
This addition also makes it a multidimensional presentation tool. For example,
in a meeting, instead of showing two separate charts (a chart by
regions and a percentage breakdown of segments by region),
you can show a single integrated visual.
Maintaining
the possibility of interaction: the user could say I see that in Asia the red
column (e.g. B2B) is dominant, now I expand Asia to see the countries , and the
tree updates.
Limitations: Spacing
is important: if you have too many nodes plus mini-charts, readability on a
single screen can suffer. It probably works best if the first level doesn't
exceed maybe 5-7 nodes, otherwise the infographic becomes tiny.
Zebra BI is a well-known provider of custom
visuals for financial and other analysis. They have visuals optimized for
FP&A, standard IBCS reports, and more. One of their visuals is listed as
Dynamic chart legends.
What it is:
Dynamic legends mean that the legend labels
(typically the series names in a chart) automatically adjust based on the
selected filter or scenario. More specifically, it seems to be used for
scenario comparisons:
For example,
in financial planning, you have a Budget vs. Actual vs. Forecast scenario.
Often, in a chart, if you select to see Budget and Actual for a certain year,
the legend might display those static names (and the user can use filters to
choose which scenario).
With this
feature, it seems that if the user filters by Scenario = Budget vs
Actual , the legend understands and changes the names
accordingly or displays those entries. Or in Year -Over- Year comparatives, if
I'm looking at 2025 vs 2024, maybe the legend will say 2025 and 2024
instead of Year 1 and Year
2 .
The context provided: eliminating confusion from static labels and useful for YOY planning
and analysis, regions etc.. So:
Before: A static legend could say e.g. Current vs Previous and the user had
to understand that Current = 2025, Previous = 2024 for the selected.
Now: the legend can expand Current (2025) vs Previous (2024)
automatically, or even change to Budget vs Actual if I have those
categories.
Zebra BI Charts already had a lot of
flexibility, here it appears as a refinement to make the charts more
self-explanatory.
For report designers, this is great in
contexts where a single chart template analyzes multiple scenarios/years: the
dynamic legend prevents interpretation errors. The end user doesn't have
to remember what the blue and red series represent depending on the filter;
they read it directly.
Implications: Less
need to add titles or explanatory notes, more immediacy. Especially in self-service
contexts, if a user plays with scenario filters on a financial report, the
visual remains clear without having to be customized for each user.
Drill Down Bubble PRO by
ZoomCharts.
Interactive multi-level
bubble charts.
ZoomCharts is another vendor known for its interactive visuals with great
animations and smooth drill-downs. Drill Down Bubble PRO is a new visual
that allows you to create bubble charts with drill-down capabilities into
hierarchies.
Characteristics:
Multi-level
drill-down on bubbles: Categories can be represented
as bubbles (circles of size relative to a value), and if a bubble represents a
group (e.g. Electronics aggregates products), the user can click it to go
down to the next level (see the sub-bubbles of Electronics, e.g. TV, Computer,
Phone).
options for
bubbles: colors, sizes, and even custom shapes (marks).
Images as
markers: Perhaps a bubble can have an image or icon
inside, or the bubble can be represented by an image, which is useful in many
contexts (e.g. brand logos).
Trendlines,
thresholds, area shading: Very interesting, this means that in addition to
bubble points, they could support trendlines if the data has a time
dimension, and thresholds with colored areas (perhaps to highlight quadrants
for X/Y values). This enriches the analysis, almost becoming an advanced scatter
plot with extra features.
Intuitive
UI Interface: ZoomCharts
typically has features like wheel zoom, panning, and built-in filters. They
likely offer a more customized experience here as well (perhaps if there are a
lot of bubbles, there's a mini-map or size slider).
Typical uses: Bubble
charts are used to represent two-dimensional distributions (different X and Y
variables) with a third value as the bubble size (e.g., sales as X, margin as
Y, bubble size = number of customers). Integrating drill-down:
You could
start with bubbles by market sector, then clicking the Industry sector gets
bubbles for individual companies in that sector, etc.
Or
geographically: continent bubble, country drill, city drill.
For analysts, such
a tool allows business users to explore data in a discovery way: they see large
circles, click on an interesting one, and navigate within it without having to
change pages or views.
Compared to native scatterplot:
PBI has basic scatterplots with bubbles, play
animations, but no hierarchy drill (you can have one point per
aggregate and you don't have a click to show subpoints).
So this fills a gap. ZoomCharts
generally focuses on performance even with many points, thanks to its
efficient canvas.
License:
ZoomCharts PRO usually requires licensing for production
use (not free beyond dev or small limits ). The summary at the end mentions
Includes free Developer License for Desktop , which suggests: you can try on
Desktop without restrictions, but to publish you may need to purchase a
license. This is to be considered in real use; however
many enterprises invest in such visuals if it is worth it.
End-user impression:
It's definitely a visual wow factor.
Seeing bubbles expand with a smooth animation when you click them, perhaps with
images inside, is captivating and brings the presentation to life. However, you
need a suitable dataset and story (don't overuse it if you don't need it).
In conclusion of this Part
1 dedicated to visuals:
Activity
Gauge provides a new way to visualize progress across
multiple objects simultaneously, useful for directly comparing multiple KPIs.
Decomposition
Tree enables multi-dimensional analysis in one go,
enhancing an already powerful visual.
Zebra BI continues to refine itself with presentation refinements (dynamic
legends) that make the difference in financial reporting.
ZoomCharts Bubble
enriches the visual library with an exploratory tool for complex data,
combining aesthetics and interactivity.
All these additions aim to give report creators
more options for communicating information clearly, engagingly, and
effectively. Each of these improved visuals/apps has its own ideal scenarios:
the key is knowing how to choose them and configure them appropriately for your
target audience.
In the next chapter (Part 2) we will continue
with further visualization innovations, completing the picture.
We continue our review of the new visualization
features with the second part, which includes:
Power BI
Theme Generator (a revamped tool for generating consistent and accessible
themes).
Power Gantt
Chart by Nova Silva (updated with task dependency management).
Synoptic
Panel by OKVIZ (visual for interactive custom maps, evolved with new features).
Customizing themes (colors, fonts,
visual style) is important for creating reports with a corporate identity and
for complying with design and accessibility guidelines. Microsoft and the
community have long had tools for generating theme JSON files, but they've
often been a bit technical.
The Power BI Theme Generator mentioned
appears to refer to a tool called BIBB Report Theme Generator, which has been
rebuilt and expanded. It's likely a web app or add-in that helps you create
themes visually, and it's now integrated with AI capabilities and new
options.
Features mentioned:
Smarter
color engine with contrast control: Helps generate
palettes that meet contrast criteria (essential for accessibility, WCAG). It
could flag if two colors have insufficient contrast and suggest alternatives.
Creating
Gradients: Allows you to easily create graduated
ranges of colors, useful for heat maps or diverging series.
AI-generated
themes: You can likely give input (like a logo, image,
or keywords) and the AI will automatically suggest a color theme, perhaps
inspired by your brand.
Brand
Color Presets: Perhaps it contains palette collections
from famous brands or industry types.
Applying
the theme via Fabric: Perhaps integrated with the
Fabric interface to apply it on the org ? Or publishing it to Fabric
OneLake ? Not very clear, but it sounds like there's now
some awareness of the Fabric environment around the theme generator.
Font
Picker: An interface to easily select theme fonts
(previously you had to write in JSON).
Live
preview: As you edit, see a preview of a sample report
or test visuals.
Advanced
BI.ST Mode: This is curious. BIBB is an author, BI.ST
Mode is perhaps a proper name, I don't know. Could it be a "Business
Intelligence Style Template" design mode with advanced parameters? I don't
have the full context, but presumably a set of advanced options for those who
want fine-grained control.
In short, an improved theme generator app, made
more user-friendly and powerful.
Practical utility:
For report
creators who aren't design experts, having a tool like this allows you to
easily generate a professional-looking theme while maintaining consistency. For
example, you can enter the codes for 2-3 corporate colors, and the generator
produces extensive palettes (tints and shades) and ensures readability.
BI Center
of Excellence managers, it's
a way to define a unified style for all company reports and share it as
a themed.json file. And with built-in accessibility
checks, it ensures that all themed reports are readable even for colorblind
users or on projectors in bright environments.
Example of use: A
bank's BI team wants all dashboards to consistently use the corporate colors of
blue and orange. Using Theme Generator, they upload the logo (does AI extract
color palettes from the logo?) or specify key colors. The tool generates
possible combinations and previews of how a pie chart, bar graph, etc., will
look. The team adjusts a couple of tones, checks that the white text on blue
has sufficient contrast (the tool indicates OK/NO). They choose a corporate
font (which appears in the list). Finally, they export a JSON of the theme.
This file is distributed to report developers and set as the default theme in
every PBIT file. The result: uniformity and visual compliance.
AI Implications: If
artificial intelligence really exists, perhaps it could generate something,
given the description "modern, elegant theme with cool colors." This
could speed up the creative process for those unsure where to start.
Ultimately, this update pushes for reports to
be not only data-accurate, but also beautiful and usable. Beautiful design
increases adoption and understanding.
Nova Silva's Power Gantt
Chart is a specialized visual for project timelines ( Gantt charts ).
With the November 2025 update, it introduces Finish-to-Start (FS)
dependencies between tasks, visualizing them with arrows connecting the
bars.
What it means: In
projects, one task can depend on the completion of another (FS = Finish A to
start B). Now the Gantt chart can graphically represent these dependencies:
An arrow will
be drawn from the end of the previous task bar to the beginning of the next
dependent task bar, usually.
This allows
you to easily see the sequence and understand if there is a critical
path or which activities are connected.
Implications:
Without
dependencies, a Gantt chart only shows parallel and temporally sequential
tasks, but you don't know if two consecutive tasks are unrelated or if the
second is waiting for the first. Now you know.
It therefore
allows you to understand bottlenecks: for example, if you see many
arrows converging on a task, you know that that is a crucial node (if it delays
that one, all the subsequent ones will delay).
Helps with project monitoring: if a
dependency has elapsed (task A finished late?) and B is attacked, you
immediately see the impact.
Practical use: A project
manager can use this visual in a Power BI report to monitor the status of
the project portfolio. For example, have a slicer for each project and view the
Gantt chart complete with dependencies. In the progress meeting, sharing the
Power BI report, they can filter on project X and discuss: "We see that
the 'Implementation' task depends on 'Design Completed' (arrow). If the design
is late, move the implementation forward." This is a representation that
is traditionally available only in tools like MS Project, but having it in
Power BI allows for cross-referencing with other data (e.g., costs,
resources) in a single platform.
Technical considerations:
To use dependencies, the dataset will need a way to specify them
(perhaps a " PredecessorID " field for each
task). The visual will assume these columns and draw lines accordingly. Nova
Silva, as a certified partner, will have instructions on how to format the
data.
Benefits vs. limitations:
The main
advantage is that it turns the Gantt chart into a realistic scheduling tool,
not just a static timeline.
One
limitation is that it currently only supports FS (Finish-to-Start) and not
other types of relationships (Start-to-Start, Finish-to-Finish, etc.). FS is
the most common, but complex projects will require the others as well. Perhaps
future updates will include them.
However, FS
covers most cases, and understanding that order is
already a step forward.
Conclusion on Gantt:
With this feature, Power BI can replace the need to export project data
to Gantt PDFs, offering an interactive view. This is very useful in PMO
(Project Management Office) contexts that require portfolio dashboards.
Synoptic Panel by OKVIZ SVG Image Data:
Interactive Maps and More
Synoptic Panel is a
visual from OKVIZ (a well-known visualization provider, Gianmarco, etc. ) that
allows you to take a vector image (SVG) and transform it into an interactive
map linked to data. In practice, it allows you to color regions of a drawing
based on data values, click on them, etc.
The November update announces an evolution of
this visual:
Transform
any SVG image into an interactive visual : reiterates the basic functionality
of being able to use any drawing as a base. Examples: floor plans, technical
diagrams, custom geographic maps, organizational charts, etc.
Multi-
level navigation: You can have different levels and
navigate (perhaps like drilling down on areas? For example, level 1 world map,
I click on Europe zooms to Europe map with details).
Dynamic
switching: Perhaps dynamically switching images based
on selections? Certainly not, or it will vary the visible layers based on the
data.
Map Editor
and Label Designer: Integrated tools for easily
drawing maps and adding labels. Previously, creating the right SVG required
using external tools (Illustrator, Inkscape ). Perhaps now there's an in-app
editor or a companion app for defining interactive areas and labels.
Advanced
coloring: Ability to color in advanced ways, such as
gradients, patterns, or based on threshold values.
Map search: if it's a floor plan, does it search for an element by name and
highlight it? This is useful for large maps (e.g., type "Meeting Room
1" and the visual will find it).
Integrated
security: I'm not sure exactly, perhaps it supports
shared map security? Or perhaps it means the mapping data is kept securely
within the PBI environment. Or maybe it's alluding to preventing content
injection from SVG.
Use cases:
Building
Floor Plans: An office can have a floor plan and use
Synoptic Panel to color occupied vs. vacant offices, and click on them to see
info.
Assembly
line or production process: when loading a diagram,
color parts of the line by intensity (machines with multiple outputs
highlighted).
Interactive
organizational chart: silhouette figures or blocks for
units, colored by metrics (e.g. performance).
Custom
geographic map: E.g., manually drawn sales area
boundaries (not necessarily corresponding to known administrative areas) can be
implemented via SVG, then linked to sales data.
Implications for business users:
It opens up highly personalized data
storytelling. Instead of limiting itself to generic geographic maps, a
company can present results precisely according to its structure. For example,
a plant manager could see the outline of the plant in a report and
understand which departments are experiencing high or low production by color.
It's more intuitive than reading a table by department.
Complexity: Configuring
Synoptic Panel traditionally required preparing an SVG file with appropriately
named elements. With the new integrated editors, it could become simpler:
perhaps upload a base image (even a non-SVG one), then draw clickable areas
over it and name those areas in the visual. This would make it accessible even
to non-designers.
Limitations: It
still requires an initial design if one doesn't exist (you can't automatically
generate the map, you need a reference image). It works best with
qualitative/categorical data (discrete colors by state, intensity of a
measurement over regions). Not for precise numerical analysis, but for
monitoring and navigation.
Security mention: Perhaps
it indicates that this visual supports built-in RLS
( Row -Level Security), meaning if a user shouldn't see
certain elements, it doesn't show them? Or the result map is safe
( no injection ), which was perhaps
a concern with custom visuals and external inputs.
Summarizing the views of this Part 2:
Theme
Generator: A tool for easily improving the aesthetics
and accessibility of reports, even using AI.
Gantt with
Dependencies: enriches a
project management view, bringing PBI closer to tools like MS Project for
timeline and constraint overviews.
Synoptic
Panel update: gives you the flexibility to display
data on any image, opening the door to highly customizable dashboards (e.g.,
plant maps, technical diagrams).
They all aim for one goal: better
communication. Whether it's through a uniform and pleasant style (themes),
through clear planning ( Gantt FS), or through customized interfaces ( Synoptic
maps ), these innovations help to convey the right information in the most
understandable way for the end user.
We've come to the end of this eBook on what's
new in Power BI in the November 2025 Update. Throughout the chapters,
we've taken a deep dive into each section of the original presentation,
dedicating a chapter to each slide and fleshing out each topic with
explanations, context, practical implications, and use cases.
To briefly recap the key points covered:
We initially
highlighted the surrounding events and announcements: Fabric Data Days,
which offered training and community opportunities, and the announcement of
FabCon 2026, a major conference for Power BI
and Microsoft Fabric enthusiasts.
We then
discussed a strategic shift with the deprecation of R/Python visuals in public
embed scenarios ( app owns data ), emphasizing the importance of
planning alternatives by May 2026.
We've been
diving into the many new features of Copilot and AI, such as the arrival
of Copilot on mobile to query data anywhere, improvements to Copilot
chat and reporting to make analysis more intelligent and contextual, the
evolution of Verified Answers for increasingly reliable and
contextualized AI responses, and openness to external agents with the MCP
server.
In the reporting
space, we've seen how small touches (e.g. auto -
resize
matrices )
and big additions (the new Card view with hero images, images with
interactive states, OneLake integration)
have made it easier to integrate with existing features. Catalog ) make it
more efficient to create beautiful reports that are integrated with the data
ecosystem.
modeling front, the focus was on tools for professionals: dataset versioning to
avoid work losses, the TMDL editor in Visual Studio Code to manage the
model as code, and even a taste of AI applied to modeling with the local MCP
server.
Data
connectivity has been enhanced with faster and more
robust Arrow-based Spark/Impala connectors, ensuring Power BI continues to play
well with modern big data platforms.
Finally, a
large group of visualizations and visual design tools: new custom
visuals (Activity Gauge, Drilldown Bubble ), improvements to existing visuals (
Decomp Tree with parallel breakdown, Zebra BI with dynamic legends, Gantt with
dependencies), and design tools such as the Theme Generator and the
revamped Synoptic Panel, which help present data in the most intuitive
and business-friendly way.
Together, these new developments reveal a clear
trend: Power BI is becoming smarter, more integrated with AI and cloud
fabric, more visually rich, and more user-friendly for both developers and
report users. The push toward AI (Copilot Everywhere, Verified
Answers ) promises to further democratize data analysis,
while usability improvements (from visuals to connectors to small fixes like
the correct scrollbar ) show the constant attention to
user feedback.
Implications for different users:
For BI
professionals and developers: these updates mean new tools to learn (VS
Code extensions, MCP APIs), but they can increase productivity and solution
quality. It's worth experimenting with Copilot to accelerate prototyping
phases, and adopt version control practices with Version History and TMDL.
Particular attention should be paid to R/Python deprecation the removal of
those components, if present, should be included in the roadmap.
For analysts
and data scientists: with Verified Answers and Q&A improvements will
likely see a reduction in ad-hoc requests because end users will be able to get
answers on their own more reliably. Furthermore, new visuals offer more ways to
communicate insights you can be creative with Synoptic Panel, or effective with
Activity Gauge. Storytelling will need to be prioritized: AI can create
graphs, visuals are abundant, and the analyst's added value will be piecing
together the right story and choosing the right visualizations.
For business
users and decision makers: many of these innovations aim to improve their
direct experience. They will be able to request data on the fly from
their phones instead of waiting for formal analysis. They will find more
interactive and tailored reports (images and formatting that speak their
company's language). They must be encouraged to adopt these new paradigms (for
example, training managers to use Copilot and to trust it, while understanding
its limitations).
Next steps and suggestions:
Tools
Update: Make sure you are using the latest version of
Power BI Desktop ( November 2025 or later) to
immediately take advantage of GA card visuals, self-expanding matrices, etc.
and if your company has PBI Report Server, evaluate a cloud adoption timeline
to have these features.
Training: Consider internal enablement sessions on Copilot (it's new to many
users) and workshops on new visuals for reporting teams. For example,
demonstrate how a visual like the All-Expanding Decomposition Tree can solve an
analysis problem that previously required multiple steps.
AI
Governance: With the massive arrival of AI, define
guidelines: which datasets to enable for Copilot ( AI
prep ), how to validate verified responses, how to monitor
usage (to avoid wrong conclusions from a Copilot used with outdated data, for
example).
R/Python
Roadmap: Inventory the use of R and Python in the
enterprise BI ecosystem: if limited to external embeds, address it; if used
within PBI for internal analytics, no immediate impact, but consider more
integrated alternatives (e.g., using Python in Power Query or Notebook).
Controlled
Experimentation: Design a small pilot using the Remote
MCP server to understand how to integrate Power BI Q&A into an internal
decision support app; or try TMDL VS Code on a new model to understand its pros
and cons compared to Desktop.