Webinar - Unlock the Power of Embedded Analytics with Power BI

In the demo portal, people sign in on the company's own domain, with its logo and text, using either Microsoft Entra ID (with MFA) or an email and password. The report behaves exactly as in the Power BI service, with cross-filtering and slicers, and the portal adds its own navigation, full screen, a dark theme and export to PDF, PowerPoint or PNG. Only the report area is Power BI; everything around it is ordinary web development, and it can look any way you like. The app and the report stay in sync: page names, for example, come straight from the published report. (10:20)

Branding. Customers log in on your domain, with your logo, instead of reaching app.powerbi.com with a guest account.

Simplicity. Most viewers need a fraction of the service's buttons; an app can show just those, or add features that have nothing to do with analytics.

Licensing. Below F64, every viewer of the Power BI service needs a Pro licence (14 USD per user per month since April 2025). Embedded content runs on a capacity instead, so the people viewing it need no Pro or Premium Per User licence.

Integration. Analytics sits inside the business app instead of next to it.

(15:00) For the full comparison, see ways to share Power BI externally.

The app talks to the Power BI service through a single technical identity, a service principal with at least the Member role on the workspace, however many people use the app. Who may see which report is decided inside the app, in its own users, roles and access rights; Power BI neither knows nor cares. When a user is allowed a report, the app obtains an embed token, and the service returns the content: a report, a single visual, a paginated report, a dashboard or a tile. App users don't need access to the workspace, and they don't even need a Microsoft account.

The embed configuration has a setting for each; setting them to false and running the code hides both. In the Embedsy Portal, a React app, the same change looks slightly different, because it uses Microsoft's React library for Power BI, where page navigation is a setting called contentPaneEnabled. Same result, different abstraction: a first taste of the details that come with building your own solution. (24:24)

Through the API the app can read every page and every visual of a report, go to a page and act on a visual. Vlad navigates to a page and sets a stock slicer to Microsoft only, with a basic filter that names the table, the column and the values; the rest of the page filters as usual. Two practical lessons: pages and visuals carry generated internal names, so give visuals clear titles in Power BI Desktop, because the title is the easiest way to find the right one in code; and a basic filter fits a drop-down slicer where an advanced filter doesn't. (26:06)

Everyone claims "insights to action"; embedding makes it literal. In the Playground's example, a button takes the people filtered in a report, say everyone in France, out of Power BI and into the app, where they can be sent a coupon or a message. Connect that button to your mailing tool or Slack and you act on the data in the same place you read it, something the Power BI service on its own can't do. (37:03)

Every company produces data, and any company can turn it into analytics for its customers, suppliers or partners, and charge for it. The example is a publicly traded construction company worth around 11 billion USD that has shipped Power BI Embedded inside its own application for years, analytics section included. Lawyers, accountants and other service firms could do the same, and smaller companies are often better placed to. (39:49)

Security. A Publish to web iframe is public: anyone can copy it from a page's source and use it anywhere. Secure embedding needs authentication and tokens for every request. (44:23)

Not as such. The limit is what the capacity can serve: every Fabric capacity supports embedding, even F2, but the smaller the capacity, the better optimised the models need to be for the data volume and the number of people using it at the same time. (45:45)

The app has to handle it. Every token has a limited lifetime that can be configured only up to a maximum, so the app must refresh tokens in code; otherwise users are logged out of the report after a while. (47:14)

Exactly as in regular Power BI, with both static and dynamic roles. Although the service only ever sees the technical user, the app knows who is logged in and puts that identity into the token; if it doesn't match the dynamic rule, no data comes back. More in our guide to row-level security in Power BI Embedded. (49:00)

As of January 2025, the Embedsy Portal is a responsive Next.js web app rather than a native mobile app. It shows a report's mobile layout on small screens, so the real work is designing that mobile layout in Power BI Desktop. (50:43)

The portal installs into the customer's own Azure subscription: an App Service that runs the app, a small Azure SQL database for users and settings, a storage account for logos and files, plus a Fabric or Power BI Embedded capacity. Keeping it all in Azure avoids authentication problems with Microsoft. (52:42) See the Azure resources in the documentation.

Because Power BI is driven entirely through an API, an agent trained on that API could change a report from a request in plain English. At the time of the webinar, Copilot was not available in embedded Power BI. (54:42)