Power BI Embedded glossary

A Power BI Embedded capacity bought in Azure, in sizes A1 to A6. It is billed by the hour with no commitment and can be paused when nobody uses it. A SKUs are not being retired; what Microsoft is retiring are the Power BI Premium P SKUs.

The way of embedding in which your application talks to Power BI through its own identity, a service principal, and decides itself who may see what. The people viewing the reports need no Power BI account and no licence, which is why customer and partner portals use it.

Dedicated compute in Microsoft's cloud that Power BI content runs on. You pay for the capacity rather than for each viewer, so embedded content always needs one: a Fabric capacity (F SKU) or a Power BI Embedded capacity (A SKU).

The unit Microsoft Fabric measures compute in. An F SKU's number is its capacity units, so an F2 has 2 CU and an F64 has 64, and pay-as-you-go prices are set per capacity unit per hour.

A short-lived token your application requests for each user and each piece of content. It tells Power BI what that user may open and, with row-level security, whose data to show. Tokens expire, so the application has to renew them while someone is working.

A Microsoft Fabric capacity, from F2 upwards. It runs Power BI Embedded and every other Fabric workload, is billed per second while running, can be paused or reserved, and is the only capacity that supports Direct Lake semantic models. It is where Microsoft is investing.

Microsoft's analytics platform, which Power BI is now part of. Its capacities, the F SKUs, run Power BI reports, including embedded ones, alongside data engineering and other workloads.

Serving several customers from one setup while keeping each one's data apart. The two common ways are a workspace per customer and one shared model with row-level security, and many portals combine them.

A page-based report laid out to print exactly, for invoices, statements and long lists. It is exported to PDF or Excel rather than explored, and it can be embedded just like an interactive report.

Stopping a pay-as-you-go capacity while nobody uses it, so no compute is billed for that time, and resuming it when someone needs it again. Analytics often sits idle at night and at weekends, so pausing can save a lot. It doesn't combine with a reservation, which is paid every hour.

Pay-as-you-go bills only the time a capacity runs and lets you pause it. A one- or three-year reservation gives a much lower rate, but you pay for the reserved capacity around the clock. Reserve what runs all the time; pay as you go for what doesn't.

Microsoft's way of showing Power BI content (reports, dashboards, single visuals, paginated reports) inside your own application or website instead of app.powerbi.com. The app loads the content through the Power BI REST API, and everything around it is your own web development.

Microsoft's per-user Power BI licences. In the Power BI service, everyone viewing shared content needs one unless the content sits on a capacity of F64 or larger. The people who build and publish reports always need one; the people viewing embedded content for customers don't.

A Power BI option that makes a report public: anyone with the link or the embed code can open it without signing in, and the underlying data can be reached too. Fine for a chart on a public website, never for confidential data or reports with row-level security.

Roles in a semantic model whose filters decide which rows each person sees. A static role has a fixed filter; a dynamic role filters by the signed-in person's identity, so one role can serve everyone. When embedding, your application passes that identity in the embed token.

A Power BI iframe for pages you already run, such as SharePoint or an intranet. Viewers sign in with a Microsoft account and their own Power BI permissions apply, so each needs a licence unless the content is on F64 or larger. In practice an internal tool.

The data behind a Power BI report: tables, relationships, measures and row-level security roles. It was called a dataset until 2023. Reports query it, whether they are viewed in Power BI or embedded.

An app identity in Microsoft Entra ID that your application uses to talk to Power BI. It needs at least the Member role on each workspace it embeds from, and a Fabric admin has to switch on the tenant setting “Service principals can use Fabric APIs”.

The way of embedding in which each viewer signs in with their own Microsoft account and their own Power BI permissions and licence apply. It suits internal applications, not customer portals.

A portal that carries your brand rather than the vendor's: your logo, colours and domain, with neither Power BI nor the portal's maker on show. Some portals can also give each customer their own look.

A container in the Power BI service for reports and semantic models. For embedding, it has to sit on a capacity, and the service principal needs access to it.