Skip to content
Tutorials Center
My learning

Topics

Lesson 44 of 56 ยท Service, Sharing and Security

Publishing to the Power BI Service

Publish a report from Power BI Desktop to the service: workspaces, reports versus dashboards, pinning tiles, apps, Q&A, subscriptions and data alerts.

A report on your laptop helps only you. The Power BI service (the web part, at app.powerbi.com) is where reports are shared, refreshed on a schedule, secured and discussed. Publishing is one button in Desktop, but what you publish into, and what your colleagues then see, depends on a handful of concepts: workspaces, semantic models, reports, dashboards and apps.

This lesson walks through those pieces and what you do with them. The service is a hosted product, so nothing here can run on this site: the steps are described in words and any code is read-only. The runnable examples are Python stand-ins that model the ideas (who sees what, how a tile gets pinned, when an alert fires), to make the rules concrete. Licences and menu names change from time to time, so treat the exact labels as a guide and check the current interface.

What happens when you publish

In Desktop, choose Publish on the Home ribbon, sign in and pick a workspace. Power BI uploads two things from your .pbix file:

  1. The semantic model (formerly called a dataset): the tables, relationships, measures and imported data.
  2. The report: the pages and visuals, which reference that model.

They appear as two separate items in the workspace, and that separation is useful. Several reports can share one semantic model, and you can refresh the model without touching the reports. If you publish the same file again to the same workspace, it replaces the earlier versions of both.

flowchart LR
    A["Power BI Desktop<br/>.pbix file"] -->|Publish| B["Workspace"]
    B --> C["Semantic model"]
    B --> D["Report"]
    D -->|Pin visual| E["Dashboard"]
    B -->|Package| F["App"]
    F --> G["Audience"]

Publishing creates a semantic model and a report; dashboards and apps are built from them inside the service.

Workspaces

A workspace is a shared container for content: models, reports, dashboards, dataflows. My workspace is a personal one that only you can see; use it for experiments, not for sharing. Team content belongs in a regular workspace, where members are given a role (Admin, Member, Contributor or Viewer), described in Governance, Deployment Pipelines and Best Practices.

A few practical rules:

  • Create workspaces for a team or purpose ("Sales Analytics"), not for a person.
  • Separate development from production, for example by using two workspaces, and later with deployment pipelines.
  • Which features you get depends on the workspace's licence mode: Pro, Premium per user or a capacity (Premium or Fabric). With a Pro workspace, people who view content generally need a Pro or PPU licence too, whereas content in a capacity with sufficient size can be viewed by free users.

From file to audience

Desktop Build the model and report.
Publish Upload to a workspace.
Pin Gather key visuals on a dashboard.
App Hand a polished bundle to readers.
1 / 2

Creating a report in the service

You can also build a report without Desktop. In a workspace choose New > Report and pick a data source: a semantic model that is already published, a pasted table, an uploaded Excel or CSV file, or an existing dataset. Power BI opens the editing view in the browser, which has the same Visualizations, Fields and Filters panes as Desktop but fewer modelling features (no Power Query, no relationships view; you can add simple measures).

The shortcut for beginners is auto-create report: choose a semantic model, select Create a report > Auto-create (or start from pasted data), and Power BI picks a sensible set of visuals for you. Treat that report as a draft. It names visuals generically, may chart fields you do not care about, and it is saved to the workspace only when you choose Save. Until you save, it is a temporary view. Reports built in the service use the live semantic model, so they never contain a copy of the data. If you need transformations, build in Desktop instead (Build Your First Report in 10 Minutes).

Reports versus dashboards

New users often use these words interchangeably. In Power BI they are different things.

Report Dashboard
Pages One or many Exactly one canvas
Source One semantic model Tiles from many reports and models
Built in Desktop or the service The service only
Interaction Slicers, filters, drill-through, bookmarks Click a tile to jump to its source
Typical use Exploration and analysis At-a-glance monitoring

A dashboard is made of tiles. Most tiles are visuals you pin from a report: open the report, hover a visual and choose the pin icon, then pick an existing dashboard or create one. You can also pin a whole report page as a live tile, and add text boxes, images, videos or web content as tiles. A pinned visual keeps its connection to the report: clicking the tile opens the source report, and the tile updates when the data refreshes.

Tiles do not have all the interactivity of the report. There are no slicers on a dashboard, and cross-filtering between tiles does not happen. Everything a dashboard shows is an answer you chose in advance. Its power is putting related facts from different reports side by side, and supporting data alerts (below).

A dashboard in the service is not the same as the "dashboard" people mean when they talk about a one-page report. Many teams build a one-page report in Desktop for interactivity and never create a service dashboard at all. Use a dashboard when you need tiles from several sources, alerts, or a very simple landing page.

Building a dashboard: pinning tiles and combining reports

A dashboard is built in the service, from content that already exists. A practical walk-through:

  1. Pin a visual. Open a report, hover over a chart, click the pin icon, choose Existing dashboard or New dashboard, and click Pin. The visual becomes a tile that remembers its report.
  2. Pin a live page. Instead of a single visual, choose Pin to a dashboard on the report's top bar and then Pin live page to pin an entire page as one tile. A live tile keeps its interactivity inside the tile and updates when the data refreshes. A single-visual tile is a static answer, so a live page is the way to keep slicers and cross-filtering inside one dashboard tile.
  3. Add other content. In the dashboard, Edit > Add a tile lets you add a text box, an image, a video, a web page (an embed) or a streaming data tile for values that arrive in real time.
  4. Pin from several reports. Open each report you need and pin one visual from it to the same dashboard. This is how one dashboard combines sales, support and HR facts: each tile's source stays a separate report, possibly on separate semantic models.
  5. Tidy and set up. Drag the tile corners to resize and drag tiles to move them. Use Edit details on a tile to set its title and subtitle and a custom link, so that clicking the tile opens a different report or a page instead of the report it came from. Set a phone layout for mobile readers, a featured dashboard as your landing page and a dashboard theme if you want a different look.

What a dashboard is not

  • A tile shows an answer as of the last refresh of its source. The dashboard has no refresh schedule of its own.
  • A tile does not filter its neighbors. Interactivity lives in the reports.
  • Row-level security and sensitivity labels of the sources still apply: two people looking at the same dashboard can see different numbers on the same tile.
  • If you delete or rename a pinned visual's source report, the tile breaks. Keep a list of which reports feed which dashboards.

Designing the dashboard

A dashboard is read in seconds, so its design is about priorities:

  • Put the most important number top left and read order from there.
  • Use cards and KPIs at the top for the headline figures, and a few charts below for context. Keep it to about six to ten tiles.
  • Give each tile a title that states the point ("Open tickets, this week"), and use consistent colors.
  • Pin the visual as the reader needs to see it: filtered and formatted in the report first, then pinned.
  • Do not duplicate the report: a dashboard answers "do I need to look closer?", and the tile leads the reader to the report that does.

Try it: checking a dashboard's layout and sources

This stand-in stores tiles on a grid, checks that no two overlap and lists which reports feed the dashboard, which is the inventory you would keep for support.

Python Stand-in: tile layout and source inventory

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The Welcome note tile overlaps both Sales by month (at column 2, row 1) and Headcount (at column 3, row 1), which is the kind of mistake you make when you resize tiles by hand: the service prevents overlaps in the real editor, but a layout spreadsheet like this one is handy when you plan a dashboard. The inventory shows two of the five tiles come from the Sales report, so a change there affects 40 percent of this dashboard.

Apps: a polished package for readers

Giving every colleague access to a workspace is a poor way to share. They would see work in progress and need member roles. An app packages selected content from a workspace, gives it a name, an icon, a navigation menu and an audience (people or groups), and publishes it as a stable, read-only experience.

The usual flow:

  1. Build and test in the workspace.
  2. Choose Create app and decide which reports and dashboards are included.
  3. Add the audience, which can be specific groups with different content visible to each.
  4. Publish. Readers install it from Apps in the left navigation, or get a link.

When you change content in the workspace, readers do not see the changes until you update the app. That is the point: you can iterate safely, then release. Apps also let you control a navigation experience, such as which report is the landing page, and hide items from some audiences.

Sharing options

Besides apps, you can share a single report or dashboard by link (to people in your organisation, selected people, or those with existing access), or by email. Sharing works with the permissions of the viewer, so viewers need access to the underlying semantic model as well, and row-level security (see Row-Level Security) still applies.

Other options include Publish to web and embedding. Publish to web creates a public link that anyone on the internet can open without signing in. Use it only for data that is truly public, because there is no way to restrict it. Embedding inside your own website or application, or in Microsoft Teams and SharePoint, has its own licensing and sign-in rules; consult your administrator.

Q&A: ask in plain language

The Q&A feature lets people type a question such as "total sales by region last year" and get a visual. It uses the model's column and table names and synonyms you define in the model (see Model View, Hierarchies, Formatting and Display Folders). Good names and a few synonyms improve results a lot. Q&A can be added as a visual in a report or as a box on a dashboard. Feature availability and the exact experience (including newer Copilot-based options) vary by licence and tenant settings, so check what your organisation has switched on.

Subscriptions and data alerts

Subscriptions send a snapshot of a report page or dashboard to your inbox on a schedule (daily, weekly, after data refresh). You can subscribe yourself, and with the right permissions, other people. The email contains an image and a link back to the live report.

Data alerts work on dashboard tiles that show a single number (cards, gauges, KPIs). You set a rule such as "notify me when sales rise above 100,000", choose how often to check and whether to get an email, and Power BI evaluates the rule when the data refreshes. Alerts are a feature of dashboards, not of reports, which is one more reason a small dashboard can be worth building.

Stand-in: who can see what

Roles and sharing are rules about people and content. This example builds a tiny permission table and answers "can this person open this item?" with the same shape of logic as workspace roles and app audiences. It is not the real permission engine; it only makes the rules explicit.

Python Stand-in: workspace roles and app audiences

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The example shows the key idea behind apps: Dana and Eli can read the published app without being in the workspace at all, while Ben can edit in the workspace but is not in the app audience, so his own access to the app depends on who adds him.

Stand-in: pinning and alerts

The second example builds a mini dashboard as a list of tiles, each tile storing its source report and a value, then evaluates an alert rule against a refresh. This mirrors the way tiles remember where they came from and how alerts watch a number.

Python Stand-in: tiles and a data alert

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

In the real service, the alert also records whether it already notified you, so that you are not emailed on every refresh while the value stays above the line. The stand-in checks the condition only.

A publishing checklist

Before you publish for other people:

  • Name things well. Report, model and workspace names are what users search for.
  • Hide technical fields and set sensible formats, so Q&A and analysts see clean fields.
  • Check the model size against your licence, and the refresh plan (see Scheduled Refresh and Gateways).
  • Set up security (row-level security roles, and who can see the model) before you share, not after.
  • Test as a viewer. Open the report from the app, in the browser, with a test account if you can.
  • Decide on a landing page, and put the key measures first.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. What does publishing from Power BI Desktop upload to a workspace?

    Publishing creates a semantic model and a report that references it. Dashboards are created separately by pinning visuals in the service.
  2. 2. Which is true of a dashboard compared with a report?

    A dashboard is one canvas of tiles, built in the service, and the tiles can come from different reports and models. Slicers and multiple pages are report features.
  3. 3. Why use an app instead of giving colleagues access to the workspace?

    An app bundles selected content for a defined audience and is updated deliberately, so you can keep developing in the workspace without disturbing readers.
  4. 4. Where can you create a data alert?

    Data alerts watch numeric tiles on dashboards (cards, KPIs, gauges) and notify you when a threshold is crossed after a refresh.

Summary

  • Publishing uploads a semantic model and a report to a workspace; My workspace is for personal use only.
  • A report is multi-page and interactive on one model; a dashboard is one canvas of tiles pinned from many sources.
  • Apps package workspace content for an audience as a stable, read-only experience that you update on purpose.
  • Sharing respects permissions and row-level security; Publish to web makes data public to everyone, so use it with care.
  • Q&A relies on good names and synonyms; subscriptions email snapshots; data alerts watch single-number dashboard tiles.
  • The service is a hosted product: the Python examples here are stand-ins for the rules, not the real service.

Next, learn how the data in your published model stays current in Scheduled Refresh and Gateways.