Prerequisites

System

  • Windows 10 / 11 (x64), the application is a native windows interface.
  • .NET 8 runtime installed on the machine.

To audit a Power BI Desktop model

  • Power BI Desktop open with the model you want to analyze.
  • No additional authentication required, the connection is local.

To audit a model on the Power BI Service

  • A Microsoft Entra ID (Azure AD) account with access to the target workspace.
  • A Fabric or Premium capacity with the XMLA endpoint enabled in read/write mode.
  • The Workspace administrator role or equivalent rights on the semantic model.
Performance analysis (measuring actual DAX execution times) requires XMLA trace creation rights on the capacity. Without these rights, the audit runs without the performance domain, and the score is recalculated accordingly.

Connect to Power BI Desktop

This is the simplest mode. It requires no authentication.

  1. Open Power BI Desktop Ensure your .pbix file is open and the model is loaded.
  2. Launch Power BI Audit Start the application. On the home screen, click "Connect".
  3. Select desktop mode The connection window displays the list of Power BI Desktop instances currently open on your machine. Each instance corresponds to an open .pbix file.
  4. Select your instance Click on the name of the model you want to audit, then confirm.
If multiple .pbix files are open simultaneously, each instance appears separately in the list with its filename.

Connect to Power BI Service

This mode allows you to audit a model published on the Service via the XMLA protocol.

  1. Retrieve the XMLA endpoint of your workspace In the Power BI portal, open Workspace settings → tabPremium → copy the XMLA connection URL. It looks like:
    powerbi://api.powerbi.com/v1.0/myOrganization/myWorkspace
  2. Select service mode in the connection window Paste the XMLA URL into the provided field, then click "Connect".
  3. Authenticate yourself A Microsoft login window opens. Log in with your professional account that has rights to the workspace.
  4. Select the semantic model The list of models available in this workspace is displayed. Select the one to audit and confirm.
Your credentials are never stored in plain text. Power BI Audit uses the standard Microsoft authentication mechanism (MSAL) with a secure encrypted cache on your machine. Once connected, subsequent openings will not require re-authentication as long as the session is valid.

Launch an audit

Once connected to the model, the main dashboard is displayed. The audit does not start automatically; you remain in control.

  1. Check the connected model The semantic model name and connection mode (Desktop or Service) are displayed at the top of the interface.
  2. Click on "Launch audit" The analysis starts. A progress bar indicates advancement domain by domain: Modeling, DAX, VertiPaq, Performance, Governance, Power Query.
  3. Wait for the end of the analysis Duration varies according to model size. For a standard size model (< 50 tables, < 200 measures), the audit generally takes 5 to 30 seconds. Performance analysis may take more time if your model contains many measures.

Read the results

Once the audit is finished, the dashboard displays:

Global score

A score out of 100 displayed prominently, with a color visual indicator (green, orange, red). It is the weighted average of all analyzed domains.

Domain cards

Each domain (Modeling, DAX, VertiPaq, etc.) has its own card with its score and the number of findings detected. Click on a card to filter the list of findings for that domain.

List of findings

Each finding indicates:

  • The domain, a colored icon indicating the category.
  • The severity, High, Medium or Low.
  • The rule identifier, for example DAX003, MOD005.
  • The object concerned, name of the table, measure, column or M query involved.
  • The recommendation, a concrete action to take to correct the problem.

Filter and sort

You can filter findings by domain, by severity, or search for an object by name. Use sorting by severity to prioritize the most impactful corrections first.

Export results to Excel for sharing with your team or for long-term tracking. See the results export page for more information.