2.1.2. Data Source Settings: Credentials and Privacy Levels
💡 First Principle: Power Query stores two separate facts about every source it touches: how to authenticate (credentials) and how sensitive the data is (privacy level). Credentials decide whether you can read the source at all; privacy levels decide whether data from that source may be sent to a different source while Power Query combines them.
Data source settings (Home > Transform data > Data source settings in Desktop) lists every source used in the current file or globally on your machine. From here you can Change Source (repoint a query from a test server to production, or to a moved file, without editing every step), Edit Permissions (update credentials and the privacy level), and Clear Permissions (force Power BI to prompt again, which is the fix when a password has changed or the wrong account was cached). Authentication types depend on the connector: Windows, Database (SQL login), Microsoft account (organizational/Entra ID), Basic, Anonymous, or key/token. In the service, credentials for a published semantic model are set separately on its settings page; desktop credentials don't travel with the file.
Privacy levels exist because of query folding. To combine a local Excel list with a SQL table efficiently, Power Query might send the Excel values to SQL Server inside a query. If the Excel file contained confidential salaries and the other source was a public web API, that would leak data. Privacy levels tell Power Query which sources may see each other's data:
| Privacy level | Meaning | Can its data be sent to other sources? |
|---|---|---|
| Private | Confidential or sensitive (salaries, personal data) | No, isolated from all other sources |
| Organizational | Internal to the organization | Only to other Organizational sources |
| Public | Freely available (public web data, open datasets) | Yes, to any source |
| None | Not yet set | Treated per the file's privacy options |
When a query combines sources with incompatible levels, Power Query either evaluates the combination locally (slower, no folding across the boundary) or raises a Formula.Firewall error explaining that a query references other queries or steps in a way that would combine data across privacy boundaries. Common fixes are setting accurate privacy levels, restructuring so each source is loaded in its own query before combining, or, for trusted personal work only, choosing Ignore the Privacy Levels in Options > Current File > Privacy, which can improve performance but removes the protection.
⚠️ Exam Trap: Privacy levels don't control who can see data in reports. They control data movement between sources during evaluation. When a stem asks how to stop report users from seeing salaries, the answer is RLS, workspace roles, or removing the column, never "set the source to Private."
Reflection Question: A query merges a Private HR spreadsheet with an Organizational SQL table and suddenly refreshes much more slowly than before the privacy level was set. Explain what Power Query stopped doing, and why that change is deliberate.
