Power BI datasets and dataflows are the two native data sources for Power BI reports. Connecting to a datasets allows a report to be built against an existing Power BI dataset in place, and dataflows represent a source of data that has had transformations applied to it. When connecting to Power BI dataflows, data is imported, into a data model but the connection to a Power BI dataset is a direct connection.

The two sources handle identity in drastically different ways, and this can lead to confusion when dealing with multiple accounts and tenants. This post is an attempt to help clarify this confusion

Connecting to a dataset

To connect to a Power BI dataset, select the “Get Data” button from the ribbon, select the “Power BI” tab, select Power BI dataset, and finally the “Connect” button.

Next, if a user is signed into Power BI Desktop a list of workspaces and their datasets are presented. If not, the user is prompted to sign in. The dataset can then be connected.

The important thing to notice here is the list of workspaces itself. The list presented is a list of the workspaces available in the tenant belonging to the currently signed in user. It is the same list of workspaces that can be chosen as a publishing target. It should also be noted that the identity of the user is displayed in the upper right corner of the dialog box, and the identity can be changed directly from there.

Connecting to a dataset in a different tenant

The signed in user can be changed by selecting “Sign in” at the upper right of the Power BI Desktop client, or within the connection dialog itself. If the user signs into a different account (in a different tenant), a different list of workspaces and datasets will then be offered. The dataset source is hard linked to the currently signed in user. In this way, the Power BI dataset source behaves differently than all other data sources, which maintain connection credentials separately.

Connecting to a Dataflow

Connecting to a dataflow follows the same steps as a dataset, with the exception that the “Power BI dataflows” option is chosen.

At this point, a Power BI data connection dialog will be shown.

There is only one authentication option because Power BI dataflows only support one authentication option.

Unlike datasets, dataflows are NOT linked to the currently signed in user. The connection is authenticated, not the current user. The “Sign In” button must be selected, and authentication completed to connect to a Power BI dataflow.

Once signed in, selecting the “Connect” button will display a list of workspaces that contain dataflows. Expanding the workspace and then the dataflow will expose a list of entities that can be imported into the Power BI data model.

The connection information for the dataflow is cached with Power BI Desktop, and subsequent connections to dataflows will not require the user to sign in. The same authentication credentials will be used.

it should be noted that unlike the dataset connection dialog, this one does not show the current credentials and does not allow those credentials to be changed. This makes the process of changing credentials to use dataflows in multiple tenant somewhat less than intuitive.

Connecting to a dataflow in a different tenant

With datasets, changing the currently signed in user will result in a different set of datasets being presented when the dataset option is chosen. This is different with dataflows. No matter what user is currently logged in, the cached credentials will be used.

This behaviour can be confusing when multiple tenants need to be accessed. With most other data sources, the cached credentials are linked with the specific data source. For example, when two different SQL databases are connected, Power BI caches two different sets of credentials.

To connect to dataflows in a different tenant, the current connection information needs to be cleared. This can be done with any data source, but it is particularly important to dataflows as it is the only way to switch connection credentials.

To clear the credentials for the dataset, select “File”, “Options and Settings” and the “Data Source Settings”. The Data source settings dialog will then be presented.

Unlike most other data sources which can have multiple entries in the list, one for each unique data source, there will only be one source for dataflows. It is named “Power BI dataflows. For example, if the current instance of Power BI Desktop has authenticated to 3 different SQL servers, there will be three SQL Server connections in this list, but there will only be one for dataflows, no matter how many tenants that have been connected.

To switch tenants, the current credentials must be either cleared, or edited. The cached credentials can be fully removed by selecting “Clear Permissions” or they can be changed by selecting “Edit Permissions”. If cleared, the user will be prompted for credentials the next time the dataflow option is selected. If edited, the new credentials will be stored.

Conclusions, and recommendations

It is possible to work with multiple tenants for both connected datasets and dataflows. However, the methods for doing so are completely different for either option. This can obviously lead to some confusion.

It is my opinion that this behaviour should be changed, and that the behaviour or connected datasets is the more intuitive. If the credentials for the currently logged in user were user for both types of connection, it would be much more intuitive, and also easier to user for report designers.