Data Source and Semantic Model Connections

Identify the source you need

The first connection decision is not how to load, but what kind of repository holds the required data. Power BI can obtain data from repositories that differ in structure and connection requirements, so the connector must match the source.

The durable rule is simple: identify the repository first, then choose its connector.

Azure Cosmos DB is a NoSQL store that can contain JSON documents. Because JSON is often nested or unstructured, its records may need to be expanded and normalized before they are useful for analysis.

For example, a shipping application can store one JSON document per shipment in Cosmos DB. After connecting, Navigator may show records rather than a ready-made grid. Opening the data in Power Query lets the author expand the record fields and turn them into rows and columns that can relate to other model data.

This makes the source categories meaningfully different:

• SQL Server is the relational database source.
• Excel is a file source.
• Azure Analysis Services supplies tabular semantic models.
• Azure Cosmos DB is the NoSQL JSON-document store.

Start the connection with the matching connector

In Power BI Desktop, the command that begins a connection to a new source is Get data. Select the connector that matches the repository, then supply the location or service details that connector requests. The workflow adapts to the source rather than treating everything as a generic file or database.

After those details are accepted, Power BI can authenticate and expose the source objects. Navigator is not the command that starts a new connection: it appears after Power BI has connected successfully.

A connection can also require authentication. The documented SQL Server sign-in choices are Windows, Database, and Microsoft account.

These are credential choices, not loading or modeling commands:

Keeping these stages separate prevents a common mistake: a sign-in method authenticates access, while a connector locates the source and a later action decides what to do with its data.

Inspect before you load

After Power BI connects to a file or database, the Navigator window shows the available data. Navigator lets you preview and select the tables or entities to bring into Power BI before loading them.

The preview is a verification step. Select a table or entity and inspect its contents to confirm that it is the intended data; then select the check boxes for the objects the model actually needs. Navigator does not change a moved file's path, reset connection permissions, or move calculations from an external semantic model.

A practical sequence is:

1. Connect to the source.
2. Use Navigator to inspect the available objects.
3. Select the required tables or entities.
4. Choose Load or Transform Data.

That last decision depends on whether the selected data is already fit for the model.

Choose Load to place the selected data into the Power BI model in its current state. Choose Transform Data when the selected data needs review, cleaning, or reshaping in Power Query before loading.

Server details and credentials are supplied while establishing the connection; they are not reasons to choose Transform Data in Navigator. Likewise, Analysis Services Import versus Connect live is selected in that source's connection flow, not by using Transform Data.

Choose a file location with refresh behavior in mind

The same Excel workbook behaves differently depending on how Power BI reaches it. For a local Excel file, Power BI creates a new semantic model and loads the file data into it, but changes to the original file are not reflected automatically. A local import therefore suits data that does not change frequently.

A OneDrive for Business connection is designed to keep the file and related Power BI content synchronized when changes are found. The related content can include the semantic model, reports, and dashboards. This cloud-based path is preferable when the workbook continues to change and the Power BI content should follow those changes.

A file can also move while a report is still being developed. If a source file moves during development, open Data source settings in Power Query, select the file, choose Change Source, and update the file path.

Changing the address is safe only when the replacement remains structurally compatible. The replacement file must keep the same structure because deleted or renamed source columns can break the reporting model.

Suppose Employees.xlsx moves from a desktop folder to a project folder while keeping the same worksheets and columns. Update its path through Change Source. Do not rebuild it as an Analysis Services connection, use Navigator to rename model columns, or retain the obsolete path.

Reuse a published semantic model

A semantic model is a queryable data resource optimized for analytics. Unlike a raw file or database table, a model can already contain tables, relationships, and calculations that report authors consume.

Once a Power BI model is published to a workspace in the Power BI service, it becomes a source for visualizations in reports and dashboards. This allows a governed analytical design to be used more than once instead of every author recreating its relationships and calculations.

The model and the presentation remain distinct:

Once it has been published to a workspace, another report author can use the semantic model as a reusable source for report visualizations. A model created and published by another team is therefore the reusable source to choose when its maintained tables, relationships, and calculations are needed.

That source is not any of the connection machinery around it:

• A SQL Server credential authenticates access; it does not package analytical calculations.
• A Navigator selection is a temporary choice of source objects; it is not a shared analytical asset.
• A local file path locates raw data; it does not provide another team's maintained model logic.

Reuse is the key distinction. Report authors can build new reports from a shared model without treating its existing calculations as raw fields that must be rebuilt.

Connect to Azure Analysis Services

Azure Analysis Services provides enterprise-grade tabular semantic models. Such a model can already hold data from multiple sources, relationships, metrics, security, and DAX calculations in one trusted analytical layer.

To connect from Power BI Desktop, use Get data, select Analysis Services, enter the server address and database name, and choose either Import or Connect live.

With Connect live, both the data and DAX calculations remain in Azure Analysis Services instead of being imported into Power BI. This centralizes the semantic modeling and calculations in the upstream model while Power BI provides the report experience.

The refresh behavior follows that ownership. When the Azure Analysis Services model refreshes, a live-connected Power BI report receives the refreshed results without its own Power BI refresh schedule.

This is why Connect live fits a team that wants reports to use an Azure Analysis Services model and reflect its refreshes without importing another copy. Import would bring data into Power BI and would not provide this documented live-refresh behavior.

Do not confuse the source with unrelated workflows. An Analysis Services model is not a local Excel file to load through a file path, and it does not need to be converted into JSON records. JSON expansion belongs to nested NoSQL data; Analysis Services already supplies a tabular semantic model.

Key takeaways

• Match Get data to the repository: files, relational databases, applications and online services, NoSQL stores, and semantic models require source-appropriate connectors and details.
• Recognize the stages: Get data starts the connection, credentials authenticate it, Navigator previews and selects objects, Load accepts the current state, and Transform Data opens Power Query for preparation.
• Treat file location as part of the design: local Excel changes are not reflected automatically, OneDrive for Business supports synchronization, and a moved file is repaired through Data source settings → Change Source without altering its structure.
• Reuse modeled assets: a published semantic model can supply existing tables, relationships, and calculations to reports and dashboards created by other authors.
• Choose Connect live for upstream ownership: Azure Analysis Services retains its data and DAX calculations, and its refreshes reach live-connected reports without a separate Power BI refresh schedule.