Impact Analysis, XMLA and Reusable Assets

What it is

Three skills keep a growing Fabric estate manageable.

• Impact analysis shows what depends on an item before you change it.
• The XMLA endpoint lets professional tools read and write semantic models published in a workspace.
• Reusable assets let people start from a standard: Power BI templates (.pbit), Power BI data source files (.pbids) and shared semantic models.

How it works

Impact analysis

Every item has a lineage, from data sources through lakehouses, warehouses, dataflows and semantic models to reports and dashboards. The workspace lineage view draws this chain.

From a lakehouse, warehouse, dataflow or semantic model, you open Impact analysis. It lists the downstream items, how many workspaces they sit in, and, for reports and dashboards, how often they are viewed. Downstream items in other workspaces are included.

The Notify contacts option sends a message to the contacts of the affected workspaces. Use it before a breaking change, such as renaming a warehouse column that many reports use.

The XMLA endpoint

Workspaces on a Fabric or Premium capacity expose semantic models through an XMLA endpoint. The connection string is shown in workspace settings and has this form:

powerbi://api.powerbi.com/v1.0/myorg/Sales Analytics

The capacity setting for the endpoint can be Off, Read Only or Read Write. Querying needs read; deploying, editing and refreshing parts of a model need Read Write. The tenant setting that allows XMLA endpoints must also be on.

A key reason to use XMLA is to deploy metadata without replacing data. Publishing a .pbix from Desktop replaces the whole model, while an XMLA deployment can add a measure and keep existing partitions loaded.

XMLA also lets you refresh a single table or partition with TMSL:

{
  "refresh": {
    "type": "full",
    "objects": [
      {
        "database": "Sales Model",
        "table": "FactSales",
        "partition": "FactSales 2026"
      }
    ]
  }
}

Reusable assets

Opening a .pbit asks for its parameter values and then loads data. Opening a .pbids file starts Power BI Desktop at the connection step with the source already filled in; the user still signs in.

{
  "version": "0.1",
  "connections": [
    {
      "details": {
        "protocol": "tds",
        "address": {
          "server": "contoso-sql.database.windows.net",
          "database": "SalesDW"
        }
      },
      "mode": "DirectQuery"
    }
  ]
}

A shared semantic model is published once, usually certified, and report authors connect to it with a live connection. Authors need Build permission. Updating the shared model updates every report built on it, so impact analysis matters most here.

A worked example

A data engineer must rename CustID to CustomerID in a warehouse table.

1. Run impact analysis on the warehouse. It shows two semantic models and 14 reports in three workspaces.
2. Use Notify contacts to warn the owners.
3. Rename the column, then update the two models. For the large model, deploy the change with Tabular Editor or ALM Toolkit over the XMLA endpoint, so existing partitions stay loaded.
4. Refresh only the affected table with a TMSL command instead of the whole model.

How the exam asks about this

• "Identify which reports in other workspaces use this lakehouse." Impact analysis on the lakehouse.
• "Give report authors a standard starting point with parameters but no data." A Power BI template (.pbit).
• "Refresh one partition of a large model on demand." Run a TMSL refresh through the XMLA endpoint.