What We Collect

Prizm extracts metadata from Power BI using a combination of the XMLA endpoint and the Power BI REST/Scanner API. Semantic model schema (tables, columns, measures, relationships, hierarchies, calculation groups, and DAX/M expressions) is read over the XMLA endpoint, which requires a Premium Per User, Premium capacity, or Fabric capacity workspace (see Setup). Workspace, report, dashboard, dataflow, and data source inventory, along with refresh history, is read through the REST/Scanner API. The tables below describe every field collected for each asset type.

Workspaces

Workspaces are the top-level containers in Power BI. Prizm uses workspace metadata as the organizational root for all cataloged Power BI assets.

Field	Description
Workspace ID	Unique GUID assigned by Power BI
Workspace Name	Display name of the workspace
Type	Workspace type: Personal, Group, or Premium
State	Active or Deleted
Capacity	Linked Premium capacity, if assigned

Semantic Models (Datasets)

A semantic model (formerly called a dataset) is the data layer that reports and dashboards query against. It contains tables, columns, measures, relationships, and transformation logic built in Power Query. Prizm catalogs semantic models as a parent asset with each table and column cataloged as child assets.

Semantic Model — Top Level

Field	Description
Model ID	Unique GUID assigned by Power BI
Model Name	Display name of the semantic model
Workspace	The workspace the model belongs to
Configured By	User or service principal that owns the model
Created Date	When the model was first published
Modified Date	When the model was last modified
Storage Mode	Import, DirectQuery, Direct Lake, or Composite
Target Storage Mode	The target storage mode, if a migration is in progress
Is Refreshable	Whether the model supports scheduled refresh
Is On-Premises Gateway Required	Whether an on-premises data gateway is needed
Gateway ID	ID of the on-premises gateway, if applicable

Semantic Model Building Blocks

For each semantic model, Prizm extracts the following internal building blocks and catalogs them as child objects:

Building Block	What Prizm collects
Tables	Table name, schema, storage mode (Import / DirectQuery / Direct Lake), hidden state, row count (where available)
Columns	Column name, data type, format string, hidden/visible state, data category, lineage tag (stable GUID across renames)
Measures	Measure name, DAX expression, data type, format string, description
Calculated Columns	Column name, DAX expression, data type
Relationships	Source table, source column, target table, target column, cardinality (1:1, 1:many, many:many), cross-filter direction (single/bidirectional), active/inactive state
Hierarchies	Hierarchy name, levels (column references in order), table
Data Sources / M Expressions	Power Query connection string, source type, server/database/schema path, full M expression (for upstream lineage parsing)
Calculation Groups	Calculation group name, precedence, calculation items, and their DAX expressions
Storage Mode Config	Per-table storage mode configuration for Composite models
Partitions	Partition name, mode (Import / DirectQuery), source expression, refresh policy (for incremental refresh)

Tables, columns, measures, relationships, hierarchies, calculation groups, partitions, and their DAX/M expressions are extracted over the XMLA endpoint using AMO/TOM metadata queries, not the Scanner API. M expressions are required for upstream lineage stitching — Prizm parses them to resolve which source system table each semantic model table draws from.

Refresh History

For each semantic model, Prizm collects the last N refresh runs to power Freshness and Reliability metrics:

Field	Description
Request ID	Unique ID of the refresh request
Refresh Type	On Demand, Scheduled, Incremental, or REST API
Start Time	When the refresh began
End Time	When the refresh completed (or failed)
Status	Completed, Failed, or Cancelled
Error Message	Failure reason, if the refresh did not complete

Reports

Reports are visual presentations built on top of a semantic model. Each report page queries one semantic model and renders results using visualizations.

Field	Description
Report ID	Unique GUID assigned by Power BI
Report Name	Display name
Workspace	The workspace the report belongs to
Dataset ID	The semantic model this report queries
Web URL	Direct link to the report in Power BI Service
Embed URL	Embeddable URL for the report
Created Date	When the report was first published
Modified Date	When the report was last modified
Report Type	PowerBIReport or PaginatedReport

Dashboards

Dashboards aggregate tiles from one or more reports into a single view. Prizm catalogs dashboards and links them to their upstream reports and semantic models through lineage.

Field	Description
Dashboard ID	Unique GUID assigned by Power BI
Dashboard Name	Display name
Workspace	The workspace the dashboard belongs to
Is Read Only	Whether the dashboard is published as read-only
Web URL	Direct link to the dashboard in Power BI Service
Embed URL	Embeddable URL
Tiles	List of tile IDs, dataset IDs, and report IDs the dashboard draws from

Dataflows

Dataflows are Power Query transformation layers that sit between raw data sources and semantic models. They allow shared, reusable transformation logic across multiple semantic models in a workspace. Prizm catalogs dataflows when Extract Dataflows is enabled in the connector configuration.

Field	Description
Dataflow ID	Unique GUID assigned by Power BI
Dataflow Name	Display name
Workspace	The workspace the dataflow belongs to
Configured By	User or service principal that owns the dataflow
Modified Date	When the dataflow was last modified
Entities	Tables produced by the dataflow (entity name, data type, M expression)

Data Sources

Data sources represent the external connections behind a semantic model or dataflow. Prizm uses data source metadata to stitch upstream lineage to cataloged warehouse assets.

Field	Description
Datasource ID	Unique GUID
Datasource Type	Source system type (e.g., Snowflake, SQL Server, Azure Synapse, File)
Connection Details	Server, database, path — normalized per source type
Gateway ID	On-premises data gateway, if the source requires one

Usage & Adoption

Usage data is collected as part of the Operational job (see Setup) and powers the Report View Count metric on reports, plus its cumulative rollup on semantic models. There is no separate, standalone usage-collection job for Power BI, and no usage metric is computed for dashboards or semantic models directly.

Field	Description
Report View Count	Number of times the report was opened, over a 14-day or 28-day window
Report View Count (Cumulative)	Sum of the Report View Count across every report built on a given semantic model, for the same window

API Methods Used

Prizm uses the following methods to collect the metadata above:

Endpoint / Protocol	Data Collected
XMLA endpoint (AMO/TOM metadata queries)	Tables, columns, measures, calculated columns, relationships, hierarchies, calculation groups, partitions, and DAX/M expressions for in-scope semantic models
`GET /admin/groups`	Workspace list and metadata
`GET /admin/groups/{groupId}/datasets`	Semantic model inventory per workspace
`POST /admin/workspaces/getInfo` with `datasourceDetails=True`, `lineage=True`	Data source details and lineage edges (Scanner API — async)
`GET /admin/groups/{groupId}/reports`	Report inventory
`GET /admin/groups/{groupId}/dashboards`	Dashboard inventory
`GET /admin/groups/{groupId}/dataflows`	Dataflow inventory
`GET /datasets/{datasetId}/refreshes`	Refresh history per semantic model
`GET /admin/activityevents`	Usage and adoption activity log, polled as part of the Operational job

Get Started

Core Concepts

Platform

Architecture

Data Sources

Integrations

Security & Access Control

Deployment

Help

Settings

MCP

Release Notes

API Reference

Workspaces

Semantic Models (Datasets)

Semantic Model — Top Level

Semantic Model Building Blocks

Refresh History

Reports

Dashboards

Dataflows

Data Sources

Usage & Adoption

API Methods Used

​Workspaces

​Semantic Models (Datasets)

​Semantic Model — Top Level

​Semantic Model Building Blocks

​Refresh History

​Reports

​Dashboards

​Dataflows

​Data Sources

​Usage & Adoption

​API Methods Used

Workspaces

Semantic Models (Datasets)

Semantic Model — Top Level

Semantic Model Building Blocks

Refresh History

Reports

Dashboards

Dataflows

Data Sources

Usage & Adoption

API Methods Used