BI Pixie Workload in Microsoft Fabric
Public Preview

OneLake and Storage

BI Pixie in Microsoft Fabric stores your telemetry in your Fabric capacity's region and integrates with OneLake so you can use the data alongside the rest of your Fabric estate. This page describes where event data lives, how to expose it to other Fabric items via OneLake shortcuts, and how lineage and sensitivity labels apply.

Where Your Event Telemetry Is Stored

When a report tracked by a BI Pixie item receives an event — a page view, a visual click, a filter selection, a survey response — the pixel posts the event to BI Pixie's intake endpoint. The event is buffered through Event Hubs and written to a Fabric-aligned data lake bucket scoped to your tenant and workspace.

  • Region. Telemetry is stored in the same Azure region as the Fabric capacity that backs the workspace where your BI Pixie item lives. Move the workspace to a capacity in another region and new telemetry follows.
  • Format. Events are persisted as partitioned Parquet under a date hierarchy (yyyy/MM/dd) for fast time-range queries.
  • Scope. Each BI Pixie item has its own logical container. Telemetry from one item in Workspace X is never co-mingled with another item in Workspace Y.
  • Retention. Default retention is 30 days on Standard and 365 days on Pro. Enterprise extends retention up to 1095 days (3 years). See Data Management for the full retention matrix.

OneLake is built on Azure Data Lake Storage Gen2 and uses the same APIs as the lake bucket BI Pixie writes to. That shared substrate is what makes the OneLake integrations below possible.

Exposing Telemetry via OneLake Shortcuts

A OneLake shortcut is a pointer from a Fabric item — typically a lakehouse — to data stored in another location. With a shortcut to your BI Pixie telemetry container, you can query events in a Fabric notebook, build a custom semantic model, or join telemetry with other lakehouse tables — without copying or duplicating data.

Create a shortcut from a lakehouse

  1. Open the BI Pixie item and select Data Management from the sidebar. Under OneLake, copy the Telemetry container path. This is the ADLS Gen2 URL of your item's data folder.
  2. Open the lakehouse in your workspace where you want to access the data.
  3. In the lakehouse explorer, right-click Files (or Tables for Delta data) and choose New shortcut > Azure Data Lake Storage Gen2.
  4. Paste the container path from BI Pixie. Authenticate with the same Microsoft Entra account you use for Fabric.
  5. Give the shortcut a meaningful name, for example bi_pixie_events, and click Create.

The shortcut now appears under Files in the lakehouse. You can read it from a notebook with Spark, query it from the SQL endpoint, or build a Direct Lake semantic model on top of it.

Screenshot: A Fabric lakehouse explorer with a bi_pixie_events OneLake shortcut listed under Files.

Read the data from a notebook

In a Fabric notebook, the shortcut behaves like any other lakehouse Files entry. PySpark example:

df = spark.read.parquet("Files/bi_pixie_events/")
df.createOrReplaceTempView("events")
spark.sql("""
  SELECT report_id, COUNT(*) AS sessions
  FROM events
  WHERE event_type = 'page_view'
    AND event_date >= date_sub(current_date(), 30)
  GROUP BY report_id
  ORDER BY sessions DESC
""").show()

Use it in a custom semantic model

Once the data is reachable through the lakehouse, you can build a Direct Lake or import semantic model on top of it. This is useful when you want to combine BI Pixie telemetry with operational data — for example, joining session counts to your CRM table to see which reports are used by your top accounts.

The BI Pixie Dashboard is the supported, packaged way to consume telemetry without writing your own queries. Use OneLake shortcuts when you need analyses the dashboard does not cover.

Data Lineage

A BI Pixie item participates in Microsoft Fabric's lineage view. Open Lineage on a workspace and you will see the BI Pixie node connected to:

  • The Power BI reports it is currently tracking (downstream from the report).
  • Any lakehouse that has a OneLake shortcut into the item's telemetry container (downstream from the BI Pixie item).
  • Any semantic model and report built on top of those shortcuts.

Lineage helps you answer questions like: "If I rename this report, what dashboards downstream are affected?" Use the standard Fabric lineage tools to explore the graph.

Sensitivity Labels

BI Pixie items support Microsoft Information Protection sensitivity labels. Apply a label the same way you would for a lakehouse or semantic model:

  1. In the workspace, open the BI Pixie item's context menu and choose Apply sensitivity label.
  2. Pick a label from your tenant's published label set (for example, Internal, Confidential — Finance).
  3. The label propagates to telemetry exports, OneLake shortcuts that pull from the item, and downstream semantic models created from those shortcuts.

Encryption-protected labels (those that enforce rights via Microsoft Purview) are honored on exports and on file downloads from the Events tab.

What's Available Now vs. Coming in GA

BI Pixie in Microsoft Fabric is being released in stages. Where a feature on this page is not yet shipping, we say so plainly.

Available now

  • Telemetry stored in the Fabric capacity region that backs your workspace.
  • OneLake shortcuts from a lakehouse to the item's telemetry container.
  • Reading telemetry from notebooks via the lakehouse Files path.
  • Lineage participation for the BI Pixie item and its tracked reports.
  • Apply, view, and export with sensitivity labels.

Coming in GA

  • Native OneLake catalog entry. Telemetry will be discoverable directly from the OneLake catalog without first creating a shortcut. We are completing the catalog manifest as part of certification.
  • Direct Lake-ready Delta tables. Telemetry is currently written as partitioned Parquet. We will publish equivalent Delta tables alongside Parquet so Direct Lake semantic models can consume telemetry without a Spark conversion step.
  • Lineage for shortcut consumers. Lineage edges from a downstream lakehouse and any semantic model built on a shortcut will surface in Fabric's lineage view automatically. Today this is a manual step in the lineage tool.
  • Label inheritance from workspace defaults. Sensitivity labels are applied per item today. We are wiring up workspace-default inheritance so newly created BI Pixie items inherit the workspace's default label.

What's Next