Scaling BI Pixie

Scaling the BI Pixie Dashboard

Background

The BI Pixie Dashboard is a Microsoft-certified Power BI Template App that lets you analyze your Power BI reports through the lens of adoption, engagement, user satisfaction, business outcome, performance and security. Behind it, BI Pixie collects usage telemetry from every report with Pixies and stores it in your own storage, and Power BI loads that telemetry into the dashboard's semantic model for analysis.

BI Pixie comes in several deployment variations that process this telemetry and make it ready for analysis in the dashboard. As your organization generates more usage data, each refresh of the dashboard has more data to load, and how that data is loaded is what you scale. The most scalable variations use the BI Pixie ETL, a Spark job that transforms the raw telemetry into compact, pre-aggregated tables for the dashboard. The ETL runs the same logic on Microsoft Fabric and on Azure Databricks.

Scaling step by step

There are several ways to deploy BI Pixie Dashboard. If you start with the Power BI app, the dashboard's semantic model imports usage data through Power Query. If you run the dashboard in a workspace assigned to a Power BI Pro license, a refresh can reach the two-hour refresh timeout when you have a high volume of usage events. When a refresh reaches that timeout, work through these options in order:

1. Limit the data range. Change the Last N Days parameter from -1 to 30 to import only the last 30 days, then refresh again.
2. Scope the heatmap. If you activated the heatmap, turn it on only for the specific reports that need it.
3. Upgrade Power Platform → Azure. If you run BI Pixie on the Power Platform, move to the Azure deployment. On Azure the dashboard imports data far more efficiently through the Azure Storage Account; the Power Platform version's Dataverse connector is intended for low usage volumes.
4. Optimize performance parameters. On the Azure deployment with the Performance feature enabled, lower the Last N Days Log Analytics parameter (default 30) to reduce refresh time.
5. Use the PBIX semantic model. Customers with source-code access can refresh the provided PBIX file outside the Power BI Pro service limit.
6. Assign a Fabric capacity. Assigning the BI Pixie workspace to a Fabric capacity raises the scheduled refresh time-out from two hours to five hours. A refresh you start through the Power BI enhanced refresh API can run for up to 24 hours, including retries. A Fabric capacity also lets you use incremental refresh for selected tables, or split the load across multiple dashboards filtered by project or workspace.
7. Move to the ETL Notebook + Direct Lake (most scalable). With a Fabric capacity, through the BI Pixie Workload or a self-hosted deployment, you run the BI Pixie ETL as a PySpark notebook. Each run ingests only new telemetry into a Lakehouse, with Spark compute that scales to years of data. Power BI reads the Lakehouse into the semantic model through Direct Lake, so there is no import step and no Power Query refresh, and no refresh timeout, whatever the data volume or retention period. This is the default architecture for the BI Pixie Workload's Direct Lake dashboard (deployed for you automatically) and the recommended target for self-hosted customers who have a Fabric capacity. Contact support to enable it for your environment.
8. Run the ETL on your own Azure Databricks (no Fabric required). If you deploy BI Pixie on Azure but do not have a Fabric capacity, the same BI Pixie ETL can run on Azure Databricks, reading your BI Pixie storage account directly and writing compact, pre-aggregated tables so the dashboard stays small and refreshes within the Power BI Pro limit even at high volume. You connect the same report and semantic model to Databricks in Import, DirectQuery, or Direct Lake mode depending on your license and freshness needs. It suits you best when you already run Databricks, and our team delivers it as a professional-services project rather than as a self-serve option. Learn how BI Pixie on Power BI over Azure Databricks works, then open a support ticket to scope it.
9. Professional services. If the options above don't meet your requirements, open a support ticket and we can offer paid training or professional services to build your analytics solution on the data BI Pixie collects.

We recommend that you open a support ticket, so that our team can guide you to the best resolution for your environment. The following table summarizes the options and support coverage: