How can we make great Power BI dashboards portable instead of rewiring them for every target model? In this Proof of Concept, I demonstrate an approach where a Fabric Notebook with semPy parameterizes and programmatically re-binds field references. This allows a dashboard template to be rolled out to a different semantic model with minimal manual effort. The outcome is less a one-off report and more a small dashboard factory.
I submitted this approach to the Fabric Notebooks for Power BI – August Contest, and the entry is now published in the Notebook Gallery. The goal is to better separate presentation (report) from content (semantic model) and to make the transition between models reproducible and automated. Templates (.pbit) are valuable, but they remain too static if you want to systematically deploy across multiple models without manual adjustments for each target.
A point raised in the community discussion was: “Isn’t this what template files were invented for years ago? If table and field names are identical, you can just rebind the report—two lines of PowerShell. Or just edit the .pbit files. Wouldn’t it be better to align on standards (conformed dimensions, bus matrix, Kimball/BEAM) instead of building a technical fix?”*
My take: Yes to standards – and yes to automation. Naming conventions, conformed dimensions, and a bus matrix certainly reduce the need for rebinding. But in practice, we often deal with heterogeneous environments (legacy models, partner solutions, tenant-specific variations). In such cases, the choice isn’t “standards or automation” but rather standards plus orchestrated automation. The PoC bundles many small, manual steps into a repeatable workflow—including design rules (placeholders, home tables for report measures, Visual Calculations). This makes visual standards (e.g., variance analysis) easier to reuse across different models: it doesn’t matter whether you analyze sales variances by article, customer, or finance by cost center. What really matters is how you present variances, where you offer drillthrough or bookmarks, and how you structure navigation. By decoupling reports from the underlying model, we increase reusability and can roll out design changes consistently across many models.
Idea and approach
- A template report uses consistently named placeholders (including a dedicated “home table” for report-specific measures).
- A mapping assigns these placeholders to the target model’s fields, columns, and measures.
- A Fabric Notebook (Python +
semPy) programmatically updates the report definition – load → map → apply → validate. - Visual Calculations reduce model coupling, report-specific measures work through a defined home table, and placeholders must be unique and consistent (since
visual.jsonalso references them).
How to try it
- Place the notebook in a Fabric workspace (capacity required).
- In the configuration section, set the variables for the target model:
workspace_id,dataset_id,act_measure,bud_measure,dimension,report_name_override. - Run the notebook.
- Open the generated Monitoring Dashboard, validate it—and share your thoughts.
I’d be delighted if you check out the project, leave kudos in the Gallery, and share feedback: which edge cases should be handled next, where should the mapping logic become more robust (CSV/JSON registry?), and what scenarios would you like to see supported? This project was developed with the assistance of ChatGPT—and it is meant as an invitation for constructive community-driven improvement.
Resources & credits
- Notebook Gallery (contest entry):
https://community.fabric.microsoft.com/t5/Notebook-Gallery/Portable-Power-BI-Dashboard-Templates-with-Fabric-Notebooks/td-p/4810409 - Code & template (GitHub repo):
https://github.com/MarcusWegener/pbi-dashboard-template-notebook - Video inspiration (Patrick LeBlanc): Creative way to use semPy to update a Power BI report definition
https://youtu.be/HT_J1QiBLwA?si=MpxYYYWvGIBtUhTM - FUAM (Fabric Unified Admin Monitoring):
https://github.com/microsoft/fabric-toolbox/tree/main/monitoring/fabric-unified-admin-monitoring
Disclaimer
This notebook was created with best intentions and serves as a Proof of Concept. It is provided as-is without any warranties. Use at your own risk; no liability is assumed for any damages or consequences arising from its use.
