Forum Discussion
Deployment pipeline not detecting report changes
Hi nandic ,
Thank you for clarifying. Deployment pipelines are able to reliably detect changes when a report or semantic model is republished from Power BI Desktop. However, if changes are made directly in Power BI Service such as editing titles, filters, or formatting these are considered service-side metadata changes and may not always trigger change detection in deployment pipelines.
This means you might see the change indicator appear inconsistently when making edits in the service.
To ensure consistent detection in the pipeline, it is best to republish the report from Power BI Desktop before deploying.
Thank You.
GilbertQ and Olufemi7 , thank you for your responses.
Until now, deployment pipeline recognized when the report was edited in Power BI service.
This is how we have been working on one project: business analysts can do some minor changes on dev workspace (change text, update default filters) and then we deploy changes to prod workspace using deployment pipeline.
Each time, when going to deployment pipeline, the report that was edited by business analysts was marked as different between dev and prod.
Note: i agree, deployment pipeline doesn't show what are actual changes on report between dev and prod, but it detected changes. And for changes on semantic model, it also shows actual changes.
But since yesterday, it doesn't work like this anymore..
Hello nandic,
Thanks for clarifying, you’re correct.
Service-side report edits (text, filters, formatting) have historically been detected by deployment pipelines as metadata differences. If this suddenly stopped across multiple pipelines and workspaces, that’s not expected behavior and may indicate a backend/service issue rather than a design limitation.
I will recommend checking the Fabric/Power BI Service Health dashboard and, if the issue persists, opening a Microsoft support ticket with workspace IDs and timestamps so they can trace it.
If it was consistently working before and stopped “since yesterday,” that strongly suggests a service-side regression.