Reports in the web that are authored independently of the dataset are far too sensitive to changes when a new dataset is published. I would expect the report to 'break' for removed columns, renames, etc. However, the entire report is lost if anything in the model seems to change. Background below. We have a data model with ~20 tables. Since users can edit reports in the Power BI Service, we had the idea of publishing the model by itself and then using the web for report creation (we thought by decoupling the model from the reports in PBI Desktop that we could potentially overwriting newer report changes made in the web when just a model change has occurred). However, this isn't working out. Anytime a change occurs and the dataset is re-uploaded, reports that referred to *anything* related to the table updated seem to be wiped out in the Service. Example 1: I changed Member ID back to being non-aggregatable (separate issue-I won't digress) and republished the model. A report uses a concatenated version of that column which had the aggregation setting changed. That report canvas was cleared completely in the Service. Example 2: I added 2 new calculated columns in a table. Though nothing was removed or changed, other columns in that same table were used in a report. Again, that report canvas was cleared completely in the Service.
... View more