Dataflows Question

This is really more of an open ended question than I need help with a specific thing or scenario.

 

We have users who are looking to "mix and match" data from across several of our enterprise tabular models, without exposing the underlying data warehouse to thousands of users for them all to refresh Customer, Sales, Item, etc. countless times a day.  Was thinking that using dataflows to provide users access to the data while also controlling/limiting the overhead on Synapse.  Spent the last couple days kicking the tires on them...set a couple up and waited.  Waited a really long time for them to refresh from Synapse.  And then waited again attempting to import data from them into Power BI desktop to author some sample reports.  The underlying views were fairly large but not unduly so (25M or so records on the fact tables)

 

The amount of time spent waiting has me believing that dataflows really aren't much use for anything.  But they do serve a purpose (I really believe they do!).  I'm inclined to believe this comes down to operator error on my part.  Like I missed something fundamental in configuring them that would have drastically increased performance.  The only thing I didn't configure was incremental refresh (will do tomorrow).

 

Can anyone offer some pointers on how to improve performance for refreshing a dataflow?  And also how to get desktop to import data more quickly?  My users will never accept these as a possibility like this and I can't say I'd blame them.

 

Thanks in advance!