Well, one could argue that data warehouse is not complete without a semantic model, which is what Analysis Services is. Your end users will most likely need data enhanced with calculations (Variance, Year to Date, etc.) so it is always good to think of it as a complete solution and not just data alone.
I would need to test it but I believe that identity is not passed to SQL Server databases (Azure or not), only to Analysis Services.
I probably need to educate myself on what Analysis Service provides; perhaps it would make it easier for end users to understand and use the data, which has been a concern.
Thanks for your input.
You may be able to get away with a power pivot data model stored on a One Drive so that you can change the model and have the model changes automatically update Power Bi. I've implemented that for some clients with no Data warehouses.
I've already started building out the DW in Azure SQL. I'm a developer/DB guy and these BI tools are new to me. The interaction between Power View, Power Query, Power BI, Power BI Designer and Power Pivot is pretty confusing, and the descriptions of what they do I find on the web aren't at the level of abtraction I find most useful. Overall, it seems that this technology is mostly geared towards peer-to-peer sharing of spreadsheets, rather than the top-down distribution of data that I'm envisioning.
The analysts I'm working with are using Tableau, but there high level encouragement to move away from this tool due, in part, to the challenge of user id management. We'd like single sign on capability so that we don't have to construct a lot of separate user ids and roles. We're already using Office 365 and would rather key access off of these existing ids.
This is high livel, but directionally accurate.
Power Pivot is the only Excel tool that works the way I think a BI should, but my notes include these problems: