Power BI - Centralisation

So it looks like the future of Power BI development is in the Power BI service and that the Power BI desktop may become redundant. To move development in that direction, here are some of the key steps:

1. Attempt to create one source of the truth, using Data Flows. With Data Flows, you create containers for your tables that you will re-use for to create your data sets and Power BI Reports (but see below, you should try to move to maybe 1 or just a few data sets). Note: Queries can still be used by creating in Power BI desktop and copying the M code to a query within a Data Flow. Multiple data flows help organise the tables into manageable spaces. I'm using groupings of 4 and alphabetic ranges maybe a good way of assigning the tables. Data Flows are created in Workspaces (not available in 'My workspace'. Learn More about Data Flows

2. Use the Power Query Editor within the Power BI Service, so your data preparation can be shared with others in the organisation and re-used without it having to be replicated again and again. The Power Query is available within Data Flows.

3. Create One model of your data in Power BI desktop (note, this will be possible later in the Power BI service later using Data Marts), so watch out for that as things become more centralised. Access it via the data hub button in Power BI desktop home. You can then just use Power BI for report creation (or you could create them on Power BI service).

4. Use M Code in the Power Query Editor in the Power BI service and don't add any table columns in DAX, so these changes will be held centrally and editable in one place, rather than distributed between Power Query and DAX column additions. Learn More about M Code

5. Get ready for DataMarts, which will allow you to create a model of your data (essentially a data warehouse), from Data Flows within the Power BI Service. 

The goal is to minimize the data that needs to be refreshed, to provide one source of the truth, to allow sharing and visibility of the data cleaning and massaging, and a centralised data model that end users can create any self-service report on.