Forum Discussion
Incremental refresh on computed tables in dataflows
in that scenario there is no point in having a second dataflow, or having incremental refresh in it. (You can of course still do either if you want).
Apologies, I should have specified that I would normally be doing more than simply aggregating, and would usually be combining dimension entities, for example, from other dataflows, with the extracted fact table. It is actually recommended by MS to have a second dataflow and I have found that I have a lot less issues if I follow that practice. The only thing missing in this article is the question I am asking here about incremental refresh.
Look at it from the maintenance perspective. Any additional piece of process will require additional maintenance. They can recommend that and probably have good motives, but it adds to the maintenance.
OK, thanks, but in my experience the oposite is true. As I mentioned above, I have found that I have a lot less issues if I follow their best practice. Most of the issues I have with dataflows are due to bugs in the service, and these reduced significantly when I limited each dataflow to a single datasource, i.e. a single ODBC datasource in the staging layer, and Power BI Dataflows in the transformation layer.