I see what you mean now. What a hack! While I was able to complete the original test with comparable stats as the first test, this “feature” didn’t work when making the schema a bit more complicated by adding DimProduct and FactSales (so we have two fact tables and three dimensions as shown in the attached file). First, Tableau didn’t detect a join between DimProduct and FactOnineSales. Second, it imports tables sequentially. Third, it froze after 3:46 min (I guess after imported all the data) and pegged CPU 90%, so I had to kill it. So, I guess this is not officially supported. https://uploads.disquscdn.com/images/16719a6f602d5d69997447096cf159102b2baf6ca4b60770143adb9823579c1c.png

just press the add button on data source once you have all the tables you want from any data source (sql, oracl, GBQ…). then choose sfdc. log into SFDC with a free developer account you have setup if your company doesn’t already use it (google that). thats it. dont add any tables from SFDC because you don’t want any of them! Tableau never hid a feature, They just didn’t get around to adding a Tableau like way of accessing this feature yet. I think there was blog post from Tamas Foldi on the nature of TDE a few years ago showing this capability.

Hi Brent,
Thanks for the tip. In this case, the data is coming from a SQL Server database. Can you point me out to more detail steps? Why would Tableau hide this feature?

TDE and Hyper have always supported normalised extracts! Tableau just never made the button in the interface… If you want to do this test against a normalised extract add a SFDC connection in the data pane (trust me) before you press extract. You can prove this out by dragging the file in as a data source and then see the individual tables.
Happy testing

No idea. I think Tableau started with the single dataset easy way and continues on that path.