Exactly as the question, I have an odd one that I cannot even replicate using Superstore Sales; I am blending two separate sources one with two relationships [date (at MY), Channel] the other with three [date (MY), Channel, Region).

Each time the data is refreshed or the report is uploaded to server, Tableau deletes the date relationship resulting in in crazy calculations.

I have attached a superstore mock-up though this is using just one blended source, the secondary source provides forecast percentages the primary forecast must be multiplied by, without the date, Tableau aggregates at a higher scope resulting in 1000% calculations.

Forecast data is loaded into the primary source at company level so the department that handles new customer acquisitions only makes-up a fraction of the full company forecast is providing me with the percentages first broken down to channel level (ie acquisitions by web and by sales person etc) and by month and then an additional level atop this time at regional level and channel and month hence the two differing forecasts and the need to blend multiple sources - just to add, this is not a debate as to the best methods of providing this data.

The calc looks something like: [Company Forecast] x [New Customer Acquisitions] x [Forecast by Channel and Region]

As the date relationship is critical to the correct granularity, you can see from the mock-up that losing this relationship fundamentally breaks the report.

Any idea's as to what could be the cause of this.

We are still languishing away on both 8.1.10 for both desktop and server

Steve