Slice measure group by dimension related to another measure group

Question

Hello everyone,
My users are wanting to slice one measure group via a dimension related to a different measure group.  I have regular relationships here: Dim_SalesStatus <- Fact_Sales -> DD_Link <- Fact_Mfg, and Fact_Mfg has a many-to-many relationship
to Dim_SalesStatus  (through intermediate measure group, Fact_Sales). The DD_Link is a degenerate dimension that I created with the LinkId's that exist in the Fact_Sales table.  The Fact_Mfg records may or may not have a LinkId. Processed great,
no problem. I was testing the configuration by finding multiple LinkId's that *are* in both fact tables in the data warehouse, then seeing if I can slice. If I slice by only the LinkId, they show the correct counts on both sides as 1.  In my following
example, I have two LinkId's (12345 and 67890).
DD_LinkId, Fact Sales count, Fact Mfg count, MfgQuantity
12345, 1, 1, 387
67890, 1, 1, 50
As soon as I try to slice by the SalesStatus, the Fact_Mfg count goes to null on some (not all). It seems like it can’t find the data through the relationship.
Sales Status, DD_LinkId, Fact Sales count, Fact Mfg count, MfgQuantity
Incomplete, 12345, 1, (null), (null)
Incomplete, 67890, 1, 1, 50
I'm not sure what could be the issue, it looks to be all setup correctly to me.  LinkId 12345 is not behaving the way I would expect, but 67890 is, and I'm not seeing any differences in the way the data is setup in the DW.  What can I check/verify?
Thanks in advance for your help!
Jarret

Answer 1

Hello Bill,
Neither of those are true.  The FactMfgCount is the default measure added when creating a measure group, and the quantity is a simple column based summed measure.
If I drop the DateEntered dimension altogether, the results returned are shown as expected.  When I then recreate it, with a regular relationship to both measure groups, it stops working.  If I remove the relationship from DateEntered to Sales,
the expected results are presented.  Or, if I remove the relationship from DateEntered to Manufacturing, the expected results are presented.
From what I can tell, as long as a dimension is not related to the fact on both sides of the many to many, the expected results are presented.  As soon as I relate a single dimension to both sides of a many-to-many, the data in the data warehouse must
be the same in the related columns, or the results are "(null)".
Is this supposed to be true?  If you look at my cube structure image above, that's my setup.
Here is the MDX that I was running, which does not reference a date dimension.
//Without Status, expected results.
SELECT NON EMPTY { [Measures].[Fact Sales Count], [Measures].[Fact Manufacturing Count] } ON COLUMNS,
NON EMPTY { ([DD_Link].[Order Link Id].[Order Link Id].ALLMEMBERS ) } ON ROWS
FROM ( SELECT ( {[DD_Link].[Order Link Id].&[18440402] , [DD_Link].[Order Link Id].&[18446846]} ) ON COLUMNS
FROM [TestCube]
go
//With Status, unexpected results.
SELECT NON EMPTY { [Measures].[Fact Sales Count], [Measures].[Fact Manufacturing Count] } ON COLUMNS,
NON EMPTY { ([DD_Link].[Order Link Id].[Order Link Id].ALLMEMBERS * [Sales Status].[On Time Delivery Status Description].[On Time Delivery Status Description].ALLMEMBERS ) } ON ROWS
FROM (
SELECT ( {[DD_Link].[Order Link Id].&[18440402] , [DD_Link].[Order Link Id].&[18446846]} ) ON COLUMNS
FROM [TestCube]
go