Dimension Attribute not related to the key directly or indirectly.

Question

I am running into an issue when trying to relate two fact tables to a dimension using two different attributes.  Due to the design of the tables I must join from my fact tables to intermediate tables in order to join to the dimension table.  The
problem is the fact tables use different intermediate tables to join to the final dimension table.
For example: FactTable1-->IT1-->DimensionTable<--IT2<--FactTable2
The  intermediate value from the first table is a foreign key to the dimension table and the intermediate value from the second table is also it's own foreign key to the dimension table.
Below is some test code to put my problem into SQL.  It's simple enough to join them together but I cannot for the life of me figure out how to properly set up the relationships between attributes in SSAS.
CREATE TABLE dbo.FactTest1(
FactTest1Id Int Identity,
IT1Id INT)
CREATE TABLE dbo.IT1(
IT1Id INT IDENTITY,
DimId1 INT)
CREATE TABLE dbo.Dim(
DimId INT IDENTITY,
DimId1 INT,
DimId2 INT,
Description VARCHAR(50))
CREATE TABLE dbo.IT2(
It2Id INT IDENTITY,
DimId2 INT)
CREATE TABLE dbo.FactTest2(
FactTest2Id Int Identity,
IT2Id INT)
INSERT dbo.FactTest1
SELECT 1
UNION
SELECT 2
UNION
SELECT 3
INSERT dbo.IT1
SELECT 10
UNION
SELECT 20
UNION
SELECT 30
INSERT dbo.FactTest2
SELECT 1
UNION
SELECT 2
UNION
SELECT 3
INSERT dbo.IT2
SELECT 40
UNION
SELECT 50
UNION
SELECT 60
INSERT dbo.Dim
SELECT 10,40, 'Atrribute1'
UNION
SELECT 20,50, 'Attribute2'
UNION
SELECT 30,60, 'Atrribute3'
SELECT *
FROM dbo.FactTest1 ft
JOIN dbo.IT1 it
ON ft.IT1Id = it.IT1Id
JOIN dbo.Dim dim
ON it.DimId1 = dim.DimId1
SELECT *
FROM dbo.FactTest2 ft
JOIN dbo.IT2 it
ON ft.IT2Id = it.IT2Id
JOIN dbo.Dim dim
ON it.DimId2 = dim.DimId2

Answer 1

It depends on the cardinality between the two Fact Tables and intermediate tables. If they are 1 to 1  relationships then you can simply create a view of the fact and intermediate table and then create a direct relationship to the dimension. A more
efficient solution would be to alter your ETL to just insert the appropriate dimension key directly into the fact table instead of using the intermediate ID.
You can join fact tables to dimensions at different levels of granularity. eg you could join one fact table to a date dimension at the day level and another at the month level. You just have to consider what you want to happen to the monthly data when viewed
at a daily level, you can either repeat the monthly value or display a blank or look at overriding the value with some sort of calculation.
If the relationship between the facts and intermediate tables is many to many then there is an excellent white paper called the "Many to Many Revolution" which shows all the different ways in which many to many relationships can be used in SSAS.
http://darren.gosbell.com - please mark correct answers