I have defined a cube which uses the customer dimension.

When I initially defined the customer dimension I did not have any
attributes on the customer dimension now I go back I had a few
attributes to the dimension like Name of customer, address, region and
city,etc.

After adding these attributes I go back and create a virtual dimension
for customer name,etc now when I go back and view the data by customer
all the numbers of the measures are wrong.

I just go back to the previous version of the cube without the
customer attributes and process the cube from the same data from the
base tables the numbers look good.

It is rather bewildering why by just adding a few attributes analysis
services is reporting all wrong numbers when I remove these attributes
everything looks fine.

The same happens when I go and define a new hierarchy on the existing
dimension and try to view the data by the new hierarchy created.

I would greatly appreciate any help in resolving this. Our users are
now forced to live without the vital attributes on some of the
dimensions.

Thanks
Karen

I used to have a saying..."Virtual Dimensions are the spawn of Satan".
It's not that virtual dimensions are all bad, the issue is that they
are deceptively easy and alluring and end up resulting in bad
calculations if you aren't very careful.

The problem is that if for the attribute you are adding to a level in
your customer dimension...if there is not one and only one distinct
value for that attribute for a given key value of that level...then
Analysis Services will pick the first value it happens to find. There
is no warning message at all...AS just does this out of it's divine
wisdom. Then...if you place a virtual dimension on that
attribute...your calcs will be off as the items that happened to not be
1st will not be represented in your aggregations.

Sometimes I create a leaf level for my dimension that is keyed by the
surrogate key of the dimension table. I then add all my attributes
that may end up as virtual dimensions to that leaf level. I then mark
that leaf level as invisible as I have no need to see it. This works
fairly well, but the invisible leaf level has been known to create some
issues with some OLAP front-ends.

One simple way to check if this is your issue...is to go into the
advanced properties of your Virtual Dimension. There is a property
called "Virtual" that will be set to True...set that value to False and
rebuild your dimension. That one property change will rebuild you
dimension as a regular dimension. If the aggregation issue goes
away...then I can say with 99% assurance that your issue is the one I
describe above.