I just posted this bug on the Microsoft Connect site. Please vote it if you think it would be a useful change. This is the issue description.

If you set the”Aggregation Usage” to true for an attribute of a dimension (i.e. Customer), you cannot run the aggregation design wizard on a partition of a measure group that is not linked to that measure group.
You receive an error message, but it would be preferable a Yes/No dialog box that allow the user to continue considering an “Aggregation Usage” set to “Default” for that dimension only for this measure group.
Actually you can workaround this issue in two ways:

  1. Changing the Aggregation Usage before/after using the aggregation design wizard (setting the desired value for the measure group you want to optimize)
  2. Manually designing the aggregation (not a very comfortable practice)

I understand it is not a real bug, but the message is not coherent with what you can do anyway by changing aggregation usage setting.

I’m working on optimizing aggregations in these days and this limitation has been very annoying today, requiring me to write a set of specification just to explain how to redesign aggregations if in the future someone will make a structural change to the cube…

Originally appeared on: http://sqlblog.com/blogs/marco_russo/archive/2007/04/27/aggregation-usage-issues-with-a-measure-group-not-linked-to-a-dimension.aspx