Hi,
Adding few more points to what my friends raman and suman suggested.
Always make sure to keep the related characteristic in one dimension lets say all customer related information in customer dimension like customer no.customer group and so on.
Similarly material related information in material dimension like material no.material group category and so on.
Another thing to keep in mind while designing an infocube and deciding the dimension that characteristic which has 1:n relation can be kept in one dimension and which has m:n relation should not be kept in one dimension.
For eg:You have 1000 products and 1000 customers so if you keep it in one dimension then total no of rows will 1000*1000 so that is m:n relation.
Now again it depends based on the business of the customer so based on the data you can decide the same.
Why its good to have DSO below cube ?
DSO is supposed to be the SPOT(Single Point Of Truth) which holds the complete data of ECC so at any point of time you have issues with the data then you can reconcile through it.
Hope this helps.
Regards,
AL