I ahve a cube which measures the Cost per visit and analyses the paid
amount by an insuarance company by different dimensions.I have data
about the membership of the members of the company which is at a
monthly level and not at a visit level.I divide the number of memebr
months by the number of visits to get the months per visit and
allocate it to the record at the fact level to get the member months
per visit at the visit level for that visit.

To get the right allocation of member months whcih have a dependency
on the region and the Line of Business and the month on which the
Member months were calculated i calculate teh member months for that
combination and didvide it by the number of visits for that
combination.

So far I am bale to get the correct numbers.

Now when I am also asked to account for the Doctor or provider , I
find that there are some providers which have members but to whoom
there have been no vists for that combination of incur date , Line of
business and Region .

So the Number of visits are null or zero for that combination and the
membermonths divided by vists are very very large.Or infinity....


Any idea how I can solve this problem?

Would creating 2 fact tables with different grain be the answer...what
would happen to the measueres which depend on other measures and we do
not have any values for the denominator.....


Has anyone come across such problems......


Please let me know....


Thnaks in advance....



Ajay