cube design

Suppose the following telco situation:

For each subscriber you have a monthly fact record describing the characteristics of the subscriber, together with the generated revenue for that subscriber.

Handset switches are also being registered. The fact table contains a flag indicating the switch and also the date the switch occurred. These 2 fields are only filled the month of the switch.

I would like to design a cube to analyse the handset switches. I would like to analyze the average revenue 3 months before the switch and 3 months after the switch. All this also against the other dimensions/attributes describing the subscriber.

Of course, a subscriber can switch handsets more than once.

I cannot figure out how to design a cube like this and I am thinking about creating a fact table with the 3months averages pre-defined.

I did try to create a calculated member (see one of my previoys posts), but this would have to be calculated on leave level, and afterwards this would have to be aggregated upwards, resulting in dramatic response times.

Thanks a lot for your response/advice!

Regards

Joos