I'm trying to build my first really complex cube and I'm not sure if I'm
going about it the right way.

Here's my scenario:

I have sales data stored in a single fact table by retailer by store by
item. Since this table changes (it's two years of data for a single vendor)
daily for most of our vendors, I need to create a cube structure that can be
populated once and then re-processed (incrementally) each day. The store
and item dimensions also need to be changing as there are stores added and
items added to the feed periodically. My first thought is to build a cube
that's only representative of the fact table, with dimensions for the
STORE_KEY, the ITEM_KEY and the TIME_KEY. From this, build cubes with only
dimensions (for the stores and items) and then create some virtual cube with
all this smashed together. This way, I could reprocess the dimension tables
without having to reprocess the fact table.

First, is this at all the right direction and second, if so, are there any
good documented examples of this?

Thank you very much!
-Malcolm