I can't remember if you touched on this in your BI methodology, but do you feel surrogate keys should be managed in your staging database?  In most implementations, I have stored both the natural key and surrogate key within the dimension table, and use this as a means to lookup the surrogate key for the fact loads.  Some places I have read that you should keep your surrogate key architecture within the staging database and perform the lookups at that level instead of the dimension- the dimension will just  inherit the surrogate key from the staging table instead of using it's own identity column.  In practice, which is the preferred method?  Re-reading some Kimball articles, he mentions having lookup or mapping tables that contain the surrogate/natural key combinations only so that it is small and efficient and can be pinned into memory, but with todays computer power and memory this is less of an issue I believe.  What are your thoughts?