Prior to the point of engaging users onto the data warehouse in any way, the data warehouse should be formally...
taken under change control. This is to minimize the possibility that the users will encounter an environment with changes that are a surprise or changes that are not well thought-out or not comprehensive to all the components affected.
Change control means more than notice. It means notice to the appropriate parties with either positive or negative assurance of the content. Changes to requirements require unanimous positive approval. Other, technical change notices are submitted giving the users the opportunity to react with disapproval of the change (negative assurance).
Those components that the user will have interface to will need change control at a user representation level. This could be the data stewards or a set of approved user representatives.
It is very important to note that initiation for changes to any of the above categories can come from a variety of sources. There should be no restriction on who may REQUEST changes to any of the components! The change approval process assures that only valid changes are ultimately accepted.
There are dependencies to consider as well. Changes to the data model almost always require changes to the ETL. Changes to requirements require changes to either the data model (and ETL) and reports/user interface, or just to the ETL.
For more information, check out SearchCRM's Best Web Links on Data Warehousing.