I have just finished a data warehouse project where the design was completed and signed off my the users responsible for the business areas involved. The key point was that they needed to understand what they were signing and if they didn't, then those items were not permitted to be deployed for development.

Practically, this meant that all tables and columns had meaningful business descriptions which everyone (IT professionals and our users) could understand.

All design work was done in a database agnostic tool (in our case Enterprise Architect by SPARX System). The meta data associated which all tables and columns was included in the DDL generated from the design repository.

The meta data was combined with the actual usage of each column (in cubes, reporting services models and reporting services reports) and this was stored in a single enriched meta data repository - a set of tables used to stored this info and the links between each item. This made finding where terms (business phrases, columns etc) were used so easy.

This was a major point of distinction between this project and most others I have been involved with. The key point being that the meta data was central to the process and was always considered to be a key deliverable from the project.