The three-layer architecture that is used to build a data warehouse consists of:

1st layer - data in various operational systems

2nd layer - a data warehouse acts as a centralized single source of data

3rd layer- data marts that built on top of the 2nd layer data warehouse but with a limit scope to particular user group. Sometimes data marts can be independent of the data warehouse or a subset of the data warehouse.

Please help me explain why it's appropriate to have the three-layer architecture. What are pros and cons?