This is an easy one to state. Design the Data Warehouse based on the requirements of the business. It sounds so obvious. The requirements should be driven by the Reporting needs, Analytics, Dashboards, Scorecards, Ad-Hoc Analysis and Data Mining. Why state this obvious principle? In all failed DW efforts, this principle is ignored.
A better way to understand Autonomous Design is by stating what not to do. Don’t design the DW by using a relational database as the source data model. Don’t make the DW look like a denormalized relational OLTP database. Don’t allow the source data models to determine the DW data model.
This is what I mean. An OLTP application has a Customer, Product and Order table. The star schema has 3 Dimension tables, Customer_Dimension, Product_Dimension and Order_Dimension. These Dimension tables have the same data elements as their source tables, except that codes are decodes. A fact table is added called Order_Fact that has 3 foreign keys, one to each Dimension table. Not much was done to create a good data model for Business Intelligence functions. The data model should be conceived from the ground up and would support the required BI functions. This is autonomous design.