While a company always has the option to build their enterprise data systems from scratch as an active enterprise data factory, most companies have large legacy data warehouses. In those cases, it is always possible, and often desirable, to migrate to the data factory. Here are the six major steps involved in a migration approach.
- Early Win. Create an “outer defense.” Build “Output Application and Reporting” model to provide near term business value and a layer of insulation so that the team can begin to work on the “production line” of the foundation layer.
- Foundation. Incrementally build the data and process portions of the Input System. Begin building the “Input Enterprise Data Model,” populating select elements of each Subject Area as needed. Put important processing in place for functions like Change Data Capture, Keying and Data Quality.
- New Tools. Introduce analytic and multi-dimensional tools that solve specific business problems and create competitive advantage with quantitative & forward thinking analytics.
- Incrementally deconstruct the unsound structure. Unplug isolated and fragmented databases that coincide with their addition to the Input Enterprise Model. Use additional capacity freed up on existing systems to act as a reporting or application “spoke” for the Analytic Data Factory.
- Factory Operations. Incrementally build the Factory Management systems. Layer operational infrastructure and around the solution during each release, creating an effective set of controls for unified process management, workflow, dimension management and data quality.