Unfortunately the ‘federated’ term has been abused by ill-conceived attempts (mostly in the fevered imaginations of software marketers rather than actual products) involving ‘virtual’ storage. In this, the global warehouse would not store a copy of all data; but rather the data would reside only in the operational systems that created it. At the center of the spider’s web a monolithic global warehouse would somehow respond to business inquiries by breaking down the query, running it against each source system, and magically combining the result. This approach, laughable to those with serious practical experience, has been unsurprisingly shown to fail. Resolving the combined results is complex technically, and even if this can be dealt with, the distributed queries place unpredictable and hence unacceptable demands for computing resource on operational systems.

What does work very effectively, however, is a federation of data warehouses, each holding a copy of a core business model and common master data, and where each higher-level data warehouse holds summarized transactions from the level below. Common master data – for example the corporate organization charts – flows downwards from the corporate (global) data warehouse, while summarized transactions – for example, the total number of sports coupes sold in the Fiorentina outlet in Italy - flow upwards from the local data warehouses. A federation of enterprise data warehouses can satisfy both the local need for flexibility and performance, and the global need for consistency and control, with each data warehouse operating independently of all the others.

The federated concept is advantageous not only in terms of daily operations, but also where implementation is concerned. When building a federation of data warehouses that are capable of adapting to change, enterprises can start with a single project, perhaps in an individual country or division of the business, then build up to a global system, adding new data warehouses in order of priority to the business. As long as the data warehouses used are capable of adapting to change, it is not necessary to fix the final architecture of the federation in advance, so a risky monolithic project can be broken down into numerous, low-risk, high-return sub-projects.

In the federated approach, local operating units incur the cost of data warehousing, but achieve payback locally while contributing to the global view, resulting in a sense of ownership and local value. The single, monolithic warehouse approach burdens local operating units with the majority of the cost, but delivers the majority of the benefits at the global level – a project manager’s nightmare!