Generic DW

Many years ago I worked for a company (in some ways, through the vagaries of mergers and divestments, I still work there) that had a commodity view of BI. The big plan was to implement a generic database on a Tandem Non-Stop machine and offer a ‘blackbox’ service to host data and provide reporting to a multitude of customers. Of course there were flaws with this approach. Some related to the technologies selected for the solution and a lack of suitable tools to build a performant data warehouse but in the main the generic model had too many design compromises to give customers what they wanted to buy, a one-stop solution that worked.
Although the DW services we offer now are tailored to a customer’s needs and cover a variety of technologies there is still a role for black boxes in the systems that we design. Instead of single solution that inputs business fact at one end and churns out reports at the other we are now talking about a chain of components that operate on a data stream. Each component can be described in abstract terms: data loader, data cleanser, publisher, aggregator, or whatever. At this high-level design stage we are only concerned with the business objective of the component, the actual implementation details are unimportant. In fact for many customers the innards of a component remain a mystery forever, how it works is not as important as that it works and works well.