A Model Driven Approach to Enterprise Data Integration

Enterprises have a lot of data locked up in their operational data stores, warehouses, documents and so on, from which it should be possible to extract a lot of useful knowledge that can help them plan their businesses better and serve their customers better. However the success of this depends entirely on the quality of data available, and unfortunately in most organizations there is very little confidence in the quality of their data assets. There are many reasons for this. A large enterprise has 1000s of data stores that are produced in different contexts under different assumptions to serve different purposes. There is uncertainty as to exactly what kind of data exists in each of these sources, what their semantics are and what kind of quality problems exist. There does not exist a single, unified, enterprise wide view of data. Traditionally the practice has been to build ad hoc, purpose specific solutions on a need to do basis, such as building a purpose specific warehouse. However, such point solutions proliferate over time, exacerbating the problem further, as they end up adding additional sources of inconsistency. We are working on a model driven approach to address some of these issues.