This is a work in progress paper where authors are involved in a Telco corporation reengineering project of a half terabyte Data Warehouse (DW).Firstly the DW context is introduced. Then we discuss the actual implicit architecture found in the runtime system, identifying the problems. Finally the proposed architecture is presented. It has two views: a static, made with filters and pipes, and a dynamic based on an event driven scheduler, as opposed to a fixed time one. In our runtime scenario, the scheduler reads the pipe configurations from the metadata repository, and then executes on the fly the planed filters, transforming data into business information.The architecture metadata offers a flexible representation of data paths whose main benefits are 1.the chance to easy obtain data path traces through the whole system, and2.the ability to change those paths on the fly without pulling down the system.