The quality of Business Intelligence (BI) depends mostly on the available information behind it. This is particularly evident given such functional improvements in the latest generation of BI tools as much better reporting and analysis functions, new visualization capabilities, dashboards, portals, data mining tools and new analytical applications like business performance management, forecasting, planning etc. Unfortunately, business users are still dissatisfied by the business information they receive every day. What the business needs is consistent, accurate, accessible, actionable and transparent information across the whole company.
Today's businesses also need flexibility and responsiveness, so that when they want to extend or change their BI either because they have a different view now or because of an actual business change this should be quickly achieved without any compromises on the quality of BI.
Why is this so difficult? There is no doubt about the importance and value of BI, and we have made an enormous progress in this field. At the same time, we are still struggling with issues like inconsistency between two reports that we received in the same morning, arguing over the completeness of data, or dissatisfied that this was exactly what we needed. There are times when we want to trace back what changes have caused our figures to appear so different this month or to audit whether our business rules are consistently applied.
Data quality is the usual suspect, but equally, we should also question the effectiveness of our data integration in data warehouses and marts. It is difficult to achieve total integrity when the business rules used for derived data are spread across multiple ETL strings as on-the-fly transformations or applied as last minute transformations in the fragmented BI components.
IT has rightly turned to enterprise data warehouses and master data management (MDM) to fix these issues. Regardless of whether they are sitting on a highly scaleable platform, implemented as a central or federated architecture, or designed in normalized or dimensional models, most of enterprise data warehouses and the data marts derived from them fail to deliver fully on their promises to the business.
This paper analyzes the issues of conventional data warehouse design process and explains how this practice can be improved using a business-model-driven process that covers not only the requirement definition and design, but also the development, operation, maintenance and change requirements of the data warehouses and data marts in support of effective BI.