BeyeBLOGS | BeyeBLOGS Home | Get Your Own Blog

« Walking on Coals | Main | Visualization »

August 4, 2007


Most of the data warehouse related projects I've been involved in that are related to a transactional system implementation, migration, or enhancement have followed the implementation of that OLTP system rather than lead or been concurrent with. I believe that's primarily because the data warehouse was also often the "reporting database" for those implementations. Part of the reason was to help justify the construction of the data warehouse - "We can off load reporting from the transactional system to the DW! The BI team can create those reports for people!" Very honorable, indeed, but those kinds of projects are often fraught with scope and requirements changes as the transactional system is implemented and people figure what it really means in the world of day to day work.

I'm in the middle of a 5 year initiative that is gutting and reimplementing the underpinnings of a all of the operational systems of a 20,000 employee organization right now. This is happening, though, because the powers-that-be in my organization decided it was part of their mission to become an "information enabled" organization. Besides implementing new and standardized applications across the company to ensure that we could collect lots and lots of information, the leadership also create a Performance Management department, which funds the Business Intelligence and Data Warehousing group. Insightful, I think.

Because Business Intelligence has such a prominent role in the organization, the implementation of reporting functionality (operational and some analytical) is happening alongside the implementation of the new ERP and billing systems. What's even more insightful, I believe, is that in some cases, there are Performance Management objectives influencing how the new systems are being rolled out - specifically in the are of master data management and conformed dimensions.

The theoretical concert between operational changes and business performance management makes a lot of sense - someone with great analytical modeling skills looks at data and determines that the business could perform better if some business process were altered. So, the business process is changed... sometimes by implementing a new application system that can help make people more productive... but that also means the business model and underlying data on which the original analysis was build changes... and the analytical model breaks. So, the data warehousing group waits for the application to go-live so they can rework their technical infrastructure to get the data and put it in some new set of data structures for the analysts to build new models on top of and go through the next iteration of process improvement. Those cycles take years to happen.

Wouldn't it make more sense if the operational applications that are built to improve business performance used a strategic analytical model as an input into what the system should look like and what kind of data it should capture and deliver back to folks doing performance management? Would it make any sense for the data warehouse to be built before the operational application, and be used as a validation that the new operational processes being implemented actually match what was designed to improve business performance? Once a business has gone through an initial cycle of process improvement, who should be pushing whom?

Posted by Paul Boal at August 4, 2007 11:15 PM


Post a comment

Remember Me?