In a market purely focused on using technology to improve business processes, it’s incredible [that some companies seem determined to] divide technical integration and process integration into two separate entities. The only reason for implementing any kind of integration is to change a process for the better. Therefore to claim that there are somehow two different reasons for doing it ignores where the ROI for integration – improved business processes – comes from. There are of course different layers involved in an integration architecture, with Business Process Management (BPM) at one end and connectivity at the other. However, as business process change is the defining focus and reason for connectivity, there is tremendous value in having an architecture where BPM and connectivity are designed to work together.
To also suggest that the business analyst is responsible for deciding which processes should be created as web services is worrying. The business analyst defines how a business process could and should work. Deciding whether that process is best implemented on a service-oriented or message oriented architecture is down to the IT architect. That decision should not be visible to the business analyst.
There is no doubt that web services will ease issues associated with connectivity but it is a mistake to assume that it will be the de facto technology solution for every business requirement, and indeed that it represents an end in its own right. It’s yet another example of IT industry malady: Hype-based overselling of new technologies for their own sake.
Contrary to jumping on the BPM bandwagon in order to stave off the competition, Vitria has been ‘staking its future’ on BPM since its foundation in 1994, which is why 66% of our customers exploit our BPM functionality today. The fact that many other vendors have followed Vitria in joining the BPM market is testament to the fact that combined EAI and BPM capabilities are what the market wants.
Director of Marketing, EMEA
Your article on predictive analytics technology (Future gazing, Information Age, September 2002) had some interesting success stories. I was particularly interested in the CenterParcs story – it sounds like one of the most triumphant analytical customer relationship management operations I have ever come across! However, the article fails to mention the important developments in Statistical Learning Theory, particularly Structured Risk Minimisation (SRM), which delivers a faster and more robust predictive analytics approach. The SRM technique enables a robust model to be produced automatically and balanced against both the estimation and validation data. This theory also means that data can be optimally coded for the business question. Three of the data mining vendors mentioned in the article – NCR, Sand Technology, and SPSS – have already acknowledged this approach by implementing analytics technology from KXEN, as have a number of other leading software vendors.
A key advantage of the SRM approach is that a large number of variables and cases can be analysed very rapidly. Typical examples show that, of 800 variables on a database, a model can be built with as few as a dozen variables, with the same model performance. The question is – which variables should be used? Ideally the best ones, and it should be the role of the analytical software that determines the most appropriate variables, not the analyst. In practice this means that once a comprehensive data mart is established, business users can engage in high-powered analytics for everyday questions, secure in the knowledge that the mathematical theory behind the technology produces a robust answer. Meanwhile, the analytics team can concentrate on more strategic business problems.
We would like to clarify some points made in the Prospectus article on VMware in the October 2002 issue of Information Age (VMware positions for server consolidation lift-off ). VMware’s CEO spells her name Diane Greene (we missed out an ‘e’), and we neglected to mention that Azure Capital Partners was one of the company’s investors. Additionally, VMware customer Conseco Finance moved from 135 2-way servers to five 8-way servers, not to seven 8-way servers as the article described. Most embarrassingly, though, our IBM mainframe historian incorrectly identified the VM operating system as having been pioneered by IBM in the 1980s. As any industry veteran should know, the VM operating system first saw the light of day in 1971 (see ‘VM Community: Past, Present and Future’).