A survey by analyst firm voke cites the key goals: prioritize, centralize, standardize I had a telephone conversation the other day with analyst Theresa Lanowitz, now of voke, and formerly of Borland, Taligent, Sun, and Gartner. (I knew her when she worked on JBuilder at Borland.) We were talking about a survey she recently did of mostly large organizations that have established performance Centers of Excellence (CoE).(For a while, I thought that this conversation wasn’t going to happen, as we called each other and got disconnected multiple times. After about 6 tries, Theresa called me from a landline and said “My iPhone can do everything except make a call without dropping it.” She’s doing better than I would, since AT&T cell phones typically get zero to one bar in my house.)In some ways, a study like this is a little like a study to establish that most traffic flows at rush hour: You have a good idea of what the results will be qualitatively, but until you see the numbers you don’t really know for sure what they are quantitatively. You would expect a performance CoE to work better than the usual ad-hoc, end-of-cycle performance testing, because measuring and improving performance takes some insight into the architecture and some software engineering and modeling skill, and in fact Theresa’s survey established that having a CoE helped to prioritize, centralize, and standardize performance testing across the enterprise. They typically returned their investment in 12 months and required a modest headcount: mostly under 19, and on average 11. A CoE not only improves performance, it improves quality awareness in the organization. A few takeaways for success: Design a performance CoE to be scalable, both for company growth and expanded scope Silos are ineffective The line of business is good at identifying bad performance and assigning it a cost A performance CoE can help to assess deployment and operation risks The study goes on to list several more concrete ways to make a performance CoE effective, some of which are motherhood and apple pie, and some of which are less obvious, but I don’t want to give away too much of Theresa’s study. This 21-page paper is available for subscribers at vokestream.com. Software Development