At Microsoft TechEd a few weeks ago, I sat at random tables at lunch and asked attendees for their thoughts about certifications. The results of my unscientific poll of a small sampling of TechEd participants is that certification doesn't matter. Wait—before you cancel your subscription to this newsletter, let me explain.

TechEd attendees included a preponderance of developers, which isn't surprising for a conference that focused on Visual Basic (VB) and Microsoft's new .NET Framework. Traditionally, developers haven't embraced certification as vigorously as people who work on networks or manage databases. For that reason, developers usually hold fewer certifications than do other IT workers. When I raised the topic, developers' responses ranged from "certification isn't necessary for my job" to "I have X years of programming experience, so I don't need a certification." The certifications' reputations must play a part as well, because I've heard of many highly paid, multi-certified consultants whose programming skills failed to impress the staff programmers they were trying to help. In all, programmers appear to have more faith in results than exam scores.

All but one of the 20 people I surveyed agreed that certification is good for landing entry-level positions, but none believed a certification to be worth more than a bachelor's degree or a couple years of job experience. Those sentiments held true for the MCSD certification especially, although respondents' reasoning appears to be based on the perception that the MCSD exams don't test the skills that programmers use in their work. Several people complained that exams pose many questions about obscure functionality that rarely applies to real applications. Others offered the usual programmer's answer of "if I need to know how something works, I look it up in the documentation." Unlike the MCSE candidates to whom I've talked over the years, programmers don't typically see a connection between certification exams and their jobs.

My results are hardly scientific, but they nevertheless cause me concern about the state of Microsoft's certification programs. First, both the MCSE and MCSD certifications are losing credibility. Many people blame the brain-dump sites and practice exams, but the real blame falls squarely on the shoulders of those who have subverted the exams' purpose by cramming instead of acquiring the skills that the exams seek to measure. Everyone I've asked has said that companies want to hire competent people, not good test takers. All of us, therefore, need to do a better job of encouraging those seeking certification to use the tests to demonstrate what they know and not what they've memorized.

Second, a disconnect seems to exist between the skills Microsoft thinks it should measure and the skills that workers use, especially for the MCSD exams. For example, I remember a question from the VB 3.0 exam that asked about the Autodraw property on a Picturebox control—a reference that's not important for those of you who are MCSEs because it's obscure even to VB programmers. I was baffled, and I'd been using VB intensively for more than a year when I took the exam. The question counted for about 2 percent of my exam score, and it had nothing to do with anything I had ever used in my work. An exam with too many of that kind of question can mean failing scores for people with years of experience or passing scores for people who have learned things that they'll never use on the job. Neither result is good for the certification.

Thousands of people are deciding whether to take the Windows 2000 MCSE exams or lose their certifications altogether. They face the question of whether the certification is relevant enough to their jobs to make the effort worthwhile. Many who landed their first jobs because of a certification now have years of experience working with Microsoft products, and they probably don't need certification to keep their jobs or to get another one. Many others will continue to use Windows NT 4.0 for years to come. Those people will have a difficult time gaining enough real-world experience to pass the exams and might have to resort to cramming. Once again, neither option is good for the certification.

I've never supported Microsoft's decision to eliminate the NT 4.0 MCSE certifications, but that's the reality we face. When the .NET Framework hits the development world next year, I expect that MCSDs will face a similar reality in 2003. What worries me most is that too many people will decide that the MCSE (or MCSD) doesn't matter enough to endure the long and expensive process of updating it. The question then will be, "if we've lost our most experienced people, does certification really matter?"