Certification can be a notable and worthwhile achievement if done right. Consider, for example, Cisco router certification, which tests not only your knowledge but also your skills. The certification process takes 3 days. On the first day, you must take an exam. On the second day, you must set up a system the way Cisco recommends in a test lab. That night, the testing staff crashes your system. On the third day, with no help from the staff, you must fix the system the way Cisco recommends. Your performance on all 3 days determines whether you achieve certification. Although parts of this exam are subjective (you must fix it Cisco's way rather than your way), the exam demands that you know the theories and have the necessary skills.
Because so many people seek Microsoft certification, Microsoft probably couldn't administer this type of exam. (To get a sense of how many people seek Microsoft certification, consider that the September/October 1997 issue of Microsoft Certified Professional Magazine reported there were 25,756 MCSEs worldwide--and that number doesn't include those who take but don't pass the exams.) But Microsoft could improve its current program. Microsoft could use sophisticated disk-imaging tools rather than written scenarios to simulate processes. For example, Microsoft could create an environment in which examinees must implement configuration procedures. Or Microsoft could present an image of a broken system and have the examinees fix it.
This approach would require more equipment, time, and testing personnel to administer than the current program, but the extra time and resources involved would be offset by the benefits gained. The resulting certification would be more meaningful to potential employers and hence more valuable to those who achieve it.