Computer-based training (CBT) has now been in existence for over two decades. It has been implemented in both the private sector and government organizations at an exponential rate. Nevertheless, many institutions, particularly educational institutions, have not yet introduced CBT. Our knowledge of what works and what does not, as well as hardware and software advances, has greatly increased in the past few years. This paper addresses many management considerations with respect to CBT. First, we consider the generic environment in which CBT might be used and then issues that affect costs and benefits, including lessons learned by the Cognitive Engineering Design and Research Team (CEDAR) of the Los Alamos National Laboratory in its assessments. The final section gives some "how-to" guidelines on increasing the probability of successfully introducing CBT into the training environment. The underlying theme of the paper is that management should be guided by what we now know about costs and benefits in its decisions regarding CBT and fight the lure of "high tech" glitter.