Rooting out the Waterfall Legacy

Computer R apid change in information technology brings with it a frequent need to undo the effects of previous culture change efforts. This process, while often challenging and frustrating, offers numerous rewards for success. Organizations can change from slow, reactive, adversarial, separated software and systems engineering processes to unified, concurrent processes. These processes better suit rapid development of dynamically changing software-intensive systems involving COTS, agent, Web, multimedia, and Internet technology. Culture changes are never easy, however. For example, in the mid-1970s, I participated in a group at TRW that created a corporate software engineering culture around the sequential requirements-driven waterfall model. The effort included corporate policies and standards, training courses administered by managers, and even stringent test on software policies. We did not realize how effective this culture change was until we tried to undo parts of it a few years later. By the early 1980s, we and other companies realized that the waterfall model was ineffective for developing user-interactive systems, in which prototypes rather than exhaustive processing specifications proved more effective in determining the most appropriate product characteristics. The “A Large Sequential-Engineering NearDisaster” sidebar provides a good example of the mismatch we found between the sequential, deductive waterfall approach and the need for emergent, synthetic approaches for user-intensive systems.