How evolution of information systems may fail: many improvements adding up to negative effects

It has been observed in case studies that computer systems evolve through three main processes: initial development, adaptive maintenance, and replacement. Often one system replaces several existing ones. Models of system evolution should therefore include the these processes for all the systems being affected. Sometimes, the cumulative effect of small improvements made during adaptive maintenance was dysfunctional; for example, recurrent changes produced spaghetti code which no one dared to change. Possible negative effects of many small changes can be foreseen through improved planning and organization of maintenance. This indicates that systems should be replaced before they deteriorate through amendments. Replacement seems to be easier to carry out than is initial development, due to easier requirements engineering. In fifteen out of sixteen processes where computer systems were replaced with newer ones, the new systems were replicas of the old systems with some functionality added. Through repeated replacements, the organizational structure is reinforced. Repeated replacements may therefore result in an inefficient and rigid organization in the long run. Thus many easy-toaccomplish replacements of program code, each which were intended to improve efficiency, may cause the counter result in the end.