Not long ago, the university installation I manage was asked by one of our control agencies to provide estimates of conversion time (hence cost) for a planned major upgrade of our systems. I took a deep breath, swallowed hard, and delegated full responsibility to our assistant director with strict instructions not to show me what he did. He is very good, and so I know that he will have carefully reviewed the 300 or so major programs we support, tried to assess their complexity, level of documentation, file structures, and so on. The resulting hodgepodge of subjective judgments and wild guesses would be painful to think about, even if the programs weren't in a constant state of flux. Fortunately, he is also very shrewd, and so I know how he must have arrived at our estimate. It would have gone something like this: Well, if I come up with, say, four years, people will panic and decide we can't afford to upgrade at all, and we will wind up procrastinating indefinitely with other things getting progressively worse. If I come up with an estimate of four weeks, our users will expect instant results and wonder what we've been doing all these years; we'll get the upgrade sooner, but then we'll be in serious trouble coping with every little glitch that pops up. The question then is, what is the maximum time we can allow to iron the wrinkles out of the conversion without placing the upgrade itself in jeopardy? Now, it would be "unprofessional" to estimate conversion time by licking a finger, holding it in the wind and saying, "Oh, about four months." However, once you've decided what the traffic will bear, the entries in an elaborate tabular arrangement of programs and files classified by lengths, complexities, and special difficulties, can be assigned appropriate coefficients to be multiplied, added, etc. Lo and behold, the final result is four months. (Of course, footnotes and disclaimers will reveal that conversion time and effort may vary for individual applications as scheduled use and unforseen problems demand.)
/lp/association-for-computing-machinery/the-components-of-conversion-cost-BpauoAdnKL