IT Failure Rates--70% or 10-15%?

T his is a closed-book quiz. Quick, now— what percentage of software projects fail? Let me hazard a guess as to your answer. Odds are, you said something like " 70 percent. " Am I right? Now, I want to analyze why you said whatever you said. A friend of mine, Nicholas Zvegintzov (the Chief Guru of software maintenance), likes to say that, if you track down the sources of the things we " know " in our field, you'll find that what we thought was a plethora of sources for that knowledge generally boils down to only a precious few. That raises the question " Where do we get our information on software failure rates? " I've seen software failure trumpeted from so many academic research papers that I had to quit counting (they tend to see a " software crisis " and then say that the research work they're describing will help eliminate it). However , if you closely examine the citations they use to support the claims of crisis, over and over again the citations boil down to one primary source, the Standish Chaos Reports. And, if you take matters one step further, the papers generally cite the 1994 report. I want to question the unquestionable status of that Standish report. That's because, you see, my own observations lead me to believe that something is terribly wrong with those Standish findings. First, there's the matter of the 1994 report. If you go back a decade and look at what it actually said, it was this—31 percent of software projects were cancelled, 53 percent were " challenged " (had difficulty meeting goals), and 16 percent were successful. Most people tend to add that 31 and 53, concluding that an astonishing 84 percent of software projects are deeply troubled. It's interesting to question whether those two numbers should be added, but at the same time, the finding that only 16 percent of software projects were successful is very damning to the field. Those academic researchers who quote those original 1994 figures have ignored the fact that Standish has updated the Chaos Report periodically. The 2000 report, for example , saw a drop in cancelled projects to 23 percent , a small drop in challenged projects to 49 percent, and a corresponding increase in successful projects to 28 percent. Still not figures to be proud of, of course. But an …