Roger Sessions has caused quite a splash in the IT world with his analysis of the global costs of IT project failure and the predictions of coming crisis that follow. The root of the debate surrounding his whitepaper is the claim within that IT failure is costing $500 billion per month and is increasing.
Of course this is sensational and no doubt intentionally so (when Sessions writes "…the fact is that, worldwide, we are already losing over USD 500 billion per month on IT failure" I hear an echo of Mike Myer's mother in "So I Married an Axe Murderer": "This paper contains facts. And this paper has the eighth highest circulation in the whole wide world. Right? Plenty of facts. 'Pregnant man gives birth.' That's a fact.")
The numbers are worthless (which Sessions acknowledges not two sentences after calling them "facts") but what is worthwhile is the analytic approach to the problem. Unlike other investigators of the failure rate issue, Sessions isn't afraid to verge into the difficult-to-quantify territory of indirect and opportunity costs. Although those might be the most complex part of the equation when it comes to determining the true costs of IT project failure, it's interesting that the most detailed critique of the paper that I have seen (by Bruce Webster, here) focuses primarily on the baseline determinations and less on the indirect cost issues, which should be the more thorny of the two.
Sessions point isn't so much the crisis as the cause he ascribes it to (complexity) and the opportunity he sees to avert it (moving toward simplicity). Regardless of his calculations, I find it hard to argue with this conclusion, and perhaps that should be the real takeaway.
The good news is that I don't think there is a "coming IT meltdown" or rather, if there is, we're already in the middle of it, and it's being handled reasonably successfully. The friction we see in the industry now between traditional provisioning and utility computing provisioning, between waterfall development and agile development, between JABOWS and SOA, are all evidence, to me, of a system that is slowly but steadily adapting to the fact that a failure rate north of fifty percent is unsustainable in the long term in a bubble-less economy.