Monday, November 1, 2010

$1B software project? For real?

The US Government routinely spends $1B or more on software projects, which almost always fail to deliver anything close to that value. In a stunningly sane move, OMB halted $20B of "financial system replacement" projects this summer. Perhaps to show they mean business, again this fall OMB and Vivek Kundra began detailed inspections of 26 troubled high-ticket IT programs worth $30B. On this list the poster child of offensiveness is the $4.7B ACE modernization project, for which IBM has been paid $2.7B to date in exchange for exactly bupkis.

High risk, high value projects can be justified, even at high cost. The US effort to ramp up its wartime footing in the 1940s, the Manhattan project, or even the race to put a man on the moon. But where do we have evidence that $1B software projects for routine IT usage have ever made any sense?  At 260 people the avionics software team for the space shuttle* is large -- and has probably cost a billion dollars over it's 40 year lifetime -- but is a rare exception.

The literature in software projects clearly documents that  risk rises much faster than team size. Ideal team sizes are measured in the 10s not in the hundreds. Even for multi-million line systems, tens of people over a decade are far less risky than 100s of people over the same time period.

How do these flagrant wastes of money and opportunity come about? A demanding user base that doesn't have to pay, a swarm of large IT service mercenaries who make a profit on "asses in seats," business school case studies hawking governance to guarantee success, a  blind congress to foot the bill,  clueless government management, and cause and effect so distant in time that there is no accountability.

* Fishman, Charles, "They write the right stuff" in Fast Company, Dec 31, 1996.

No comments:

Post a Comment