Another week, another shocking reminder of the tendency of big and complex artefacts to fail. Not the space shuttle Columbia. Even after two disasters, the shuttle's record of returning crews safely to earth is better than 98%. The failure I'm talking about was more predictable and much less heroic.
In the latest in a decade-long series of reports on big government IT projects, the National Audit Office (NAO) reported that the Libra scheme to develop a single system for England's magistrates' courts had run years late and way over budget. Just as predictable were the causes. The lord chancellor's department tried to impose a standard system on different bodies without ensuring they were singing from the same hymn sheet.
When two out of three potential contractors withdrew from the procurement process, project managers pressed on. And when things started to go wrong, there was no way out, apart from throwing more money at the contractor. And so on. Weren't we supposed to have moved on from all this?
It is three years since the McCartney report established a consensus about why big projects so often go wrong. It is two years since the government's main weapon against disasters, the gateway review procedure run by the Office of Government Commerce, became mandatory.
Gateways are short, independent scrutinies at crucial stages in a project's life, especially before much money is spent. If a project fails the test, it doesn't go forward. Libra was the first case to have gone through gateway reviews. The snag was that these took place in June 2001 and February last year, too late to do any good. Peter Gershon, the former chief executive of BAE Systems who runs the Office of Government Commerce, says that as projects that have been tested by gateway reviews come on stream, the government's record on IT projects will improve.
Gershon has also drawn on the evidence created by gateway reviews to propose new rules. They were announced by Douglas Alexander, the cabinet office minister in charge of IT, in a low-key statement last month. At a time when the prime minister is staking his reputation on big IT projects, especially in the NHS, they deserve more attention. There are four new rules.
First, civil servants have to sharpen their project management skills. By June, every government department and agency is supposed to have a "programme management centre of excellence" in place. The idea is to recognise project management as a skill in its own right, not something for Sir Humphreys to pick up on the job.
Second, senior officials such as permanent secretaries or chief executives will have to give a personal, signed assurance that a project is not repeating past mistakes.
Third, no more "big bang" projects - except in exceptional circumstances. Any such exception must go before a central scrutiny group of two ministers, Paul Boateng and Douglas Alexander, Gershon and the e-envoy, Andrew Pinder.
Finally, a project's resources must be in line with its priority. Gateway reviews have shown that many supposedly top priority projects have been delegated to junior managers. This must change, Gershon says.
Where a project has the highest priority, or is a high risk, a clearly identified minister must be in charge. In theory, ministers are supposed to be in charge. But over the past 10 years of project disasters, they have been good at distancing themselves from the IT elements of new policies.
It's a brave minister who agrees to attend a system "go live". It's even rarer for one to resign when the system on which the new policy depends goes wrong. Gershon's new rules will make dodging that responsibility that much harder. They won't end IT disasters entirely: as in space travel, complexity, politics and the laws of physics are a volatile mix. But 98% success would be a good target to aim for.
