Universal Credit - an IT experiment
It's an ambitious scheme to transform the benefits system but Universal Credit has begun to look like a classic example of how big IT projects go wrong.
A year ago the National Audit Office described the project and in particular its technology as beset by "weak management, ineffective control and poor governance".
Now the NAO has produced a follow-up report which warns that there is no guarantee that the programme will ever deliver value for money. But what this project is delivering is a valuable live experiment which may tell us how public sector IT schemes should be run in the future.
Right now, there are two different IT systems designed to deliver Universal Credit. The first, called the "live service" in the NAO report is the original plan which involved handing the job over to large contractors. This is what went so disastrously wrong, forcing the Department for Work and Pensions to "reset" the programme last year.
The second is what's called the digital service, designed to be a far more lightweight and nimble system created by bringing software skills into government rather than relying wholly on contractors. This looks much cheaper - its budget is £108m compared to £657m set down at the start as the cost of the original project.
But here is the problem. Instead of just writing off the original investment and concentrating on the new digital service the government is running a twin track approach. The "live service" is, well, live, and continuing to be used with claimants - although only around 18,000 in October compared with the two million the DWP projected to be on Universal Credit by now back in the winter of 2012.
Meanwhile, the digital service being developed in house is already six months behind schedule, partly because of difficulties in recruiting people with the necessary skills. The service has reached the beta stage which means it will now be tested on a few hundred people. The new target is that within eighteen months it will be fully operational, eventually reaching as many as 10 million claimants.
In the meantime, more money is still being spent on the old IT than the new service - the total bill so far is £344m and over £100m more will be needed to keep it going until the digital service is ready in 2016.
But the National Audit Office seems uncertain that even this target will be hit, and warns that further delays could prove very costly to the overall aims of the project, cutting the societal benefits it is supposed to deliver by £2.3bn.
The NAO says the twin-track approach has its risks and Labour's Margaret Hodge, who chairs the Public Accounts Committee, accused the government of throwing good money after bad.
But from an accounting point of view you can see its attractions. So far, just £34m of the £344m spent so far has been written off by the Department for Work and Pensions. However ineffective the old IT system might be, elements of it are still being used - and the longer that goes on, the smaller the eventual writedown will be.
The DWP continues to look on the bright side when it comes to examining an IT project that has received plenty of brickbats. A spokeswoman gave me this statement: "The investment in IT has a value that hugely outweighs the costs as it's being used, day in and day out, and will continue to be used even as we start to test an enhanced digital solution from this week."
But here's the thing - if the new digital service does work and arrives on schedule, the government will end up spending around £100m less than the original budget of £657m. It will also have proved that the old way that this government and previous ones administered big public sector IT projects was costly and ineffective. So shouldn't it be getting ready to acknowledge and celebrate that?