Universal Credit: the history of an IT project failure

Reading the Institute for Government’s report on Universal Credit, I was struck by two related things. First of all, the project was powered forward by people who didn’t bear any responsibility for its implementation. Whenever it ran into people who needed to care about how it would work, it hit opposition.

Lord Freud’s original skunkworks project within James Purnell’s DWP ran into heavy opposition from the Treasury and No.10 Downing St; the then Chancellor, Alistair Darling, had spent years running the DWP and the then Prime Minister, Gordon Brown, had spent even longer running a Treasury that intervened more in routine administration than any other. They didn’t think it could possibly work, not least because it required the PAYE systems in the Revenue to provide timely notification of changes in taxpayers’ circumstances to the DWP and they knew the system couldn’t do it. Having been turned down, Freud took his trade elsewhere and hawked his report around the thinktank world until the Tories’ Centre for Social Justice picked up on it.

Over the summer of 2010, it hit heavy traffic again; the Treasury was still suspicious that it could possibly work, and also saw it as expensive. This devolved into a head-on contest between George Osborne and Iain Duncan Smith to convince the prime minister, who seems to have backed away from making a decision. At this point, a third party got involved – the Lib Dems, who liked the idea in its gimmicky whiteboard purity and made it part of coalition politics. This is a really important point: one reason why IDS’s scheme survived was because the Lib Dems protected it with their veto. This was especially important because of Danny Alexander’s role on the key public expenditure committee.

This also leaked into the IT implementation. A succession of non-coding directors actually claimed to have created a new software development methodology, “a fusion of agile and waterfall”. It is hugely telling that the actual programmers were in Warrington when they were in the UK at all. They, at least, were well aware that the project was a mess from very early on, and this probably explains why The Register usually knew all about it.

Secondly, although IDS evidently had no idea how it would work or how to make it work, he did grasp an important truth about big IT projects. Donning my Consultancy Hat, I pronounce: all IT projects are business transformation projects! It is rare indeed that an organisation sets out to work in exactly the same way it used to, just with faster computers. Even if that was the declared aim, just observing how the work gets done now tends to change it.

One of the most baffling features of the whole UC disaster is why, after the Government Digital Service intervened in the project, IDS and those around him insisted on continuing work on the original, failed, version of the system at enormous expense when they admitted publicly that they would only bin it once the new one was ready. The IfG report is completely clear that the minister wanted to get something of UC deployed into the Jobcentres because it would change how they worked. Whether the software worked or not, it would still enforce the new rules and the new, punitive attitude to the claimant.

Notably, it would enforce the principle that payment could only be made monthly, after a mandatory seven-day wait, which plus the target processing delay meant that anyone paid weekly would have to wait six weeks for any help in the best-case scenario. Tellingly, nobody in the IfG report is willing to take responsibility for this. The author, Nicholas Timmins, uses this in his foreword as an example of something that was “neither left-wing nor right-wing” (page 8 in the report). He also says, in the body of the text, that it was meant to “accustom the recipients to financial discipline”. (This is page 33 of the report.) How he can believe this and yet believe that it was an ideologically neutral decision baffles me. But then he also says, on page 8:

The simpler it is, the more it will involve forms of rough justice

If that isn’t the remark of someone who loved UC as a beautiful abstract scheme, without a thought for the people who might be on the receiving end of the “rough justice”, and who crucially took no responsibility whatsoever for its implementation, I don’t know what it is.

Anyway, switching target back to people who certainly were responsible for it, it’s clear that somebody thought that just the fact of being paid a monthly salary made you more Protestant, in a word. The implicit churchiness points like a radar beam towards the minister. Traditionally, the welfare state used precisely this distinction between weekly wages and monthly salaries as a class-indicator. So that charmingly old-school once-a-month, 5 minutes before the hour, million-quid query batch run would enforce change in the DWP workforce, and further, in wider society. It was a technical decision heavily loaded with ideological baggage, that would have profound practical consequences.

But its history is bound up with my third point. Getting the Consultancy Hat back on, it’s a cliché that adapting an IT project built for some purpose or customer to another is usually a false economy. The history of the original UC Real-Time Information system is actually quite fascinating. In the late 2000s, while the James Purnell/David Freud proposal was being shot down by Gordon Brown and Alistair Darling, HMRC was in fact working on an IT system intended to deliver real-time PAYE reporting, just what the UC boosters wanted. The only problem was…it didn’t work, and the Revenue eventually scrapped it, leaving the contractor, Vocalink, with a dud project on their hands.

Vocalink’s director of public sector clients, Peter Seymour, was up to the challenge, though. In a fortuitious meeting in a House of Commons boozer, he pitched the project to Philip Hammond MP, then the Shadow Chief Secretary to the Treasury and now of course the Chancellor. Hammond passed him onto to the circle of UC fanatics around the Dynamic Benefits report and IDS. From this point on, they would claim to anyone who asked them that HMRC already had a technical solution that would make UC happen, and act hard-done-by that nobody at HMRC had told them. The possibility that HMRC didn’t want to talk about their turkey of a project, or didn’t want to encourage anyone to revive it, seems to have escaped them.

Vocalink’s history is a little involved, but suffice it to say that it’s the IT operating subsidiary of the organisation that runs the UK’s interbank payments system. This has changed its name a couple of times, but back when it was called APACS for Association for Payments and Clearing Systems, it employed none other than Theresa May, whose first comment on seeing IDS’s white paper was apparently “Two benefits – why not one?” I find it hard to believe this didn’t help in dumping their software zombie on DWP.

Importantly, Vocalink’s project was built in the context of HMRC, where time flows in tax years commencing on Lady Day, divided into calendar months, and aren’t you glad you don’t have to write the corrections for the varying lengths of the months. You’ve guessed it; force-fitting UC into the constraints of Vocalink’s zombie software project baked the six-week delay into it!

But here’s the really interesting bit. Many of the worst aspects of UC actually made their appearance between the climatic off-site meeting in June 2010 when it got the green light, and when the detailed documents appeared in the autumn of that year. The IfG report doesn’t discuss how this happened at all, which makes me think I should read it exoterically and infer that this is important.

3 Comments on "Universal Credit: the history of an IT project failure"

  1. Very interesting, very depressing, and horribly familiar for anyone who has ever worked on a failed IT project.

    Thanks for taking one for the team and reading through it, Alex.


  2. The bits in the report covering the actual software development are really startlingly awful. The ex-MySociety crew over at GDS come out of it quite well. I should probably write some more about that.


Leave a Reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.