Universal Credit IT: What we know; what we don't know

“There is no debacle on Universal Credit” declared Iain Duncan Smith in his opening salvo to MPs on the Work and Pensions Select Committee this week.

The beleaguered secretary of state yet again refused to accept that there is any problem with his flagship welfare reform, a programme that will define his political career in government.

A highly critical National Audit Office (NAO) report in September, and the hammering his senior officials took at the hands of the Public Accounts Committee (PAC) that same month, were dismissed as historic records of past problems that are simply no longer relevant.

“The IT is working,” declared Howard Shiplee, the director general of Universal Credit, one of the UK’s top project managers in the construction industry, brought in by Duncan Smith to turn around the programme last May having overseen the building of the Olympic Stadium.

“A shambles,” countered Labour’s shadow work and pensions secretary, Rachel Reeves.

The Department for Work and Pensions (DWP) obfuscated for months when we now know that Duncan Smith had ordered a top-to-bottom review of Universal Credit in 2012, insisting all was well when we subsequently learned it was absolutely not.

So he can hardly be surprised when critics and press alike express a certain cynicism over the true progress being made.

Much has been revealed in the past few months, and this is an opportunity to step back and look at what we know, and what we don’t know, about what is happening on Universal Credit.

A “twin-track” approach

At the PAC hearing, Howard Shiplee revealed that he was close to making a recommendation on the way forward for the IT developed as part of the Pathfinder pilot project. That IT had already been the subject of much disagreement.

The DWP admitted at the time that £34m of IT work – out of a total of £303m spent – had been written off, or “impaired” in accounting-speak. The Cabinet Office’s Major Projects Authority – brought in to review Universal Credit earlier in 2013 once the depth of problems had become clear to Duncan Smith – described the IT developed so far as “not fit for purpose”, and suggested a further £140m could be written off.

Shiplee insisted that much of the IT still had “utility” and his review would determine how much.

A behind the scenes disagreement then followed – some sources used the word “row” – between DWP and Cabinet Office, and between Duncan Smith and Cabinet Office minister Francis Maude, over the best way forward.

The Ministerial Oversight Group tasked with making decisions on Universal Credit were presented with two options: To throw away all the IT developed so far by suppliers IBM, HP, Accenture and BT, and start again; or to salvage as much as possible in the short term, while developing a new “enhanced” IT system that will support full roll-out of Universal Credit by the end of 2017.

Maude, it is said, favoured starting again, with the Government Digital Service (GDS) that he controls taking the lead. Duncan Smith felt that writing off so much IT was politically unacceptable.

While this was going on, GDS was working with DWP to design a “digital system” to take Universal Credit forward.

The end result, announced last week, was a victory for Duncan Smith – a new “twin-track” approach will see the existing IT continue to be used and developed as the Pathfinder expands to other parts of the country, while a new “enhanced” and “digital” system will be developed in parallel.

We are left free to make whatever assumptions we choose from the fact that GDS has stepped back from its involvement in developing the new system.

Both of those twin tracks warrant further analysis.

Written off or re-worked?

DWP has now “impaired” £40.1m of IT work. When MP Glenda Jackson asked during the select committee meeting what “impaired” meant, one of her colleagues muttered, “Down the drain”.

Mike Driver, DWP’s finance chief, insisted that every IT project includes a degree of “re-work” citing figures from Forrester Research that claim it is not uncommon for as much as 40% of any development to be re-worked before reaching a working solution.

I haven’t read that Forrester paper, but I suspect its definition is subtly different.

A further £91m has been declared as an asset, but will be written off over five years, instead of the planned 15. In front of the committee of MPs, Driver insisted the NAO had given a “clean audit” and approved the move.

The very next day, an NAO report on the DWP’s latest accounts described that decision as “a major change in accounting treatment” and highlighted “uncertainty over the useful economic life of the existing Universal Credit software”.

The NAO went on to say: “[DWP] now expects to write down £91m of the remaining assets to nil value by March 2018, due to the considerable reduction in their expected useful life. While this is the appropriate accounting treatment, it should not detract from the underlying issue that the department has spent £91m on assets that will only support a limited service for five years, with clear consequences for public value.”

Five years means the software will be written off by the time Universal Credit goes fully live – it is work that will play no part on the final delivery. DWP justifies this by saying the Pathfinder IT is being used to learn about how best to deliver the new benefits, and to take feedback from users and claimants during the early roll out to a very limited number of people.

According to the NAO, DWP will spend a between £37m and £58m of additional investment on the existing software as part of that process.

Everything that the £91m of software and much of the additional £37m-£58m is used to learn will eventually be delivered by an entirely new piece of software developed in parallel, not the software upon which the learning is gathered. You can decide for yourself if that is a worthwhile use of taxpayers’ money, or merely delaying the fact that cash has been mostly wasted.

Digital, but not “digital by default”

The select committee had a few particularly surreal moments when discussing the differences between the existing IT and the new “enhanced” system.

When asked to explain what had changed, Shiplee – by no means an IT expert, and asked by MPs to use layman’s terms – said that DWP had been building a “conventional system” with “lots of hard drives” and “big memory banks”.

“The digital approach is very different,” he said. It does not need “large amounts of tin”, and revealed it would use open source and “mechanisms to store and access data in that [online] environment”, which will be “much cheaper to operate and to build”.

When asked why that approach was not adopted from the start of the project, Shiplee said, “Technology is moving ahead very rapidly. Such things were not available two and a half years ago.”

That’s a claim that would rather surprise those IT experts that have been using open source for two decades, not to mention the GDS team that successfully redeveloped the entire government website estate using open source and cloud technologies over the past couple of years.

We also learned that Universal Credit will no longer be “digital by default” – the GDS digital strategy mantra for redeveloping online public services. That had been an “aspiration too far”, and Shiplee cited security concerns as being the reason for the change.

Bear in mind that GDS sees its identity assurance security programme as absolutely central to “digital by default” public services.

However, the new system will instead be just “digital”. Duncan Smith said digital by default had proved “overly ambitious”.

Welfare reform minister Lord Freud added: “When people say it’s not digital by default it’s a minor thing because there are some things we can’t do because of security reasons.”

That’s a line I expect GDS chief Mike Bracken would privately dispute with some force.

So we are left with a system that will be digital, but not by default. Your guess is as good as mine.

What’s more, the digital system will again be developed using agile methodologies – an approach that was initially lauded by Duncan Smith in the early days as the way to avoid costly IT failures, then subsequently scrapped.

DWP has recently recruited a new digital chief, Kevin Cunnington, from Vodafone, who, we must assume, brings new agile skills to the table that were previously lacking.

I’ve been told by sources that Cunnington and Shiplee have already had “disagreements”. Shiplee was asked by the select committee about rumours of disagreements on the way forward – presumably in reference to the Maude / Duncan Smith debate – and he dismissed this as nothing more than you would expect on any major project. Presumably any differences of opinion with Cunnington are the same.

The NAO this week described the new system as having “uncertainties over the exact nature of the digital solution, and in particular: How it will work; when it will be ready; how much it will cost; and who will do the work to develop and build it”.

So, nothing to worry about there.

“The IT is working”

The DWP line continues to be that the IT is working. This has been a statement made by officials again and again, even during the times when we later learned the IT was not working and was instead being deeply analysed by the government’s own Major Projects Authority to see if it was any good at all.

The issue here seems to be the definition of “working”.

The Pathfinder system, we learned from the September NAO report, has limited security features, no fraud detection, does not allow claimants to change their personal details online once submitted, and only caters for the simplest category of potential claimants.

Sources also say that the amount of manual intervention needed to process claims means it takes longer to complete a transaction than the existing systems used for benefits such as Jobseekers Allowance.

So, if you define “working” as being a system that has poor security, no fraud detection, does not meet user needs, does not cater for the vast majority of claims, requires a lot of manual intervention, and will not be used for the final roll-out, then yes, it’s working.

The DWP will say that as this is a Pathfinder, it is being used to “find a path”, to quote permanent secretary Robert Devereux. It can be used for the limited functionality for which it is required to be used.

It does not, however, work well enough to cope with the 1.7 million claimants that Duncan Smith claimed in March would be on Universal Credit by 2015.

Next year, this system will be further developed to cater for couples, and children, with all the extra complexities that requires. It will be used to “learn” about processing those more difficult claims. And it will then be thrown away and replaced by the digital system being developed in parallel before Universal Credit is rolled out to three million people during 2016/17.

Delaying the risk

Sources say that Duncan Smith had two priorities when deciding on the revised plan for Universal Credit – to minimise the IT write-off, and to get people on the system (either system) as soon as possible.

Even then, the new plan sees significantly fewer people being migrated to the new benefits by 2017 than the DWP had originally intended – 2.9 million instead of the previous target of 6.1 million.

The vast bulk of claimants will never see one screen of the existing IT system that will eventually be thrown away – their details will only be used in the new “enhanced” and “digital” system in development.

Between £25m and £32m will be spent to the end of 2014 on the digital system, with the aim of having just 100 initial claimants processed as a trial.

As a result, it has been estimated by DWP that during the 14 months leading up to the Universal Credit deadline of the end of 2017, more than 200,000 people will be migrated from old (non-Universal Credit) benefits system to Universal Credit every month.

That figure comes from DWP insiders, but is validated by the estimates made by the independent Office of Budgetary Responsibility (OBR) to accompany the government’s Autumn Statement last week.

That mass migration is entirely back-ended on the project, and its scale is such that it can only represent an enormous risk to the 2017 deadline. Sources say that DWP staff “audibly gasped” when they heard the figure of 212,000 average monthly migrations first mentioned at an internal meeting in October.

The OBR said, “There is clearly a risk that the eventual migration profile differs significantly from this new assumption”, highlighting the challenge of such a back-ended roll-out.

The NAO pointed to “significant levels of technical, cost and timetable uncertainty.”

Even if the DWP has a working digital system ready to receive 212,000 monthly migration before the end of 2016, that figure still represents a massive risk to achieving the 2017 deadline.

It’s a risk that Labour would do well to consider – whoever wins the next election in 2015 will be responsible for overseeing that migration.

What is the role of the IT suppliers?

The four main IT suppliers to Universal Credit – IBM, Accenture, HP and BT – have been notable by their absence in the debate around the mess so far. To be fair to them, they have effectively been gagged by DWP, and any queries around Universal Credit are politely (well, mostly) referred back to the department.

It is amazing how lightly they have got off so far. The main criticism has come from the Public Accounts Committee, which questioned the conflict of interest inherent in asking those suppliers to assess the value for money of their own work – “Marking their own homework”, as one MP put it.

DWP has consistently refused to publish those assessments, despite Computer Weekly requesting them under Freedom of Information laws.

With the new twin-track approach, it is unclear to what extent the four firms will be involved in the new developments, although they will almost certainly be involved in work on the existing, to-be-phased-out system.

Shiplee suggested to MPs this week that he would seek to renegotiate contracts. Currently the suppliers work on a time-and-materials basis, and were under such poor financial controls, said the NAO, that nobody was able to say what invoices related to what work delivered. Shiplee wants to move to a fixed-price contract, which in the circumstances surrounding the existing systems seems eminently sensible.

However, a fixed-price contract tends to suggest that the suppliers will not be working so much on the agile development for the digital solution – no supplier’s legal team would accept a fixed price without a specified requirement, which runs contra to the iterative principles of agile. That was the reason why time and materials was considered appropriate for the original contracts, which were meant to be for agile work.

Sources suggest that the suppliers work without co-ordination, developing software that is different in look and feel depending on which bit of the system has been developed by which supplier.

One of the suppliers told Diginomica that nobody at DWP is telling them what to do.

When asked what would be the role of those suppliers henceforth, the DWP’s response was somewhat non-specific: “We will continue to work with existing suppliers as we develop the new enhanced system and ensure existing systems build into it.”

So, what next?

Iain Duncan Smith continues to take the flak and issue denials, insisting – because he has little choice – that Universal Credit will be on time and budget, or at least on the latest time and the latest budget.

His prize is the billions of pounds he claims the welfare reform will save the public purse. Nobody disagrees that Universal Credit is, in principle, the way forward for the UK benefit system. It is only the practice that is under such scrutiny.

The NAO said that DWP “needs to learn the lessons of past failures” – that’s about as succinct and understated a summary as you will see.

It is not only the past failures of Universal Credit that need to be learned from, but those of every major government IT disaster of the last 15 years. That is what is so dispiriting about this debacle – sorry Iain, there is no other word – that all the many reports and recommendations about how to improve government IT have been so spectacularly ignored in this one project.

There is, for the first time in many years, a real sense of optimism that elsewhere, government is starting to get IT right, thanks to GDS and the work led by Mike Bracken and chief technology officer Liam Maxwell. Thanks to them, IT now has a seat at the policy table, not just delivery, and GDS has assembled a team with a startup culture, and some of the most talented developers in the country.

It’s no surprise GDS wants to distance itself from Universal Credit. But there is little doubt that Universal Credit will be a huge factor in judging the success of IT for this administration – let alone the political success of Iain Duncan Smith.

Much as the secretary of state says the rumours and claims published in the press are wrong – they are not – and wishes they would go away, there is much more to be written about Universal Credit and its IT yet.

Data Center
Data Management