The West Yorkshire snack vortex, his personal shopper, NHS computing, and FRES

I’m currently reading Oliver Campion-Awwad, Alexander Hayton, Leila
Smith, and Mark Vuaran of Cambridge Computer Lab’s case study on the NHS National Programme for IT, an old topic of this fine blog’s. There is so much in common here with Think Defence‘s epic blog series on the disastrous FRES project it’s not even funny.

In this post we’ll look into how, why, and what this has got to do with Eric Pickles.

The Labsters conceptualise the NPfIT disaster as being another case for the growing academic study of IT project failure. Obviously, they would, they’re the Cambridge Computer Lab’s MPhil in Public Policy – wait, that’s an actual degree? wow! They identify various common points from past failures and note that NPfIT had them in spades, despite the fact that a lot of what was known about project failure in 2002 was known from studying NHS IT projects. These were as follows:

Haste – an unrealistic timetable, no time to engage with users, inadequate preliminary work, failure to check progress, failure to test systems

Design – failure to recognise the risks of large IT projects, failure to recognise that the longer the project takes, the more likely it is to be overtaken, sheer ambition, project is too large to manage, confidentiality issues

Culture and skills – a lack of clear leadership, not knowing or constantly changing the aim of the project, not committing necessary funding from the outset, a lack of concern for privacy, no exit plan or alternatives, lack of project management skills, emphasis on price, suppliers depend on lowballing and charge heavily for variations to poorly written specifications

They provide a detailed history of NPfIT, starting off with the original Information for Health strategy in 1997-1998, which actually sounds quite awesome. This strategy foresaw various aims, which boil down to three functional areas – a system for keeping records and running workflows, a management information system for both managers to monitor how well it worked and clinicians to monitor how effective medical practice was, and a Web site to disseminate information about health and to help the public influence NHS policy. The nice thing about this is that the three areas have minimal interfaces, and even those are read-only (the MIS sucks up logs from the first; the Web site uses stats from the MIS).

The even nicer thing about it was that it didn’t require centralisation, indeed, its five principles explicitly required the maintenance of patient confidentiality and local ownership of the system. These were maintained as late as 2000. The even nicer thing still was that it had a sort of Stafford Beer recursive quality; medics would use evidence from it to practice evidence-based medicine, managers would use evidence from it to improve the organisation, and we, the political nation, would use evidence from it to supervise the managers.

How this went wrong, well, it’s basically got Tony Blair’s name on it. Replacing Information for Health with NPfIT was Blair’s idea. It came from Tony Zoffis, as they used to say. Microsoft (and Cisco) pitched him silly and he then imposed it on everyone else. The key meetings were in February and March 2002, exactly the apex of Blair’s self-confidence. The Wanless review, better-known for pointing out that the NHS needed money more than it needed “reforms”, played a role; Wanless thought it needed better IT, and that the central government should hold it to doing it, but he didn’t actually recommend anything like NPfIT. However, the report did serve to sell NPfIT to the Treasury, the only force in British politics that could veto a project backed by the prime minister.

I’ve got more to say about this, but let’s pause for a quote.

Management consultant Thomas Brooks, who was involved in NPfIT under contract for a number of trusts, commented that in the procurement process “the iSoft Lorenzo offering was selected from paper descriptions with minimal demonstrations of prototype software elements”

Compare:

One of the Boxer prototypes in APC configuration would participate in the Trials of Truth, joined by Véhicule Blindé de Combat Infanterie (VBCI) from Nexter (previously Giat Industries) and the General Dynamics Piranha V. It was rumoured that the head of DE&S, Lord Drayson, wanted the VBCI because it would be quickest into service.

Both Boxer and VBCI were rejected by the Army, despite what Lord Drayson wanted. We had a choice of two vehicles that would need minimal development and were already (or about to be very soon) in production, and a PowerPoint design, the PowerPoint of course won the trials. The actual vehicle trialled was not Piranha V but Piranha Evolution, a surrogate for the final design.

It’s uncanny, isn’t it? FRES was a huge unwieldy mess with too many stakeholders, aims that constantly changed, and a deliberate determination to avoid meaningful test or development on the government’s time. There was even an insult for civil servants or officers who dared suggest that the end product might be an armoured vehicle of some sort, rather than, say, a new shade of the colour blue. They were said to be “solutionising” and usually moved on. Concrete thinking was reserved to private contractors only.

NPfIT had a parallel phenomenon, the “outcomes-based specification”. Presumably this originated in some sort of vague awareness of object-oriented programming, but the silliness can be summed up by the fact they set out to draft an outcomes-based specification for a standard data interchange format, which sounds like either a sheet of A4 with “IT SHOULD WORK” on it, or else an elaborate exercise in doing the actual work without admitting to it. It didn’t go anywhere; when things eventually deployed, huge amounts of time were taken up mapping nonstandard field names.

As a result, both projects tended to drift in a cycle between ambition growing without limit, untethered from the ground truth, and stodginess, lacking inspiration because out of touch with the possibilities of the technology.

It seems hard to fault FRES for haste of all things, but I think I would. Reading through TD’s series, one thing which stands out is the combination of haste – there was never time to do it properly – with timelessness – deadlines were never allowed to bite. This is precisely what happened with NPfIT, and come to think of it, every project I’ve been involved with that failed. It’s a special kind of time when frenzy and stasis combine. NPfIT’s schedule was always wildly unrealistic – Department of Health R&D Director Sir John Pattison promised Blair delivery in 2 years, 9 months in a meeting where Pattison recalled nobody seemed to be able to say “no” to Blair – but every time the deadline came up, it was just rolled over.

In both cases, the government tried to outsource its own outsourcing. The only element of NPfIT that worked properly and that was part of The Vision was the network. BT built that and it did the job itself. Everything else was contracted out to a contractor, who then subcontracted, creating a minimum of three layers of abstraction between the customer and the supplier.

Under FRES, the government hired WS Atkins as a “systems house”, whatever one of those is, shoved in between the MOD and the various contractors. The idea was apparently that they would manage the process of managing the development of the system (no solutionising!) and also its procurement.

This is the logical end point of our friend Pickles’ worldview. Pickles claimed back in the 1980s that there was a US town council that met once a year, just to issue contracts to run all its services for the next 12 months. This turns out to have been a fairy tale, but that’s by the by. He tried to implement this in Bradford and failed, but its spirit infused the procurement reforms of the 1990s, all of which were designed on the basis that letting the government get involved in the stuff it bought would be stupid. Instead, the point would be to pick as between brands of biscuits at the supermarket. The really weird thing here, though, is that Pickles procurement differs quite dramatically from the sort of thing neoliberals like to say about stuff you buy in the supermarket. Rather than being an active and informed consumer, the government is expected to use a personal shopper. How well this works…well he did spend £10,000 on snacks.

With regard to NPfIT, this intersected with other political imperatives. The famous LSPs, the super-contracts that didn’t actually match any NHS structure and only existed to make a better size of contract for Accenture or whoever, were also intended to channel a more general management influence from Blair’s office into the NHS. The Gate Zero review of NPfIT said:

There is widespread appreciation that the programme is a change programme first and foremost albeit with significant IT elements

Sir John Pattison said in mid-2002:

there was a need to create a new tier in the procurement process “to ensure not only that technology solutions are available and accredited, but to underpin those implementations with comprehensive change management

But what was the change that needed to be managed? It is surely very telling that the Health Secretary involved was Alan Milburn, memorably described at the time as leading the “Special Republican Guard” of ultra-Blairites, inventor of foundation hospitals and independent sector treatment centres. This was the peak of Blairite confidence; on the privacy front, Milburn had just legislated himself the right to dispose of NHS information as he pleased. (Is it significant that as we now know, GCHQ was growing at a rate of knots?)

Arguably, between him and Tony Zoffis, what was wanted wasn’t a management information system but rather a management imposition system. Rather than a system that would aid in the practice of evidence-based medicine and in public scrutiny, they wanted one that would help generate policy-based evidence to defend the changes it imposed after the fact, and to protect it from public scrutiny. Part of the take-home message here is that even had NPfIT worked, there is an argument that it shouldn’t have.

Meanwhile, FRES was certainly intended to support an army with global capability. However, its eventual consequences are a reconnaissance vehicle that weighs 32 tonnes and can’t cross most of the bridges in the Home Counties. (Hey, it’s no funnier than software you decided to stop buying but that kept coming anyway.) You could make a similar case, but then, if you decided to build a super-heavy armoured behemoth of an army that could crush anything as long as it was within a few miles of the border, you’d get today’s Israeli Defence Forces.

It is possible that there is so much software in anything important today that all big projects exhibit some of the characteristics of big software projects. Technology changes aims, though, but not as much as aims change technology. I fear that the problem is different. The upshot of both these stories is that the aims of the political settlement under which we live may make these procurements impossible.