Why do we never learn? Because good practice is punished hardest when it matters most

The written evidence to the Public Administration Select Committee makes depressing reading. Much is not new. It raises the question “why do we never learn”. I alluded to the core reason in my blog comparing the proposed IT strategy for the Universal Credit with that for a Universal Patient which led to the National Plan for IT. The punishments are draconian for “speaking truth to power” at the time when it matters most: when everyone is admiring the “vision” and those at the top are at their most gullible. Anyone asking awkward questions will be condemned as a naysayer. Hence the dilemma of Mr Brickworth in C Northcote Parkinson’s little essay on “High Finance”.

My own submission to the enquiry was as follows:    





2. I have been involved in the delivery of IT systems for over forty years since I was responsible for merging and decimalising (1971) the sales ledger systems of the companies that had come together to form ICL. My MSc (London Graduate Business School) paper on “Why Computer Systems Fail” was published in 1973 and I subsequently ran the ICL-DTI Study to help the new Regional Water Authorities with a Computer Development Plan that was not only based on established good practice but was actually followed and succeeded. I have been involved in many such studies since, including the orignal computerisation of PAYE and using the experience of decimalisation to advise companies (and government) on how to avoid problems with Y2K.


3. I have been involved in public policy with regard to the use of IT for thirty years, since working on policy studies during the run-up to the 1979 election (the liberalisation and privatisation of telecoms and the micros in schools programme).


The main points in this response are:


4. The reasons why good practice is not followed are usually political. Most were well described by Machiavelli, Kipling and C Northcote Parkinson and can also be found in the Old Testament, Confucius and Sun Tse.


5. The core task is to ensure that ministers harness professional advice to find safer ways of implementing their change programmes and do not over-ride good practice because it is politically unacceptable in the short term.


6. That is not easy but improved consultation and scrutiny, including by Parliament, can help reinforce pressures to follow good practice, reduce temptations to over-ride it and limit the damage when it happens. IT itself should be better used to assist the process.


Question 3 Have past lessons from NAO and OGC reviews about unsuccessful IT programmes been learnt and applied?


7. No. That leads to the Question “Why Not?”. I have twice written well-received papers on the question of “Why do we never learn: the pre-conditions for public sector success”.


8. That for the National School of Government, was published in 2008 and is available on-line at:  https://www.computerweekly.com/blog/When-IT-Meets-Politics/Why-do-we-never-learn-The-pre-conditions-for-public-sector-systems-success 


9. Part of the argument was:


10. “Confusion and conflict over objectives and priorities and split responsibility for policy and implementation commonly mean that no one knows what success looks like or is responsible for achieving it from conception to completion. In the worst-case scenario, proposals originate from policy professionals rather than operational staff; they are worked up by officials with equally little experience of implementation or operational delivery; and turned into procurement specifications by consultants and lawyers who have never seen a project all the way through from concept to success.


11. “There will almost certainly be at least two changes of Minister and one of officials between enabling legislation and statutory instrument, let alone procurement.  The supplier’s  “A” teams will then compete to negotiate lowest cost, blame avoidance contracts, to meet a re-negotiated compromise specification that has lost sight of the original objective. Finally the winner’s “B” team arrives to do its best. Meanwhile those who are to use the system have become increasingly frustrated with delays followed by broken promises as to what the system is supposed to achieve and when.


12. “The main reason why such problems persist long after they were first identified is that those who plan clever policies using fashionable technology are promoted to repeat their mistakes elsewhere before they have time to learn. Rudyard Kipling might have had the relationship between Policy Advisors, Technology Gurus and Ministers in mind when he wrote how:


13. They sit at the feet – they hear the word – they see how truly the promise runs

They have cast their burden upon the Lord and – the Lord lays it on Martha’s sons


14. “For Kipling the “Sons of Martha” were the engineers who made the systems and infrastructures of the Empire work. Today they are long gone and we no longer train their successors to do the same for the systems and infrastructure of today.


15. “Once the planning phase has over-run, leaving the more difficult problems to be sorted out during the procurement phase, which over-runs even more in consequence, there is a rush to catch up, using the newcomers in the performance monitoring team, after those who understood the original requirement moved on during the delay. 


16. “The obsession with studies of “Best Practice”, assuming skilled and experienced staff, is itself a prime cause of the continuity of bad practice. The need is to publicise and enforce adequate practice on the part of the people you actually have available.


17. “They must then be trained to do so by those who have done it for real, with opportunities for mentored and supervised experience before they are left in sole charge of a project, let alone programme, of their own. The tendency to ‘muddle along’ delays corrective action while problems grow.


18. “Private sector projects are commonly announced only after they have been shown to work.  Moreover the commercial sector sees a reputation for reporting problems in advance and organising rapid remedial action, however brutal, as career-enhancing, not career-limiting.


19. “By contrast, public sector projects are usually announced in advance. There may be increasingly desperate attempts to fix any problems in private, without calling for additional resources, with notification up the chain of command only if the problems cannot be solved with the resources to hand. This is almost certain to limit the careers of those who called for help.


20. “The culture of private cover-up followed by public witch-hunt used not to be peculiar to the public sector. It used to be found in many large businesses. But those facing global competition can no longer afford to try to conceal problems, as opposed to earning reputations for acting fast to resolve them.”


21. The paper leads through to action plans such as:


22. “manage political expectations, beginning with what is realistic, given the time and resources available. This requires ensuring that the minister’s policy team includes advisors with relevant practical experience of delivery.


23. “confine risk to one dimension at a time. For example, if there is a high risk that the objectives or organisational structures will change you should avoid changing supplier or systems at the same time.


24. “Break the programme into modules. In the private sector, projects which take more than three months are more likely to be cancelled than to go live. If the implementation team has not worked together before on programmes of this type it is even more essential to begin with a series of small projects and quick wins to build experience and confidence.


25. “Think big, but “start small, test hard, scale fast” is the route to systems success in the private sector. It is not a new technique and has had many names over the years: from structured evolution to rapid application development and dynamic systems methodology.”


26. The National Health Service Information Authority was following professional best practice to knit together the disparate clinical systems of the NHS to enable and encourage inter-operability based on common recording standards. Progress was beginning to accelerate when the Government (at the very top) lost patience and imposed a grandiose National Plan for IT.


27. Today we still have the problems with regard to semi-incompatible standards across the records and systems used for clinical or professional purposes, despite having spent billions on recording systems that are said to be insufficiently accurate, available and reliable to support anything other than “defensive” medicine.


28. The linking of an updated PAYE system that will cope with the income fluctuations of those who move in and out of employment to a new DWP benefit systems, to provide Universal Credits is, in key respects, even more ambitious than the attempt to give on-line access, anywhere in the UK, to an accurate current patient record. It also has similar political support and impetus.


29. Those who call for independent professional review to find low risk incremental ways of introducing such radical change will be as politically unpopular as those who counselled against the NPfIT, even before it became obvious to senior clinicians that it did not relate to the realities of patient care and they would have to maintain their own parallel systems in order to meet their moral and professional obligations.


30. It is politically much easier to promise that a new system will be carefully planned and tested for several years by well-known consultants and suppliers before being introduced, than to admit that neither officials or suppliers have experience of successfully handling the scale and nature of the risks involved.


31. The UK benefits systems is uniquely centralised and complex. The UK is, however, also unique in having a single, centralised payment clearing operation. The transition of the banks, including their networks of automatic teller machines and other on-line systems, to internet protocols and common smart card standards is probably the only major change of equivalent risk to those being proposed to support the Universal Credit. The way that transition was planned and phased has many lessons for government. The scale and nature of the staff and customer education programme necessary for Universal Credit to work is, however, even greater.


32. It is therefore important to use IT itself to help prepare the way with better consultation and scrutiny, to ensure that politicians, staff and public, as well as the implementation teams, know what to expect.


Question 4 How well is IT used in the design, delivery and improvement of public services?


33. The EURIM response to this consultation by the PASC says ” IT is rarely used in the design and/or targetting of public services. We commonly “retrofit” IT to deliver a service that has been specified, in primary and secondary legislation, without testing how it is likely to work in practice. The untested specification is then put to out tender in an expensive ritual driven by consultants and lawyers paid according to time spent, or given to the incumbent contractors to implement without external scrutiny of value for money.


34. “Policy initiatives should be subject to computer modelling during the design phase to see how they are likely to work in practice. Existing public records (tax and benefits) should be collated anonymously with private sector databases (e.g. credit reference and market research) to identify how many individuals or businesses will be affected. Those scrutinising legislation should have access to such simulations when debating the proposals and any suggested changes.” 


35. I recently blogged https://www.computerweekly.com/blog/When-IT-Meets-Politics/How-to-prevent-the-DWP-Universal-Credit-from-being-yet-another-doomed-IT-project on the ideas of the late Donald Michie on how computer simulation might be used at each stage of the policy process, from formation and initial planning, through consultation and political scrutiny to procurement and implementation. The aim would be to help improve understanding of what is proposed and how it is intended to work and, at the same time, identify and resolve problems, including conflicts of priority, early enough in the process for to prevent them causing unacceptable delays and over-runs during implementation or chaos at the start of live running.


36. The process should not be used to help plan an optimised, delayed-bang “big” system. Times change, needs evolve and, by definition, what is optimal today is sub-optimal tomorrow.


37. The aim should be to help plan an evolutionary process in which the system will change incrementally over time, with each set of changes consulted on and tested before commitment, with the simulation helping ensure homogeneity and continuity of approach.


38. There are, of course, many issues around implementing such an approach but it is closer to the mainstream of current and evolving private sector best practice than the way in which big public sector systems are announced years in advance, constrained by legislation and implemented after several changes of customer (alias minister and senior departmental official responsible for the system).


39. In parallel an analysis of the live transactions flowing through the existing systems should used to inform ministers and those involved in consultation and scrutiny as to how many individuals are likely to trigger which “rules”.  This could lead to not only to better estimates of costs and impact but also better debate as to how rare conditions and exceptional hardship cases could and should be handled. 


40. During the EURIM transformational government dialogues Lord Kirkwood (former chairman of the DWP Select Committee) commented that barely 20 of the 1200 or so “rules” were needed for over 80% of claimants on one system and those responsible regarded the complexity as a matter of pride, not shame.  


41. A new approach may also help ensure that the Universal Credit is indeed successful in addressing the problem that most welfare systems assume predictability of need and PAYE was designed for those in stable employment. Meanwhile “those in most need lead lives of quiet desperation, lurching from unpredictable crisis to unpredictable crisis. Then, if and when they get their lives together, with a brief period of work and prosperity, the system catches up with them and crushes them back to poverty with its demands for payback”.


42. To really help those trying to help better themselves, we require systems that assume chaos and unpredictability. That entails giving front-line staff (and those in the delivery “partners”) responsibility for holistic support and the ability and authority to over-ride the “system”. It is unclear whether the currently planned reforms are intended to go that far, although some of the “Big Society” rhetoric might be seen to convey that expectation. 


43. Tasking IT professionals to design and support systems that allow humans to over-ride the “rules” in order to handle the unpredictable (and logging their decisions and reasons for audit and accountability) is unusual in central government, where there is a common presumption that the rule book and system, should cover all eventualities. We can see the consequences.


44. My personal professional “prejudice” has always been for simple systems with routines for manual over-ride, tools to help authorised human beings to decide how and when to do so and a log to record their actions and reasons. Having worked alongside some world-class systems engineers I would not, however, elevate that view to being anything more than “reasonably well-informed prejudice”. 


45. I do, however, believe that this problem goes to the heart of why so many big and well-intentioned central government systems fail.