SUBS0000018
Post Office Horizon IT Inquiry
On behalf of Core Participants:
Nichola Arch, Lee Castleton, Tracy Felstead, Seema Misra, and Janet Skinner
Closing Statement
Phase 2 - Horizon IT System: procurement, design, pilot, roll out and modification
Introduction
1. This Phase of the Inquiry has established as fact that the Horizon IT system
rolled out between 1999 and 2001 was flawed, such that the accounts it
produced were unreliable, and that was known at the time. It was also known
by some that the accounts it produced could be used as evidence in criminal
and civil proceedings taken by the Post Office.
2. This Statement considers how high those two strands of knowledge went up
the chain of command within the three responsible hierarchies: ICL/Fujitsu,
the Post Office and Her Majesty’s Government. We will consider knowledge
within the framework of the criminal law, which allows a tribunal of fact to
attribute knowledge if the evidence suggests that a person “deliberately shut
their eyes to the obvious, or refrained from inquiry because they suspected the
truth but did not wish to have their suspicion confirmed”'. We will also
? Per Lord Bridge in Westminster City Council v Croyalgrange Ltd (1986) 83 Cr App R 155 at 164
1
SUBS0000018
consider recklessness as to the result that Horizon produced unreliable
accounts, and the result that such accounts were used without corroboration
to prosecute and extract money from subpostmasters. As per G?, a person is
reckless as to a result when they are aware of a risk that it will occur, and it is
— in the circumstances known to them — unreasonable to take the risk.
3. We use the criminal law framework for considering knowledge and
recklessness because this may allow for a consistent approach to be taken
across the Phases of the Inquiry. We acknowledge that the law of tort offers
another framework, to a large extent complementary, for considering states of
mind, but the above concepts of knowledge and recklessness suffice for the
Submissions we intend to make. Self-evidently, those who hold positions of
power should not knowingly or recklessly exercise their power in a way which
brings about harmful results. Whether or not such conduct is criminal will
depend upon more than the person’s state of mind, and we do not identify any
criminal conduct in this Statement, but we consider it likely that evidence of
criminal conduct will be explored in later Phases, and for that reason a
consistent approach to knowledge and recklessness will assist.
Knowledge and Recklessness within ICL/Fujitsu
Unreliable Accounts
4. Any person who knew about Acceptance Incident 376 knew that the Horizon
system that was rolled out between 1999 and 2001 was producing unreliable
accounts.’ The question is whether such people believed that the problem had
been fixed by 2001. Those who knew that the ostensible “fix” could not secure
more than 99.4% accuracy for cash accounts at best should also have given
thought to safeguards to ensure the remaining potentially unreliable cash
2 [2004] 1 AC 1034
3 POLO0043691 pp57-58
SUBS0000018
accounts would not be used improperly.‘ The latter issue leads into knowledge
of the Post Office’s prosecutorial function, which we will return to below.
5. The evidence of David McDonnell and Jan Holmes, supported by
contemporaneous documents, shows that within ICL/Fujitsu managers knew
that the problem with unreliable cash accounts had not been fixed by the close
of 1999; and as they went into full rollout in 2000, they were, at the least, closing
their eyes to the truth. Further evidence set out below suggests that
knowledge, or at least recklessness, went as high as Keith Todd.
6. The reliability of the Post Office’s accounts was entirely dependent on the
EPOSS element of Horizon, which had been put together in a chaotic fashion
by people who lacked the necessary skills. This was laid out in stark terms in
September 1998 in McDonnell and Holmes’ Report on EPOSS PinICL Task
Force>. The Report was considered by Charles Cipione. He said that it made
“disturbing” reading, and he corroborated the main points: the RAD technique
was totally unsuitable for developing EPOSS, which needed a highly co-
ordinated approach to design and build, and the EPOSS code cited in the
Report was “terrible”.°
7. The fact that McDonnell’s recommendation to re-write the EPOSS Code was
not followed meant that by October 1999 there was no improvement in the
quality of the product. This was established by the CSR+ Development Audit’,
which quoted the number of PinICLs which had been raised as a result of
problems with EPOSS over the preceding months. There was no evidence of a
decline. Jan Holmes, who authored the CSR+ Audit, stated that “The figures
indicate that the problems facing EPOSS during the Task Force period have
“The Second Supplemental Agreement, Schedule 4 Part B (i), provided that over a monitoring period of six
weeks in late 1999, cash accounts with discrepancies would not exceed 0.6% - FUJO0118149
5 FUJO0080690
° https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-11/POH%2017%20November%202022.pdf
pp150-158
7 FUJO0079782
SUBS0000018
not diminished”, and he recommended that re-writing the code be re-
considered.®
8. The CSR+ Development Audit was addressed to John Bennett, Mike Coombs and
Terry Austin, amongst others. According to the subsequent Corrective Actions
document, Terry Austin, with input from Mike Coombs and John Bennett,
again decided not to re-write the code, and in May 2000 they closed down the
recommendation.? Mr McDonnell, when showed the Corrective Actions
document, and Mr Austin’s reasoning for the decision, said it was “based on a
logic which is ‘we've done lots of things, remedial work, code improvement’,
et cetera, et cetera but those improvements are not reflected in the PinICL
statistics. So how, if you were the client POCL... how could you accept that as
an argument, unless you didn’t have visibility of the PinICL statistics” .!° There
is no evidence that ICL/Fujitsu revealed the PinICL statistics from the CSR+
Development Audit to anyone at POCL.
9. In the period that the EPOSS code was under scrutiny but not re-written -
September 1998 to May 2000 - live trials and rollout got underway. Acceptance
Incident 376 was a bone of contention from August 1999 onwards," and it
made it clear to all concerned, and particularly to ICL/Fujitsu, that Horizon
was producing unreliable accounts. When Mr McDonnell was shown the
AI376 report he said he thought it highly likely that the problems encountered
were caused by the EPOSS code because “we saw this stuff all the time”. It
was therefore not only the internal evidence which told the ICL managers that
EPOSS would not be able to produce reliable accounts, but also the external
8 FUJ00079782 p20
° WITN04600104 pp9-10
*° https://www.postofficehorizoninquiry.org.ul/sites/default/files/2022-11/POH%2016%20November%202022.pdf
p94
11 POLO0043691 pp57-58
” https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-
11/POH%2016%20November%202022.pdf
pp91-92
SUBS0000018
evidence of AI376 which told them that, as a result, and as a matter of fact, the
Horizon system was not producing reliable accounts. The continual fixes and
workarounds which were offered up as the resolution for AI376" were part of
the problem, not the solution, as they knew. Jan Holmes put it in this way: “It
was just, you know, fix and move, fix and move, fix and move and, you know,
we all know that that's not necessarily the right way to do anything.”"*
10. When giving evidence, Terry Austin said that he could not make the decision
not to re-write the code alone, and that it was a decision taken by the Pathway
management team: Mike Coombs, John Bennett, Tony Oppenheim and Mike
Bennett. (The reference to Mike Bennett was probably intended to be a
reference to Martyn Bennett, one of the addressees of the CSR+ Report.) Later,
Mr Austin named two other members of the management team as being party
to the decision: Peter Jeram and Stephen Muchow.'°
11. John Bennett was Managing Director of ICL Pathway until early 2000, and the
most senior person named by Terry Austin as being part of the decision not to
re-write the EPOSS code. The contemporaneous documents show that he not
only received the CSR+ Development Audit but his initials appeared on the
Corrective Actions relating to the decision. Nevertheless, he denied being part
of it.” He was so intent on denying responsibility for this decision that he
would not even accept the common-sense proposition that re-writing the code
would have caused additional delay and cost.'* His denials — in the face of
evidence from Terry Austin and the contemporaneous documents — indicate
that he was well aware of the importance of this decision, and the harm it
13 See POLO0028337 and FUJO0079178 as examples of documentation describing such workarounds and fixes.
» https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-11/POH%2016%20November%202022.pdf
p125
25 https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-10/POH%2027%200ctober%202022.pdf
p22
6 Ibid p30
1 Ibid p1s4
28 Ibid pp175-177
SUBS0000018
caused. (The evidence of Mike Coombs on this issue was so heavily dependent
on the documents that it added little. He appeared to accept his own
involvement in the decisions, based on the documents.!’)
12. We submit that it strains credulity that John Bennett was not a party to this
decision. He was the Managing Director of a single-purpose subsidiary, which
had one focus in late 1999/early 2000: Horizon rollout. The six week monitoring
period for ensuring that AI376 would not bring it to a halt took place between
3 October and 14 November 1999, and that would have been at the forefront
of minds at the exact time that Mr Holmes inconveniently raised the damning
EPOSS PinICL findings. Shortly after this period, in Mr Bennett’s Monthly
Report dated 20 December, he said “We now move on_ to
the EPOSS reconciliation facility, which is required for AI376 and is one of the
other critical criteria to be reached to enable National Rollout to restart”.”° This
was a reference to the new checking facility settled upon in the Third
Supplemental Agreement and shows that he knew that EPOSS was at the heart
of AI376. His team must have sought a steer from him on the central question
of whether it needed to be re-written.
13. The likelihood is that after receiving the CSR+ Development Audit on 28
October, the whole ICL Pathway Management Team were holding their breath
until 14 November, hoping that the current fixes to EPOSS would hold long
enough for them to achieve less than 0.6% of cash accounts with errors by the
end of the six weeks of monitoring: fix and move, fix and move, fix and move.
Their plan, as set out in Terry Austin’s email explaining their decision, was to
rectify as many problems as possible at the end of the pipe: “Effectively as a
management team we have accepted the ongoing cost of maintenance rather
2° https://www.postofficehorizoninquiry.org.uk/sites/default/files/202211/POH%201%20November%202022. pdf
p79
2° FUJO0058188 p6
SUBS0000018
than the cost of a rewrite.”?! Maintenance in this context meant fixing problems
raised by PinICLs, which arose after the event. The software was being rolled
out into thousands of terminals, and, as the November monthly report
showed, even at that stage EPOSS was handling 12.5 million transactions per
month.” It would be absurd to assume that every problem would lead to a
PinICL and a prompt, successful fix. From the Managing Director downwards,
the managers of ICL Pathway knew that Horizon as rolled out between 1999
and 2001 would produce unreliable accounts.
14. The question is whether that knowledge also went up to ICL and Fujitsu. There
are indications that it did. First, and most importantly, there is the fact that
Fujitsu coders arrived from Japan to oversee what was happening with regard
to AI376 and the EPOSS code, as agreed by both Mr Austin® and Mr
McDonnell*. There is no dispute that Fujitsu, Japan conducted an audit of
Horizon as Mr McDonnell alleges but FSL do not apparently hold any records
of the audit's findings. However, surprising that may be, there are references
to the audit in the following documents:
a. FUJ00079795
b. FUJ00079796 (this relates to the PA Consulting review in 1997)
c. FUJ00058188 (this relates to audit by PA Consulting in November / December
1999)
d. FUJ00003630
15. Second, there are hints in the Board paperwork. The September ICL Board was
told of the three high level acceptance incidents.* Then on 27 September,
shortly before Jan Holmes sent the CSR+ Development Audit, John Bennett sent
21 WITNO4600104 p10
2? FUJO00S8188 p27
23 https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-10/POH%2027%200ctober%202022. pdf
pp37-38
24 https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-11/POH%2016%20November%202022. pdf
pp20-22
?5 FUJ00003631 p2
SUBS0000018
a Memo to Keith Todd, Richard Christou, and ICL Board Members who sat on
the “EMC”. It said he was delighted to confirm Acceptance, as the Second
Supplemental Agreement (which settled the monitoring period) had been
signed on 24 September. It also said “Acceptance has been a tough process —
but the good news is that it is behind us”.”° That was something of a hostage
to fortune.
16. Keith Todd in particular had a reasonably thorough grasp of AI376 and the
problem of unreliable accounts, because he read the ICL Pathway Monthly
Reports, which raised the issue regularly.” In his evidence he said that his
Reports to the ICL Board provided a synopsis of the “main factors” raised
therein.”* (Unfortunately, we do not have a complete set of Mr Todd’s Reports
to the ICL Board.) It is interesting, therefore, that the question of re-writing the
EPOSS code was never directly addressed in the Pathway Monthly Reports.
This may be seen as surprising, given that it was described in the Corrective
Actions document as an action that “falls within Development but requires
higher level drive”, and after it was raised in the October 1999 CSR+
Development Audit it rambled on until May 2000.
17. We question whether leaving it out of the Monthly Reports was intentional, to
provide Keith Todd and the ICL Board with a cloak of deniability. The ICL
Board meeting on 25 November 1999, just days after the monitoring period
ended on 14 November, may have been a moment when Mr Todd felt able to
give the Board at least a hint of the problems, given that it now seemed likely
that they were going to be able to muddle through them. The relevant Chief
Executive’s Report which Mr Todd prepared for this meeting is one which
does not appear to have been disclosed on the Relativity platform, but the
26 FUJO0079189 p2
27 For relevant sections of the Pathway Monthly Reports see Sections 13 and 17 of EXPG0000001
28 https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-10/POH%2025%200ctober%202022. pdf
pss
SUBS0000018
Minutes record that after the Pathway team received congratulations for
achieving the rollout milestones which released £69 million and £90 million, it
was noted that there were “difficult points to be resolved before rollout could
re-commence in January, but Mr Christou was now hopeful these could be
resolved in time”.*? Having scraped through the monitoring period, there was
indeed reason to be hopeful.
18. Whether or not the Board as a whole had knowledge, we submit that if Mr
Todd did not actually know of the problems with the EPOSS Code, he was
closing his eyes to the obvious truth. The problems with EPOSS were first
raised in Pathway Monthly Reports from March 1998.” Given the cumulative
effect of these Reports, and Mr Todd’s use of them to compile his own monthly
Chief Executive’s Reports, he was very well aware of the persistent
unreliability of EPOSS and the effect that would have on accounts.
19.Mr Todd’s knowledge of the problems with EPOSS (and perhaps more
generally) is indicated by his repeated attempts in late 1998 to minimise
testing, first to the level of model office only, and then to a small number of
live offices. As Sarah Graham said in her evidence: “I think, just in a common
sense way, something didn't seem right in this constant... sort of eating away,
trying to make things easier”.*! The documentary evidence that he made these
attempts is irrefutable: in a letter supportive of his position, Ian McCartney
said “the company have moved substantially from their 9 November position
and have now conceded that acceptance will follow live trials based on the
NR2 software release at 300 offices”. In his evidence, Mr Todd claimed not to
2° FUJ00003595 p2
3° FUJO0058170 pS
31 https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-11/POH%2024%20November%202022. pdf
p144
3? BEISO000400 p2
SUBS0000018
have any recollection of these attempts.* We would suggest that this may have
been a convenient lapse of memory.
20. His positioning throughout 1998 and early 1999 took little account of the
known and obvious problems with the Horizon system, which was already
very delayed, and that in itself indicates recklessness if nothing else. He
claimed in July 1998 that “the system is technically sound” and “deliverable
within a relatively short timescale” .* In other letters during June and July 1998
he claimed that Release 1c was “sufficiently robust for a rollout to several
thousand offices”. This bullishness was undercut by his own Managing
Director, because on 7 August 1998 Mr Bennett identified several weeks of
necessary work before further rollout of Release 1c.* There was, as the Inquiry
has explored, a concerted effort to recategorize, even minimise acceptance
incidents requiring discourse, co-operation (and mutual knowledge of the
issues) driven by Pathway’s and POL’s joint interest in delivering the project**.
This must have been known at Board level by Mr Todd and Mr Bennett.
21. Itis plain from the contemporaneous documents that Mr Todd was absolutely
intent on getting the Horizon project over the line, and in a meeting with Steve
Robson in May 1999 he was “clearly quite worked up”. It is also difficult to
accept, despite his protestations to the contrary, that Mr Todd’s position as
Chief Executive was not in question, because of the planned flotation of the
company, and the importance of the Pathway contract to those plans.* The
* https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-10/POH%2025%200ctober%202022.pdf
p139
34 DWP00000174 p1
35 POL00031127 p22 para 411
3© POLO0028359
37 HMT00000013 p1
38 https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-10/POH%2025%200ctober%202022.pdf
pp115-117
10
SUBS0000018
reputational and commercial impact of an aborted project of this importance
would have been immense.*”
22. After the event people are often reluctant to admit reckless self-interest, or
ruthless commercial partisanship, especially in a case where such motives
have led to a terrible tragedy, but we submit that Keith Todd’s determination
to secure the Pathway project, closing his eyes to the obvious truth that AI376
was caused by the unreliability of EPOSS, is beyond question. We further
submit that he was at best reckless as to the dire consequences arising from the
instability of the Horizon system as rolled out between 1999 and 2001, in that
it produced unreliable accounts, which (over the entire network) would result
in subpostmasters being held responsible for entirely fictitious shortfalls,
which were treated as incontrovertible losses by POL.
Unreliable accounts to be used as evidence
23. Clauses 4.1.8 and 4.1.9 of the Codified Agreement of 28 July 1999" are now
well known, and they speak for themselves. ICL Pathway was committed to
providing audit trails and other information necessary for POCL to commence
investigations and prosecutions. Only Tony Oppenheim, of all the Pathway
managers, admitted to knowing about this provision”, but it was plain to see,
and its forerunners were also in plain sight. Indeed in the original iteration of
the agreement the need to support fraud prosecutions was central, and from
as early as November 1994 ICL was boasting of its capability to support
prosecutions and provide an audit trail.” It may have been an afterthought in
the re-negotiations that took place in 1999, but there must have been conscious
thought, otherwise Clauses 4.1.8 and 4.1.9 would not have found a place in the
> https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-10/POH%2025%200ctober%202022.pdf
pp115-117
4° FUJ00000071
“1 https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-10/POH%2026%200ctober%202022. pdf
p46
#2 FUJ00098230 p&
11
SUBS0000018
Codified Agreement. It is easy to see why no one but Mr Oppenheim was
prepared to admit to knowledge. The questions he faced on the subject must
have been uncomfortable. The unwillingness to admit to knowledge in the face
of the evidence is in itself an indication not only of knowledge, but of an
awareness of how important that knowledge was.
The Post Office
Unreliable accounts
24. John Roberts made a bold attempt to testify that he and the Post Office Board
believed that problems with cash accounts were about training rather than
technology, but following an interjection from the Chair, he accepted that by
the time the Third Supplemental Agreement was signed on 19 January 2000,
they were aware that it was not just a training issue.* In fact that much was
absolutely clear not only from the Agreement itself, but also from the January
Board Minutes“, and the Board Paper from Stuart Sweetman which
underpinned them*. Going further back to how AI376 was portrayed in July
1999, it is right to say that the Minutes record David Miller as saying that he
“considered the system robust and fit for service’”**. However, his evidence
was that if he made this statement it was incorrect, and he pointed out that on
the next page of the Minutes, and under the same heading, it was recorded that
“members were concerned that a number of technical issues remained
unresolved”. “” More importantly, this was against a backdrop of many years
of delays, and a recent period of negotiation during which Keith Todd had
3 https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-10/POH%2020%200ctober%202022. pdf
p109
* POLO0000336 pp10-11
*5 POLO0089770 p2
*© POL00000352 p11
47 https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-11/POH%2028%200ctober%202022.pdf
p66 and 69. He also picked up on the underlying Board Paper, which was not at that stage before the Inquiry,
but it is on Relativity at POLO0028491, and unfortunately it sheds no further light on the Board’s knowledge of
technical issues.
12
25.
SUBS0000018
repeatedly attempted to downgrade testing of the Horizon system. Again, we
submit that the attempt to deny knowledge of the technical issues underlying
the unreliable cash accounts, which are plainly set out in the contemporaneous
Board documents and Agreements, indicates not only knowledge but an
awareness of the significance of that knowledge.
Below Board level Post Office management were incontrovertibly aware that
the Horizon system being rolled out from 1999 to 2001 would produce
unreliable accounts. Not only was everyone from Stuart Sweetman down well-
versed in the resolution of AI376 through the Supplemental Agreements,
which manifestly permitted unreliability in some branch accounts, some were
also aware of the Project Mentor’s Report dated 18 December 1998. The
headline summary was that the team who prepared it were “deeply concerned
that their findings show a serious problem with the way in which Pathway
hove developed the system. The impact of this is likely to be that there will be
failures to meet essential user requirements, causing the need for extensive re-
work before the system can be accepted and, potentially, operational problems
if the system is rolled out’ *’. It was addressed to Paul Rich, and a hand-written
cover note shows that it was given to David Miller. The latter, at least, knew
that these concerns chimed with Jeremy Folkes’ concerns that he had been
denied access to the Horizon design documents. Mr Folkes’ testified that he
raised this regularly within the technical assurance team”, led by John
Meagher. According to Mr Meagher’s statement, these concerns culminated in
a meeting which David Miller chaired early in 1999, during which he and Mr
Folkes voted to terminate the project.® In his testimony Mr Meagher said that
around February 1999 all the indications were that Horizon would “fail to
#8 POL00038829 pS
49 https:
www. postofficehorizoninquiry.org.uk/sites/default/files/2022-
11/POH%202%20November%202022.pdf p98-102, within which Mr Folkes was referred to his document
written on 1 October 1998, WITNOS970118
5° WITNO4150100 para 29
13
26.
27.
5 httpsy
SUBS0000018
balance satisfactorily” and that amongst the people working on the project at
the time, “everyone knew that”.*! Mr Folkes concerns were then crystallised in
his leaving note, A Reflection on the Past Five Years, which was discussed with
David Miller.*
In these circumstances, there was no excuse for ignoring the Project Mentors
Report. Whatever scepticism there may have been about the motivations of
those who commissioned it, the findings chimed with everything that was
known about the development of the project to date. We submit it is highly
likely to have been discussed with Stuart Sweetman, but evidently neither he
nor David Miller took it any further. This speaks of a complete disregard for
subpostmasters, because as 1999 went on they were faced with an avalanche
of concerns raised by members of the NFSP™, which again chimed exactly with
the Project Mentors finding that the system would “fail to meet essential user
requirements”.
We submit that at the time the Codified Agreement was signed in May 1999,
Post Office management crossed the rubicon. Up until that point, they had not
wanted to continue with ICL if the Benefit Payment Card was scrapped. They
knew that the Horizon system may not have met their needs, but the
Government forced their hand, and they submitted. From that point onwards
they were committed. Mr Roberts, Mr Sweetman and Mr Miller must have
been ready for the inevitable technical issues to arise, just as they had done
over the years leading up to 1999, but after the removal of the Benefits Agency
from the deal, they were on a clock to find alternative income streams. The
regular un-minuted meetings between Mr Roberts and Stuart Sweetman must
‘www.postofficehorizoninquiry.org.uk/sites/default/files/2022-11/POH%2015%20November%202022.pdf
p94
52 WITNO5970123
53 https:
‘www.postofficehorizoninquiry.org.uk/sites/default/files/2022-11/POH%203%20November%202022.pdf
ps
54 eg. NFSP00000471 p10 p12 pp21-22 — NB both Mr Sweetman and Mr Miller attended this meeting.
14
SUBS0000018
have centred on securing those income streams, and how they were going to
make sure that technical problems — and the consequent delays - did not stand
in their way. If Mr Roberts did not know that AI376 was caused by technical
problems, he must have been closing his eyes to the obvious. The next question
is whether he believed that it had been fixed by 2001.
28. Stuart Sweetman’s testimony is of relevance to this question. He had a clear
understanding of the extent to which AI376 had been fixed: the effect of the
Supplemental Agreements fixed the problem at Board level, because errors in
the accounts were now minimal enough to allow the Post Office accounts to be
“true and fair”, but they had not fixed the problem at Branch level, and
subpostmasters still had sums of money going missing from their accounts.®
He was asked whether this was a matter which he discussed with the Board or
with John Roberts and he said this: “I don't recall but it is the sort of thing we
would have talked about because I know, when we reported up to the board,
they expressed concerns about technical issues that were still outstanding.
When we went back to, was this project fit for purpose, I know David Miller
said -- he is quoted as saying it was robust but the board, in its detailed
discussions, expressed concern over the outstanding technical issues, which is
what the board should do and the nonexecutives on the board would have
understood that and then they remitted that to John Roberts and the Chairman
to manage and monitor.”*
29. We therefore submit that the Post Office Board knew that the Horizon system
rolled out between 1999 to 2001 produced unreliable accounts at the Branch
level. If not, they, and specifically John Roberts, closed their eyes to the truth.
The inescapable conclusion, which Tony Oppenheim had already confirmed,
55 https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-11/POH%201%20November%202022.pdf
pp133-135
5 https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-11/POH%201%20November%202022.pdf
pp13-136
15
SUBS0000018
was that POL saw automation as an existential issue. Automation, as Mr
Sweetman in his closing remarks made clear,” was akin to a creed or article of
faith. In the drive for automation, the needs of the subpostmasters, mere
agents, were not considered. They did not even have the facility to challenge
the data Horizon generated
Unreliable accounts to be used as evidence
30. When Jonathan Evans claimed that he was not aware of prosecutions based on
Horizon data taking place, the Chair expressed surprise, pointing out that if he
had been aware of prosecutions taking place using paper evidence he must
have been aware that later prosecutions would rely upon Horizon.** Mr Evans
had tried to distance the Board, when suggesting that local management
ultimately decided to prosecute and exercised its discretion on an individual
case by case basis.
31. We respectfully submit that his evidence was unreliable. He must have known,
and the same applies to all the Board members. They knew that they had an
investigations department and a prosecutorial function. Likewise they knew
that the contracts with subpostmasters provided for the Post Office to claim
shortfalls in the branch accounts from the subpostmaster. Ignorance of these
fundamentals would be unforgiveable.
32. In any event, as stated above, the Codified Agreement shows that at some
stage conscious thought was given to the issue of prosecutions. Mr Sweetman
said that these clauses were inserted by “the lawyers” (although he admitted
that he knew of the prosecutorial function)®, but these provisions are hardly
57 https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-11/POH%202%20November%202022.pdf
p29
58 https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-11/POH%204%20November%202022. pdf
p97
5° https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-11/POH%201%20November%202022.pdf
00147-149
16
SUBS0000018
“boiler plate”. No lawyer lower than Board level would have the authority to
impose these obligations on ICL Pathway unless under instructions.
33. It is therefore damning that there is no evidence that the Board discussed how
Horizon would impact upon prosecutions, investigations and shortfall claims
at any stage, and particularly in the context of AI376. That none of them made
the obvious connection between accounting integrity and the use of those
accounts vis-a-vis subpostmasters is implausible. Their silence suggests
deliberate avoidance of a problem standing in the way of automation. It seems
there was fear of opening a Pandora’s Box, which could delay the move to
Network Banking, Government Gateway, Golden Cloud, and the other airy
ideas for filling the ACT hole. A parallel exists in Colin Baker's evidence: just
as the Post Office Board were determined to automate at all costs, so was he,
and just as they did not choose to air the problems that might stand in the way,
so he did not raise the trauma being experienced by his members at the
Horizon Working Group meetings.”
34. The Board’s silence left those lower down to their own devices when
implementing the Codified Agreement, and hence the confused and highly
problematic situation emerged as set out in Jeremy Folkes’ document Review
of Horizon Cash Account System — Stage 2°'. People were prosecuted right from
the start. Nichola Arch and Tracy Felstead are amongst those we represent, but
Nick Wallis has set out a Timeline at the end of The Great Post Office Scandal,
which states that six people were convicted in 2000, 41 in 2001, and in the years
immediately thereafter it was in the order of 50 to 70 per annum.”
35.To return to the issue of implausibility concerning the Board’s denial of
knowledge, it is plain that the Post Office’s requirements were ‘second fiddle’
59 https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-12/POH%2030%20November%202022. pdf
pp68-70
5! WITNOS970134
© pp4o4-495
17
SUBS0000018
to the BPC and the specifications required by the DSS, but nevertheless, in
September 1998 it was said that “Work has been undertaken under the guise
of Business Audit, with Tom Patterson, to cover various aspects which have
a security and audit interest. Contact with this group should be maintained
to ensure that continued consideration is given to the needs of POCL staff
investigating potential dishonesty. SIE will be responsible for the formulation
of relevant procedures for the conduct of investigations, and prosecution, and
copies of the previous work by Horizon on this subject will be provided.”
36. Moreover, Mr Folkes’ Review of Horizon Cash Account System — Stage 2“ suggests
that concerns re Horizon’s limited data accessibility/provision to POCL’s
investigation service was drawn to the attention of those involved in the
management and rollout of the programme between May and July 1999. The
limitations were potentially undermining POCL’s prosecutorial function, as no
procedures were then in place for evidence production or analysis. Based upon
any analysis of Mr Folkes’ Review, such deficiencies were well known by July
1999.
37. The importance of this document re data retention/provision protocols for
prosecutions and obtaining data from ICL Pathway in 1999 confirms that the
withdrawal of the DSS had gravely compromised POCL’s prosecutorial
capability. Mr Folkes, whose comments were provided within the boxes, noted
(July 1999) that:
“Clearly, a process does need to be agreed between POCL SIE and ICL
Pathway for the commissioning of PACE certification, statements and
court appearances.”
® POL00029136 p2 by Dave King
© WITNOS970134
18
SUBS0000018
38. The conclusion of this document gives rise to obvious concerns so far as the
later decision to prosecute on Horizon data:
“There is a need to ensure that the problems relating to the audit trail for
S&IE investigations and demonstrating that the system meets the
requirements of the Police and Criminal Evidence Act have been impact
assessed as incidents and are considered by the Acceptance and Release
Authorisation Boards if not satisfactorily resolved. In addition, it will be
necessary to consider whether the current level of cash account errors
will affect the accuracy of settlement with clients, when considering the
rate at which the system should roll-out,”
39. We therefore submit that Mr Sweetman and in all likelihood the whole of the
Post Office Board knew that some branch accounts would be unreliable. Their
determination to automate in accordance with the Government's timetable
meant that they simply closed their eyes to the obvious truth that
subpostmasters would be prosecuted as a result.
The Government
Unreliable accounts
40. There is no evidence to suggest that Ministers were aware of AI376, nor that
they had specific knowledge that Horizon would produce unreliable accounts.
The question is whether they were reckless as to the result that the Horizon
system rolled out between 1999 and 2001 produced unreliable accounts. Were
the warning signs of general technical problems such that they were aware of
a risk that would occur, and — in the circumstances known to them — was it
unreasonable to take the risk?
41. The most stark evidence that the Government as a whole was aware of the risk
came from Lord Alistair Darling. He said repeatedly in his evidence and
during 1998/99 that he was not prepared to risk an unreliable system being
19
SUBS0000018
rolled out when millions of benefits claimants had their livelihoods depending
on it.® His officials, most notably Sarah Graham, were clear-eyed about ICL’s
protestations, and given performance to date they simply could not see any
reason to believe that there was suddenly going to be an improvement.® While
there were multiple reasons for the DSS determination to withdraw from the
Benefits Payment Card project, this was a valid, stand-alone reason. Moreover
it applied with equal validity to the risk that any and all aspects of the Horizon
system would be unreliable.
42. Indeed it applied with all the more validity once there was a rush to re-
engineer Horizon after the BA’s withdrawal in May 1999. As the Post Office’s
representative, Mena Rego, said at the start of the first meeting of the
McCartney Horizon Working Group, “There was a need to establish exactly
what POCL was buying, ICL have not allowed access to POCL so that
they could determine this and have insisted that it is not a contractual issue.””
Sir Ian McCartney’s woefully reckless response was to say he was
“emphatically not prepared to accept getting away from the commitment to
2001. Slippage would make the wider discussions on government usage of the
network impossible. If there were problems with software, training etc. then
these should have been flagged up earlier, and must now be resolved in a way
that enabled the 2001 timetable to be recovered.”® As Sir Ian told the Inquiry,
he was “no patsy”. It is clear that Post Office management took his words to
heart, and they were relayed back to the NFSP immediately.” Effectively, he
5 https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-11/POH%2029%20November%202022. pdf
pp125-126
& See for example POLO0028525 p4
57 NFSP00000203 p3
Ibid
® https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-12/POH%2001%20December%202022_O.pdf
p49
7° NFSP00000471 p28
20
SUBS0000018
ensured that neither he nor his Group heard any further mention of technical
problems.
43. Sir Ian and others within Government who decided to press on with Horizon
unreasonably took the risk that ICL would deliver an unreliable product
because they fell into the sunk cost fallacy. This was first raised by John
Meagher, who said that after the BA withdrew the Post Office - and by
implication the Government — should have cut its losses because the project
was “on a death march”.” It was raised again by Sir Geoffrey Mulgan when
he was asked whether there should have been a re-setting of the “black box”
approach to the Horizon solution once the BA withdrew. He said “I think most
involved want[ed] to go for the line of least resistance and get a deal done...
It's a sort of natural human psychology to want to avoid making things
complicated and renegotiating fundamental principles which would then
make it harder to get a deal. And ina sense, that's just a symptom of the larger
feature of this kind of situation, sort of sunk-cost fallacy, as economists call it:
when you put a lot of time into something, you try very hard to keep it alive,
even if the rational decision is to cut your losses.” Mena Rego was a lone and
junior voice calling for a contractual reset to find out what the Post Office was
actually buying.
44. The sunk cost fallacy, and the reckless risk-taking that went with it, went right
to the top. On the one hand, when Sir Geoffrey Mulgan wrote a minute
advising against proceeding with a system which was “possibly unreliable”,
Sir Tony Blair wrote a comment asking for a “clear view” on whether the
system was indeed flawed.” On the other hand, those concerns were simply
™ https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-11/POH%2015%20November%202022.pdf
p96
” https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-12/POH%2001%20December%202022_O.pdf
p17
73 €BO00100001_072 p4 and pS
21
SUBS0000018
allowed to drop. The official response to the minute elided them,” and the
unfounded Peter Mandelson assertion that independent experts had found the
system “robust” was allowed to go unchallenged.”?
45. It is understandable that the Prime Minister and his Ministers at the DTI were
drawn into the sunk cost fallacy. They feared damaging litigation with ICL,
they were concerned about the survival of a famous British company and both
Post Office and the NFSP were talking up the need for urgent action to halt
closures across the Post Office Network. Perhaps most importantly, they were
concerned about the implications for Japanese inward investment, given the
pressure that Fujitsu were able to bring via the Ambassador.” All of these were
valid concerns. However they do not excuse the unreasonable decision to
knowingly risk rolling out an unreliable system across the Post Office
Network.
Unreliable accounts to be used as evidence
46. To a very large extent the arms-length relationship with the Post Office was a
fiction. Alan Milburn admitted as much when he was asked why the
Government foisted the bilateral agreement on the Post Office Board, and he
replied “the real answer is that, in the end, the buck stops with the government
and, indeed, the cheque book is owned by the government, so one was going
to have to write the cheques on behalf of the taxpayer, and that is HMG.””
This was also understood by the Post Office Board. The Treasury had
effectively cut them out of negotiations, and as John Roberts said, the
consequent deal was presented to the Board “at the end of a gun”.” For that
74 €B000000009
75 €BO00100001_067 p2
76 BEISOO00278
77 https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-12/POH%2002%20December%202022_0.pdf
p41
78 https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-10/POH%2020%200ctober%202022.pdf
p161
22
SUBS0000018
reason, Her Majesty’s Government cannot disclaim responsibility for the
implications of it. Ministers and officials at the DTI must have or should have
known that the Post Office conducted its own investigations and prosecutions.
As discussed above, it was obvious that the change from paper accounts to
Horizon accounts would have profound implications in that area, so just as the
Post Office Board should have considered the issue, likewise Her Majesty's
Government, particularly those with oversight at the DTI, should have done
so. While there is no evidence that they were on notice of the specific
implications of AI376, we would submit that at this time of transition to
automated accounts, it was reckless not to exercise greater supervision of what
was a private police force in the form of the Post Office Investigations
Department. Alan Johnson testified that he was personally aware of their
tendency to be over-zealous.”
47. However genuine Mr Johnson's shock at hearing of the sad case of Pamela
Lock, in the case of Chief Constable of the Hertfordshire Police v Van Colle the
House of Lords held that “stupidity, lack of imagination and inertia do not
afford an excuse to a national authority which reasonably ought, in the light of
what it knew or was told, to make further inquiries or investigations: it is then
to be treated as knowing what such further inquiries or investigations would
have elicited.”*° This, we would submit, is illustrative of the onus which was
on the DTI to find out what the POID was going to do to investigate branch
losses once branch accounts moved off paper. It makes it plain that Ministers
knew or ought to have known there was a risk that unreliable accounts would
be used as evidence against the subpostmasters. It would always be wholly
unreasonable to run that risk.
79 https://www.postofficehorizoninquiry.org.uk/sites/default/files/2022-12/POH%2001%20December%202022_O.pdf
p122
5° Per Lord Bingham, [2008] UKHL 50, para 32
23
SUBS0000018
Conclusion
48. This Phase has established knowledge/recklessness at the highest levels of ICL,
the Post Office and Her Majesty’s Government, both with respect to the fact
that the Horizon system rolled out between 1999 and 2001 produced unreliable
accounts, and that those accounts were used as evidence.
Edward Henry KC, Mountford Chambers
Flora Page, 23ES Chambers
Hodge Jones and Allen
7 December 2022
24