POL00112903
POL00112903
Privileged
WOMBLE
BOND
DICKINSON
Bates & others v Post Office
Board Sub-committee: 21 February 2019
Speaking note
Headlines
e The trial in March is focused on the Horizon system. There
are 15 questions for the Judge to address, which fall under
three core topics:
1. Is Horizon robust or does it cause shortfalls in SPM
branches?
2. Are PO and FJ able to manipulate Horizon data in
secret and without the knowledge of SPMs?
3. And finally, a group of factual questions about certain
aspects of the system that if determined now will be a
useful foundation in other trials.
e In my assessment, the evidence weighs in favour of Horizon
being a robust system. There is little risk of the Judge saying
that Horizon is bad but he may find that it is only "ok" and/or
that in some areas it could be improved (for example, he
may find that it can sometimes encounter problems (e.g. in
the recovery process following a system crash) or he may
find that some of Post Office’s or Fujitsu’s support
processes could be criticised (e.g. there is no system to
monitor the transaction correction process with a view to
identifying problems for which Horizon may be responsible
and/or some problems in Horizon are not fixed as quickly as
would be desirable).
e This trial will come down to whose expert is more credible.
We believe the evidence of our expert, Dr Worden, has a
better methodology and is more cogently evidenced that the
Cs expert, Mr Coyne. His approach has been to pick as
many holes as possible without considering them or their
impact too carefully, which lacks balance.
AC_154286889_1 1
POL00112903
POL00112903
e The secondary battle will be keeping the Judge focused on
what is important. The Cs will attempt to draw attention
away from the experts (where they are somewhat weaker)
and towards eye-catching issues such as FJ's ability to
remotely change Horizon data and side issues such as a
perpetual complaint that PO has provided insufficient
disclosure of documents. They will use this to paint Post
Office and Fujitsu in a bad light but these points don't go to
the core of whether Horizon works.
e We also have to factor in a good deal of judicial risk as this
Judge as acted in ways that are truly extraordinary as you
will have already heard from David Cavender on the
Common Issues trial.
Our case
e The core argument we will run in Court is as follows.:
o Horizon like all IT systems is not perfect but it is a very
good system.
= This is not about proving that Horizon is perfect
and that there are no bugs in Horizon.
= Inthe context of this litigation, it is about showing
that Horizon accurately records transactions the
vast majority of the time so that PO, SPMs and
the Court can safely start from an assumption that
the branch accounting information held on
Horizon is sound.
o No-one has found a fundamental flaw in the system or
its support processes. We are nowhere near a
situation where this a bad or even average system. In
practical terms, this is a decision between whether
Horizon is good or very good.
o One would expect imperfections and bugs to occur in
any large system of this sort. But Horizon has been
designed and supported carefully over many years,
with many countermeasures in place to spot and
AC_154286889_1 2
POL00112903
POL00112903
correct problems as they may arise. The service
history of Horizon is comprehensive and shows these
countermeasures in action time and time again. They
were plainly effective in practice.
o The highest the Cs can put their case is that of the
many thousands of issues which inevitably arise on a
system of this scale, they have identified a couple of
dozen bugs over 20 years which according to them, did
or may have caused loss to a SPMR:
= Several of the "bugs" they claim to have found are
no such thing. They are not bugs at all. Indeed,
they often support PO's case that there were
effective countermeasures.
= Of the actual bugs found, many have no bearing
on the critical issue in this case, i.e. those bugs
either did not affect branch accounts or they were
picked up and fixed relatively quickly in the
ordinary course of Horizon’s operation
= Many of those bugs are also triggered by
relatively unusual events: SPMs doing
unexpected things with the system and some
cases doing things that are in breach of PO's
operating rules.
o Cs argue that there could be thousands of undetected
bugs lurking in the system, but this is based on little
more than the (obvious) point that it is not possible to
say that all bugs are always detected whenever they
appear. Cs have no evidence to justify the supposition
that that there could be many lurking bugs having a
significant impact on a significant number of branch
accounts. PO’s expert has a persuasive analysis
suggesting that there is nothing like the number of
undetected bugs in Horizon that there would need to
be in order to explain even a material portion of the
losses alleged by the Cs. The Cs’ expert does not
have a rival analysis and, although he tries to pick
holes in PO’s analysis, he does not really grapple with
it.
AC_154286889_1 3
POL00112903
POL00112903
o The simple fact is that, taking into account the size of
the Horizon estate and the period of its operation, the
volume of bugs Cs have found is tiny. About 20 bugs
over an 18 year period, against a back drop of 30,000
active users and 50m transactions a week.
= Even if one assumes that the Cs case is entirely
correct, Horizon would still be a robust system.
= This is PO's expert’s clear and well-explained
opinion.
= Cs expert cannot even bring himself to give a
clear opinion that Horizon is not robust, which is
very telling as to the truth of the matter.
e This is the narrative that we are proposing be put to the
Court and around which we will arrange our detailed
submissions and the evidence.
e The strength and formulation of this narrative may need to
be adjusted in light of the Cs supplemental expert report
served on 1 Feb this year.
o Up to Christmas, we were feeling confident about PO's
position. The Cs first expert report was poor and we
had answered it comprehensively in Dr W’s report.
o On 1 February, they served their supplemental report.
It is anything but supplemental. It raises a whole new
range of issues, including about 15 new alleged bugs
in Horizon that have not been raised before.
o Consideration was given to whether to seek an
adjournment to the trial but this would have been
impossible with our Judge who is obsessed with speed
of resolution at all costs, even if that creates an
unfairness. An adjournment would not have been
supported by Cs.
o The legal team and FJ are now working ferociously to
investigate these new points. Progress so far has
been good. Extracting relevant and reliable information
from FJ in a timely fashion can be difficult, but we are
hopeful, but not certain, that we will have credible
counter-points to raise to most of Coyne's new attacks.
AC_154286889_1 4
POL00112903
POL00112903
e My assessment of the case is therefore premised on FJ
being able to supply good ammunition to counter-act
Coyne's late arguments. This seems likely but until the
investigations are complete we cannot be certain and we
are about a week away from that point.
Risk 1: High bar.
e PO has necessarily set its case very high: that Horizon is
robust and extremely unlikely to cause shortfalls. It has
done this for two reasons:
e First, this was the finding from Deloitte's investigation into
Horizon at the outset of the litigation and on which PO
premised its legal position.
e Second, PO's operating models are built on the assumption
that Horizon works.
o The heart of this litigation is whether SPMs should be
liable for shortfalls. The Cs main line of attack is that
the shortfalls were not caused by them but by bugs in
Horizon.
o Inits interactions with SPMs, PO's starting assumption
is that Horizon works. It habitually looks first for errors
in branch before looking at Horizon. Its contracts with
SPMs are structured to in line with this approach, and
the experience of PO and FJ is that this starting
approach is the right one to take.
o This approach can only be sustained if Horizon is so
reliable to justify a starting, but rebuttable, assumption
that Horizon is accurate.
o If Horizon does not meet this bar, then that may require
PO to test the accuracy of Horizon in relation to every
branch shortfall going forward before seeking to
recover any monies.
o It would also pose a risk to its decisions over the last
18 years in relation to the recovery of shortfalls and the
AC_154286889_1 5
POL00112903
POL00112903
termination of SPMs: which heightens the possibility of
successful claims within the litigation.
e Although this bar is high, Dr Worden has felt comfortable
giving an opinion saying that Horizon is robust and
extremely unlikely to cause shortfalls in branches, which
aligns with Deloitte's earlier work.
Risk 2: Witnesses
e Unlike the Common Issues trial, this trial turns heavily on the
quality of witness evidence.
e In any litigation there is a risk that witnesses may break
under cross examination. That applies to both our evidence
and the Cs’ evidence
e The Two key risk areas are:
o Dr Worden relies on statistical models to show that
incidence of bugs in Horizon is tiny when compared to
the size of the system. Some Judges dislike data and
maths-heavy analysis and so this approach may fall on
deaf ears.
= We have mitigated this risk by ensuring that Dr
Worden also includes a qualitative opinion
alongside his quantitative views.
o The main evidence for Post Office is being given by
technical personnel from FJ. We have encountered
difficulties in the evidence from FJ.
= Ina relatively small number of places, the
evidence has been incorrect and needed to be
corrected through further statements.
= We are therefore concerned that under cross-
examination further cracks in FJ's evidence could
be exposed.
= We should make clear that FJ have been very
cooperative — this is not due to a lack of
engagement from them, particularly now.
AC_154286889_1 6
AC_154286889_1
POL00112903
POL00112903
= The problem is that they are being asked to give
factual evidence on issues that arose years,
sometimes, decades ago on highly technical and
often obscure points. In some cases, the points
they are addressing are hypothetical and so FJ
are having to speculate rather than draw the
answer from first-hand experience.
This problem is exacerbated by the fact that
several of the points now being emphasised by
Cs’ expert were either only raised or only given
emphasis in the last month or so (in a
supplemental witness statement by Richard Roll,
who worked for FJ about 15 years ago, and in Mr
Coyne’s “supplemental” expert report). For
reasons related to the way in which the Judge has
managed these proceedings, Cs’ case has been a
moving feast. The result is that not all the factual
issues are now in play that were addressed in the
FJ and PO witness statements previously served.
We have mitigated this through extensive
discussions with FJ, probing their evidence as far
as we can as lawyers with no technical knowledge
and I have even met with FJ direct on a number of
occasions. Nevertheless, it remains a real
concern whether FJ's witness will withstand cross-
examination.
The sorts of areas where FJ might be attacked
should not, in our view, necessarily lead to PO’s
overall case being undermined. For example, if
the Judge finds that FJ used its powers of
privileged access to edit or delete branch data
more frequently and with less discipline than FJ
previously said, that does not of itself mean that
there were problems with the Horizon system. So
the case does not stand or fall with the
performance of FJs witnesses on every issue, but
it needs to be recognised that if they perform
POL00112903
POL00112903
poorly it gives the Judge a route to finding against
PO if he wants to.
Risk 3: Remote access
e Weaved into the heart of the Cs case is a theory that PO
and FJ are frequently meddling with branch transaction data
in secret and that this is either deliberately or casually
causing false shortfalls to appear in branches. This has
become known as the "remote access" problem.
e The difficulty is that PO's and FJ's case on remote access
has changed over the years.
o During the mediation scheme, FJ advised PO that
remote access was not possible and PO passed that
message on to Second Sight.
o At the end of the mediation scheme, the FJ position
changed to there being a limited capability to add new
transactions to a branch's accounts, but with no
capability to edit or delete existing data.
o Deloitte's work then revealed that the editing or
deleting existing data was possible in certain
circumstances.
o Fujitsu’s witness statements indicate that various forms
of remote access have been and are possible, more
than previously admitted.
o The Cs’ expert is now saying that there are even more
ways that data can be edited and deleted — and we are
still investigating these allegations with FJ.
e Viewed in its proper context, the Remote Access problem
adds little substance to the basic question whether Horizon
is robust. But it gives the Cs lots of damaging forensic
points.
o There is no evidence of widespread data manipulation
but there are circumstances where FJ do need to
change data to correct problems in the system. This is
normal, but the history of this issue and FJ’s apparently
inability to give even POL a full and account of the
AC_154286889_1 8
fe)
POL00112903
POL00112903
remote access facility and its use feeds oxygen to a
conspiracy theory of widespread secret meddling.
It is difficult to imagine any rational reason why FJ
would use this capability to maliciously manipulate
branch data, or that it would do that so carelessly to
cause widespread problems.
But the changing position on FJ's capability to edit and
delete transaction data exposes PO to criticism of
covering up some untoward activity. The Cs will say
that they have had fight to draw out the truth from PO.
The remote access point can also be used to tarnish
the credibility of FJ's witnesses.
e At trial, this will attract disproportionate attention. Human
nature will attract the Judge and the media to this topic.
Our approach is that sunlight is the best antiseptic. We
won't be denying the existence of remote access capabilities
but we will be saying that it’s a red herring when it comes to
deciding whether Horizon is robust. But for this purpose, we
are seeking to produce a clear and comprehensive account
of the remote access facilities that have been available to FJ
over the last 18 years. This is not as easy as it we could like
it to be.
Risk mitigation
e Because of the above risks, we are building risk mitigations
into the legal strategy.
First, we will draw a clear distinction between the old version
of Horizon that existed before 2010, and new Horizon that is
currently in place. The Cs’ attacks are mostly on old
Horizon. If the Judge were to find that old Horizon was not
robust, this would be far from ideal in the context of the
litigation but it should cause less operational disruption to
PO as that system is no longer used.
Second, where possible, we are laying down markers that
the steps taken to reach this trial were unfair. The Cs have
AC_154286889_1
POL00112903
POL00112903
(again) served evidence that is not consistent with what the
Judge said he wanted. Mr Coyne's supplemental report
effectively makes a new case. With a more engaged and
even-handed Judge, such things would not have happened.
But they have happened and the possibility of appealing the
Horizon Issues judgment is limited. The only likely ground is
procedural unfairness and that is a very difficult ground to
make out, particularly in the context of group litigation, in
which Managing Judge’s are expected to fashion unusual
procedures to ensure the swiftest and most efficient means
of resolving the points in dispute.
e Third, we are taking urgent steps to investigate whether we
can produce some short witness evidence addressing some
of the issues which the Cs are now placing emphasis. FJJ
are being cooperative, but this is not an easy or quick
process, we are not yet in a position to determine how
useful it will be and there is no guarantee that the Judge will
allow any further evidence to be adduced.
e Fourth, due to the level of reliance on the quality of witness
testimony, we are planning to only submit short opening
submissions. This then allows us flexibility to adapt our
closing submissions to the quality of evidence given during
trial. But this does also come with the risk of criticism by the
Judge that we are not being clear about our case on
awkward issues and/or that when we address them have,
we will be raising new points late in the day.
Conclusion:
e In conclusion, as matters stand, I believe that the Judge will
likely find that Horizon is a robust system, but that we will
get some criticism about F's ability to remotely change data
and about the processes they apply in identifying and fixing
bugs in Horizon. This assessment assumes that FJ can
provide answers to most of Coyne's late attacks.
e The outcome will however turn heavily on what actually
happens at trial with the witnesses perform well or not.
AC_154286889_1 10
POL00112903
POL00112903
e I will therefore be keeping my assessment of the position
under continuous review over the next few weeks and
during the trial and will flag if I believe the position has
changed.
AC_154286889_1
POL00112903
POL00112903
Note for possible Q&A after above
Tony — I need to draft answers to this.
What's the problem with disclosure?
Why did we not object to Coyne 2 / the Cs evidence?
How come FJ's evidence was wrong?
[I WOULD ADD — WHY ARE THERE SAID TO BE GAPS IN
OUR EVIDENCE?]
e Can you not give me a more definitive view on the merits?
AC_154286889_1 12