POL00363851
POL00363851
From: Anthony de Garr Robinsonf,_
Sent: Sun 18/11/2018 9:14:27 PM (UTC)
To: Andrew Parsonsf._
Ce: i
Subject: Re: Robert conclusions [WBDUK-AC.FID26896945]
Not really - hence my suggestion that that may be no clear answer in para I. All I would say is to say that
something has a small chance of happening is entirely consistent with the claimant s’ case but other descriptive
terms may be less consistent with that case. For example, it is possible to say that the chance of X happening is
very small and the chance of Y happening is also very small and then to combine these two descriptive statements
together in a way that leads to a more forceful descriptive statement - e.g. that Z can only happen if X and Y
have happened, and multiplying together the two small chances of X and Y happening means that in practice the
chance of Z happening is tiny or vanishingly or infinitesimally small or similar.
The other point to mention for completeness is that, although the judge may not accept all the elements and/or
implications of the quantitative analysis that RW is using, if he does not accept these things it does not
necessarily follow that he will not accept RW’s attempts to assign mathematical probabilities to certain things
happening. I don’t have any specific probabilities in mind when I say this, however.
Tony
On 18 Nov 2018, at 20:52, Andrew Parsons i
GRO _ > wrote:
Tony
We'll work through your detailed comments below, however I have one over-arching comment which goes to
your first two points (and some others).
Where measuring the extent of something, I can only see two ways to do that. First, we do it descriptively using
language like low, or very low, or unlikely, etc. Or, second, we do it numerically, by saying this only has an X%
chance of happening, the effect is no bigger than X, etc.
Given the nervousness about relying heavily on numbers, I've been speaking to RW about how to do this
descriptively — which led to the conclusions in my email.
Is there some third way you have in mind? Sorry if I'm being thick, but I don't really know how else we do this.
A
Andrew Parsons
Partner
Womble Bond Dickinson (UK) LLP
d:
m:
t
e
4
POL00363851
POL00363851
Stay informed: sign up to our e-alerts<https://www.womblebonddickinson.com/uk/preferences>
<image2c4f42.PNG>
womblebonddickinson.com<http://www.womblebonddickinson.com>
<image77e63f.PNG><http://www.twitter.com/wbd_uk>
<imagee9 1083.PNG><http com/company/womble-bond-dickinson-uk-Ilp/>
From: Anthony de Garr Robinson [mailto
Sent: 18 November 2018 12:32
To: Andrew Parsons; Simon Henderson {
“Kmailto
Ce: Jonathan Gribben
Subject: RE: Robert conclusions [WBDUK-AC.FID26896945]
Dear Andy,
This summary represents real progress, I think. For what they are worth, my immediate devil’s advocate
thoughts are:
1. How low is very low (a recurrent question in this context to which there may be no clear answer — see my
comment on 4b&C below)? On RW’s own analysis, known bugs with an actual or potential impact on branch
accounts have happened quite a lot, and there have been lots of unknown bugs too.
2. How invariably is invariably (this raises another recurrent question)? We don't have much witness evidence on
this; as far as I recall we only have Parker saying that all the issues Fujitsu identified were fixed. Query whether
there is a conflict between all and invariably?
3, This is really important. But:
1. £300 seems a high figure to me. However, I can see problems with trying to make it lower. In the first place,
the only witness evidence we have on SPM reporting behaviour is from AVDB and her evidence is weak: her
para 187 says that when discrepancies are of a value of “several hundreds of pounds”, she would generally
expect SPMs to contact the NBSC. I’m not sure that this will be enough for RW’s purposes and I wonder
whether it might actually make it impossible for him to reduce his current £300 figure. In the second place, I see
from TG’s account of the local suspense account bug in paras 50 and 51 that even a discrepancy of £9200 was
not reported first time around — the SPM only reported it the second year it arose, in 2013.
2. I find it difficult to see what one does about middling bugs, i.e. bugs with an impact of less than £300 but more
than £10. Your para 4b may well be the best that can be done, but I would prefer not to take refuge in
quantitative arguments when dealing with unknown bugs, if that is possible. As we both know, the critical issue
is what the chances are of the figures contained in any particular set of branch accounts being wrong — i.e.
whether those figures can be taken as good prima facie evidence of the underlying reality. I don't (currently) see
how quantitative arguments help on those questions.
4a. This is also really important. But:
POL00363851
POL00363851
1. The fact that the word very is in square brackets is a little troubling?
2. In any event, how low is very low?
3. Could there ever be an answer to this recurrent question which would make us happy? We all know that very
low is consistent with the claimants’ case, which is based upon the proposition that, with a huge network like
Post Office’s, a low probability is all that they need and is exactly the sort of probability that they would expect.
4. My recollection is that RW wanted witness evidence that the Fujitsu system for identifying bugs was efficient —
ie. it had a really effective system for reviewing issues/discrepancies reported by SPMs, working out that some
of these reports may be the result of a general problem and identifying the cause of the problem, with the result
that if there was a bug causing some reported issues/discrepancies, it was almost certain that the bug would be
identified. However, we have not served such evidence. Even if we could draft some evidence along these lines, I
suspect that it would not be very convincing.
4b&c and 5. See my second comment on para 3 above. As at present advised, I don't see how quantitative
arguments help on establishing the evidential value of the figures contained in branch accounts. RW has an
argument suggesting that, given the low mean value of the financial impacts he has found, the chances of a bug
having a significant impact in any branch is vanishingly small. But the argument is counter intuitive and I doubt
that anyone who not versed in the rules of probability and statistics would be convinced by it. Moreover, we
know that several known bugs have had really big impacts — e.g. the payments mismatch bug (from memory, one
branch had a discrepancy of around £16,000) and the local suspense account bug caused a discrepancy of £9200.
Big impacts are not like lightning hitting the same tree twice, are they?
This is probably a convenient moment to share my thoughts with you on the evidence we served yesterday. The
significant points are emboldened:
1. It’s my job to raise issues for us to fix or deal with, but before doing that I want to take a moment to reflect on
how much was done in such a short space of time. Given the impossible deadline and the striking difficulties we
faced in getting clear instructions from Post Office and Fujitsu, the witness statements we served are a
considerable achievement. Jonny and the rest of the WBD team who worked on them deserve a big pat on the
back.
1. There are a number of issues on which RW asked for witness evidence. I don’t currently understand the full
reasons for or significance of witness evidence in all these areas (e.g. statistics on TCs, TC disputes, NBSC calls
received, handled and passed on) so you'll be pleased to see that I have nothing to say about them. But the
evidence we have served on some issues may not be what RW was hoping for (e.g. SPM behaviour in reporting
small and medium sized discrepancies; the priority given to issues apparently having an impact on branch
accounts, however small; and the existence and enforcement of remote data change controls) and some evidence
seems not to have been attempted (e.g. how good Fujitsu were are spotting possible bugs from issues reported
by SPMs). I’ll be interested to know whether RW tells us that he wants further and better evidence. He obviously
needs to tell us as soon as possible.
1. AVDB: -
POL00363851
POL00363851
* Are paras 18.4 and 18.5 inconsistent with each other on the question whether Horizon associated remotely
inserted transactions with user IDs?
* Para 32 is difficult to follow. Never mind, perhaps we can clarify in chief.
* Paras 33 and 50 and 123 and 172 —I have two queries about NBSC logs, which are relevant to several
statements we have served. First, the lead claimants on several occasions claim that our logs do not cover calls
that they know they made — do we have a rigorous system for logging every call of any materiality made to the
NBSC and for retaining log? If so, we may want to give evidence of this in due course. Second, what about call
logs for the HSD? I’m assuming we have no logs for calls to the other relevant departments called by SPMs, but
given that some relevant calls may have been made to HSD, what logs are made of those calls, how well are they
kept, how long are they kept for, have we reviewed them and might we want evidence on some of these points?
* In para 82, given that she is talking about Legacy Horizon, should AVDB be talking about what the guide for
Legacy Horizon says, not what the Horizon Online guide says? Should we correct this?
* If we have the paper records that Stockdale kept of all transactions done in her branch, why don't they
represent a really good opportunity to show that Horizon was reliable — that it accurately recorded all the
transactions recorded in her paper records, nothing more and nothing less — and also to show that the
discrepancies were not the fault of Horizon but were the fault of the cash kept in her branch not being what it
should have been? Am I missing something, or does this represent a wonderful chance to advance on the enemy?
Query whether this may also be relevant to the question which claims are tried in May 2019...
* Paras 165-177 raise a potentially serious problem. I thought our case was that, although an SPM can only
suspend his payment of £150 or more (because only discrepancies of that amount can be settled centrally), he can
always dispute a discrepancy of any amount. However, that is not what Dawn Phillips said and nor is it what
AVDB now says in paras 165.1 and 168 and 175 — those paras talk about SPMs being able to call the NBSC for
help in identifying the cause of any discrepancy, which is a very different thing from calling the NBSC to dispute
any discrepancy. I raised this problem in the draft I saw of AVDB’s statement and as the wording in paras 165.1
etc was not changed I infer that an SPMs cannot in fact dispute a discrepancy of below £150. However, paras
177.3 and 177.7 seem to say that, since 2005, SPMs have been able to call the NBSC to dispute sub-£150
discrepancies. What is the true position? Can it really be the case that one could not dispute a sub-£150
discrepancy before 2005 but can now? Do we need to correct AVDB’s statement?
* Para 185.5 — what do the words “to confirm” mean? Should those words be deleted?
* Para 187 — see my comments on RW’s conclusions above.
1. Torstein Godeseth:-
* TG’s statement is peppered with references to Gareth Jenkins as a source of evidence. Given that such a large
proportion of TG’s evidence is not within his knowledge but is merely passing on his understanding of what
Jenkins has written or told him, we can expect some uncomfortable questions as to why we have not called the
organ grinder and why the monkey’s evidence should be given any weight at all. I wonder whether the claimants
will have the courage to demand that we call Jenkins as a witness. As I didn't previously appreciate how every
road seems to lead back to Jenkins, I also wonder whether, depending on how much of a fuss they make before
the trial, should we review our prior decision not to call him.
* Para 15 on Cullender Square — do we need evidence that the 20 branches affected were compensated?
* Para 31 —is it a problem that TG does not explain why his previous WS did not reveal the existence of the
extra global user? It can’t be that his statement was exclusively concerned with the current state of Horizon.
* Para 33 — I do not think everything said in Coyne para 9.19 is meaningless. TG seems to misunderstand or
overlook Coyne’s main point in that para. What is our response to that point? Should we consider doing an
analysis of whether global users had done any transactions in any of the claimants’ accounts? This could be a
major opportunity to advance on the enemy or it could be a shot in our foot, but is it worth considering?
* Para 40 is hard to follow, perhaps we can clarify in chief.
POL00363851
POL00363851
* Paras 50-51, see my comments on RW’s conclusions above.
* Para 61 — can we find out what losses were identified in the outreach branches as having been caused by the
Dalmellington branch? Would this help or hurt RW’s conclusions about how large a bug has to be before an
SPM is likely to report it?
1. Tracy Mather:-
* Para 14 — does RW think we need evidence on the reliability of Credence/whether it has its own bugs resulting
in false TCs?
* Para 19 — why couldn’t Mather deal with the question whether the various FSC teams she worked in (a) ever
needed ARQ data and (b) ever refrained from requesting such data because Post Office might be charged a fee
for it? This is the key issue and it seems odd that we don't have n FSC witness addressing it.
1. Paul Smith:
* I don't yet fully understand why RW wants these TC statistics, but that does not matter. However, some of the
statistics appear to show a high number and proportion of disputed TCs and a high number/proportion of upheld
disputes — see in particular para 23. Is this damaging?
* What does para 26 mean? Is the witness intending to suggest that only the cheque investigation team does
good and thorough investigations?
1. Andy Dunks: why have we produced this statement/? Did RW ask for it?
1. Paul Parker: -
* There is much to like in this statement. We need to ascertain whether RW Is happy with the things he has said
about
i, the drafting and accuracy of KELs and Peaks (e.g. paras 41.3 and 61.5);
ii. there being more errors in Horizon in the years 200-2004 (para 42.3 — is this consistent with some of RW’s
graphs of KELs over time, for example?);
iii. deactivated KELs (para 61.6 — I didn't even know that such things existed, have the deactivated KELs been
shown to RW and Coyne?);
iv. deleted KELs (para 61.9)’
v. deleted Peaks (para 62.6).
POL00363851
POL00363851
* Paras 20.2 and 58.5 read well for us, but Parker does not actually give evidence of the change controls that
were in place or of how rigorously they were enforced. I think that Coyne has identified a report suggesting that
they needed improving and were not properly enforced. Will Parker be able to handle questions about this? Do
we need further and better evidence on these points? What does RW think?
* Para 57.1 is weak. Might there be other errors that Roll is referring to that involved data being remotely
downloaded in other ways?
* Para 64: The Coyne KEL table appears not to refer to some KELs that Coyne refers to — e.g. the Callendar
Square bug, referred to in Coyne para 5.5. Coyne does not give the KEL reference for that bug, but there must
have been one. Does the table only refer to KELs for which Coyne has given the KEL reference number? If so,
does this explain why Coyne thinks he’s referred to 76 KELs but the table only refers to 58? Are there 18 further
bugs referred to in Coyne’s report for which he has not given the Coyne ref? If so, should RW wish to ask Coyne
what the missing KEL refs are, or has he already got too much on his plate?
Best wishes,
Tony
Sent: 16 November 2018 19:45 .
To: Anthony de Garr Robinson < Simon
Henderson (¢
GRO
Ce: Jonathan Gribbe
Subject: Robert conclusions [WBDUK-AC.FID26896945]
Tony, Simon
I've been speaking with Robert over the last few days about his conclusions on the extent of bugs. This is the
current conclusion (not in exactly these words) Robert is proposing:
1. Horizon is robust (see section 7) and as a consequence the extent of bugs in Horizon which cause errors in
branch accounts is very low.
2. Ifa bug is detected which could affect branch accounts, the evidence shows that invariably it was fixed and /
or that its effects with remedied such that there was no lasting impact on branch accounts.
3. Known bugs - significant bugs (over £300 per branch per month- can probably lower this figure) will very
likely be detected, and as a result were fixed / remedied (as per point 2).
4. Unknown bugs —
* The probability of undetected bugs occurring is [very] low.
* Tn any event, if they did occur the extent of their financial impact is extremely small compared to the impact of
detected bugs.
* Micro bugs (£10 or less) do not have significant impact on branch accounts.
5. Taking the above conclusions together, in an attempt to quantify the effect of bugs, at worst known and
unknown bugs account for less than 0.05% of shortfalls.
6. Conclusion:
* Horizon is extremely unlikely to be the cause of shortfalls in branch accounts (tying back into Issue 1).
* When looking at any particular set of branch accounts taken from Horizon, it is fair to assume as a general
starting proposition that those accounts reflect the transactions input by the Subpostmaster or his assistants to a
[extremely] high degree of accuracy, unless there are special circumstances in a case.
POL00363851
POL00363851
I think this is closer to where we want to get but I'll keep working on him.
Any thoughts? Resist the temptation to draft!!!
A
Andrew Parsons
Partner
Womble Bond Dickinson (UK) LLP
d:
m:
t:
e:
Stay informed: sign up to our e-alerts<https://www.womblebonddickinson.com/uk/preferences>
<image001.png>
womblebonddickinson.com<http://www.womblebonddickinson.com>
<image002.png><http://www.twitter.com/wbd_uk>
<image003.png><https://www. linkedin.com/company/womble-bond-dickinson-uk-Ilp/>
Please consider the environment! Do you need to print this email?
The information in this e-mail and any attachments is confidential and may be legally privileged and protected by
ly is authorised to access this e-mail and any
+ as soon as possible and delete any copies.
Unauthorised us n, distribution, publication or copying of this communication or attachments is
prohibited and may be unlawful. Information about how we use personal data is in our Privacy
Policy<https://www.womblebonddickinson.com/uk/privacy-policy> on our website.
Any files attached to this e-mail will have been checked by us with virus detection software before transmission.
Womble Bond Dickinson (UK) LLP accepts no liability for any loss or damage which may be caused by software
viruses and you should carry out your own virus checks before opening any attachment.
POL00363851
POL00363851
Content of this email which does not relate to the official business of Womble Bond Dickinson (UK) LLP, is
neither given nor endorsed by it.
This email is sent by Womble Bond Dickinson (UK) LLP which is a limited liability partnership registered in
England and Wales under number OC317661. Our registered office is 4 More London Riverside, London, SE1
2AU, where a list of members' names is open to inspection. We use the term partner to refer to a member of the
LLP, or an employee or consultant who is of equivalent standing. Our VAT registration number is
GB123393627.
Womble Bond Dickinson (UK) LLP is a member of Womble Bond Dickinson (International) Limited, which
consists of independent and autonomous law firms providing services in the US, the UK, and elsewhere around
the world. Each Womble Bond Dickinson entity is a separate legal entity and is not responsible for the acts or
omissions of, nor can bind or obligate, another Womble Bond Dickinson entity. Womble Bond Dickinson
(International) Limited does not practice law. Please see
www.womblebonddickinson.com/legal<http://www.womblebonddickinson.com/legal> notices for further details.
Womble Bond Dickinson (UK) LLP is authorised and regulated by the Solicitors Regulation Authority.
This email has been scanned by the Symantec Email Security.cloud service.
For more information please visit http://www.symanteccloud.com
This email has been scanned by the Symantec Email Security.cloud service.
For more information please visit http://www.symanteccloud.com
This email has been scanned by the Symantec Email Security.cloud service.
For more information please visit http://www.symanteccloud.com
This email has been scanned by the Symantec Email Security.cloud service.
For more information please visit http://www.symanteccloud.com