WITN09950100
Witness Name: Gregg Nicholas Rowan
Statement No.: WITN09950100
Dated: 23 August 2023
THE POST OFFICE HORIZON IT INQUIRY
First Witness Statement of Gregg Nicholas Rowan
I, Gregg Nicholas Rowan of Herbert Smith Freehills LLP, Exchange House, Primrose
Street, London EC2A 2EG WILL SAY as follows:
PART 1 - INTRODUCTION
1: lam a partner of Herbert Smith Freehills LLP ("HSF") and, at the date of making
this statement, the Recognised Legal Representative ("RLR") for Post Office
Limited ("POL") in the Post Office Horizon IT Inquiry (the "Inquiry").
2. I make this witness statement in response to a request made by the Inquiry
pursuant to Rule 9 of the Inquiry Rules 2006, which I received on 31 July 2023
(the "Rule 9 Request").
3. The purpose of this statement is to respond to the questions set out in the Rule
9 Request and to provide my evidence in relation to the three disclosure issues
set out in the directions of the Chair dated 6 July 2023 (as amended on 14 July
2023) ("July Directions") and to explain certain aspects of the systems and
processes that are in place to seek to ensure that disclosure will be made and
issues of a similar nature will be avoided in the future.
Page 1 of 63
WITN09950100
4. This statement is structured as follows:
Section I Title Pages
1 Introduction 1-3
2 Background 4-11
3 The approach to disclosure by POL 11-21
4 Search terms, family review and deduplication 21-45
5 The Inquiry's questions and my answers 46 - 65
5. There is an Annex to my statement which describes in detail the remediation
steps that have been undertaken and are in hand in respect of the search terms
issues which concerns the “Policy Review" (explained below at paragraph 68)
and the review of documents that were deduplicated at an "item level". The
Annex also deals with assurance work being conducted by my firm (explained
below at 138). Since the remediation process is complete in respect of the
family review issue, and can be described succinctly, I have addressed that in
the body of my statement.
6. The explanations in my statement and the Annex in respect of POL's
remediation and assurance activities are up-to-date as at 16 August 2023. As
at the date of this statement, POL’s remediation and assurance activities are
ongoing and POL intends to produce additional documents on a rolling basis,
including in the weeks commencing 21 and 28 August 2023. POL also intends
to write to the Inquiry on 31 August 2023 to provide a further update on these
activities as at that date (and in the meantime to continue its practice of
providing weekly updates to the Inquiry by letter).
Page 2 of 63
WITN09950100
7. Except where I indicate to the contrary, I make this statement on the basis of
facts and matters within my own knowledge. Where the facts and matters in this
statement are within my own knowledge, they are true. Where the facts and
matters are not within my own knowledge, I have identified the source or
sources of my information and I believe such facts and matters to be true. As I
explain below, there is a very large team working on this case within my firm
and where appropriate I have consulted team members to assist in preparing
this witness statement.
8. Peters & Peters Solicitors LLP ("P&P") have had a substantial role in POL’s
disclosure to the Inquiry insofar as it relates to private prosecutions. Where I
refer to their work, I have based my explanations on the knowledge and
understanding of members of my team from working alongside P&P, and from
discussions that I and members of my team have had with P&P in the
preparation of my statement.
9. Many of the communications between my firm and POL, and other
communications in connection with POL's participation in the Inquiry, are
subject to legal professional privilege. I do not have authority from POL to waive
privilege over those communications and nothing in my evidence is intended to
do so.
PART 2 - BACKGROUND
My role and the role of HSF
10. I joined HSF (then Herbert Smith) in 2007 and became a partner in 2014. I am
admitted in England & Wales and Australia, and I have worked from our
Page 3 of 63
da.
12.
13.
WITN09950100
London, Melbourne and Tokyo offices. For several years before joining the firm,
I practised as a barrister on the Northern Circuit from Kings Chambers in
Manchester.
The Inquiry was established on a non-statutory basis in September 2020. My
firm had been advising POL before that date, since April 2019, on various
matters primarily relating to the Horizon-related issues, in particular the group
litigation and the establishment of the compensation scheme. My firm continues
to advise POL in the latter capacity. The compensation-related work is
undertaken by a separate team within the firm led by my partner Alan Watts.
When the Inquiry was non-statutory, POL was managing its strategy and
preparations for the Inquiry, and its day-to-day responses to and interactions
with the Inquiry, ‘in-house’. During that period my firm undertook specific, ad
hoc tasks on instructions from POL.
The Inquiry was converted to a statutory inquiry from 1 June 2021. My firm was
retained by POL to act on its behalf in the Inquiry in September 2021. In a letter
dated 30 September 2021 (WITN09950101), POL informed the Inquiry that my
firm had been retained. By this point, POL had begun to receive Rule 9
Requests. My partner Andrew Lidbetter was designated as the RLR in October
2021, and I took over the role in April 2022, having worked on the matter since
October 2021. The scope of services provided by my firm to POL has varied
over time and has mainly been focussed on Phases 1 to 4. It principally
includes:
a. engagement with the Inquiry on POL's behalf;
Page 4 of 63
14.
15.
16.
WITN09950100
b. day-to-day management and oversight of POL's involvement and
participation in the Inquiry;
Cc. assistance with and/or management of disclosure requests from, and
responses to, the Inquiry; and
d. management of witness evidence provided to the Inquiry.
Of these, as I explain below, assisting POL in responding to the Inquiry's
disclosure requests is a very significant proportion of the work that we do.
As the RLR for POL, I have overall responsibility in my firm for the matter. I am
assisted by three partners (two of whom have worked on the matter for some
time; one of whom is a recent addition to the team). There is also a very large
team of lawyers, trainees, paralegals, legal analysts, project management
personnel and other staff working on the matter: approximately 150 people in
total currently and approximately 280 over the course of the matter. We instruct
a counsel team currently comprising two King's Counsel and 6 junior barristers.
My overall responsibility for the matter means that I lead the team, I am the
primary point of contact with POL, and I am ultimately responsible for the advice
given by, and steps taken by, my firm, including in respect of disclosure. That
said, due to the size of the matter, there is an allocation of responsibility
amongst the team in relation to different workstreams. As part of that, since
around April 2022, another of the partners on the team has led the work on
Phase 4 and my involvement has been limited. I have become more involved
in Phase 4 recently in view of the disclosure issues. Insofar as my statement
Page 5 of 63
WITN09950100
relates to Phase 4, I draw upon information provided by members of my team,
POL and certain others, including P&P.
17. Whilst as a partner specialising in dispute resolution I am familiar with electronic
disclosure, as is usually the case with disputes practitioners, I am not an
eDiscovery specialist: that is a separate specialism for which firms with large
disputes practices like mine now have dedicated teams who can be engaged
adjacent to the provision of legal services. My firm is not the eDiscovery
provider to POL in this matter; that is KPMG. Insofar as I comment on specialist
eDiscovery matters in my statement, I have consulted specialist colleagues
(who were not working on the matter at the time of the events described in my
statement).
Change in the RLR
18. Whilst at the time of making this statement I am POL's RLR, by the time of the
hearing on 5 September 2023, I anticipate that the role will have transferred to
Chris Jackson of Burges Salmon LLP.
19. As the Inquiry is aware from my firm's letter of 3 February 2023
(WITN09950102), POL decided early this year to replace my firm as its legal
representatives in the Inquiry, and POL has since appointed Burges Salmon
LLP.
20. Following an extensive handover process, on 2 August 2023, my firm wrote to
the Inquiry to request that the Chair designate Mr Jackson as the RLR with
effect from 1 September 2023 (WITN09950103). After the RLR designation
Page 6 of 63
WITN09950100
transfers, I understand that POL will continue to call on my firm as required
including to assist with Phase 4 until the conclusion of that phase.
The roles of others
21. Together with POL itself, there are others involved in the disclosure process,
whose roles I summarise briefly below. Each of the principal service providers
(HSF, KPMG and P&P) contracts directly with POL and do not have
responsibility for each other's services.
(a) POL's Inquiry team
22. My firm receives its instructions principally from the Inquiry team at POL. This
is a team made up of lawyers and project managers, some of whom are and
have been secondees from law firms.
23. Fintan Canavan joined POL as Inquiry Director on secondment from DAC
Beachcroft in late October 2021 and remained in post until December 2022.
Diane Wills joined the team in January 2023 and is the current Inquiry Director.
Until recently the Inquiry Director reported to Ben Foat, POL's Group General
Counsel, who was also involved in instructing us. I am aware that Mr Foat, Ms
Wills and Mr Canavan have also received Rule 9 Requests.
24. The composition of POL's Inquiry team has changed over time. It is now
significantly larger than in 2021 and for most of 2022, and it has been
reorganised, including by the creation of three new senior roles reporting to the
Inquiry Director: the Operations and Strategy Director in October 2022 and the
Inquiry Heads of Legal in January and April 2023.
Page 7 of 63
25.
26.
27.
WITN09950100
(b) KPMG
POL's eDiscovery provider is KPMG, who are directly accountable to POL for
their work. KPMG host electronic documents for POL in various databases on
an eDiscovery platform known as Relativity. They also process and upload to
Relativity documents harvested from POL and other sources. In addition, they
are responsible for the technical elements of any document review and
production. There are a number of aspects to that, but at a broad level it involves
advising on eDiscovery matters, receiving search instructions, running
searches over databases (or subsets of documents in one or more of the
databases) in accordance with those instructions to produce a pool of
documents for review, deduplicating documents and/or applying other
refinements, migrating documents (e.g. from the large processing database to
a 'review' database), batching documents for review, resolving technical issues,
preparing document exports for production to the Inquiry and uploading
productions to the Inquiry's file sharing platform.
The KPMG team is led by Paul Tombleson and I am aware that he has also
received a Rule 9 Request for the purpose of the 5 September 2023 hearing.
(c) Peters & Peters
P&P are retained by POL to provide advice and assistance in relation to criminal
law and procedure, including in particular in the ongoing criminal appeals. The
P&P team includes lawyers at P&P, as well as a number of junior barristers
whom they instruct to undertake review work amongst other things — I use the
term "P&P" to refer to the whole team.
Page 8 of 63
28.
29.
30.
31.
WITN09950100
P&P are also retained by and assist POL in respect of Inquiry-related work as
it relates to criminal law, criminal investigations and private prosecutions. This
includes, but is not limited to, providing documents and information, which they
have collated as part of the Post-Conviction Disclosure Exercise ("PCDE"), to
assist POL in responding to requests from the Inquiry.
In late 2021, P&P were instructed by POL to undertake preparatory work
relating to issues 109 to 161 in the Inquiry's Completed List of Issues ("CLI").
This work comprised the identification of material already available as part of
the PCDE that was relevant to those issues; preparing narrative documents that
provided information relevant to the issues; obtaining additional documents and
information; and conducting the further application of search terms and/or
review of material related to aspects of the PCDE as necessary to collate
materials relevant to the issues. P&P began this work in January 2022.
When POL received Rule 9 Requests from the Inquiry relating to private
prosecutions, my firm worked with P&P to prepare POL's responses to the
requests. The level of involvement of P&P has varied, depending in particular
on the information and/or documents sought by the Rule 9 Requests. For
certain Rule 9 Requests, POL and my firm have utilised P&P's prior work
product.
For the Rule 9 Requests in which P&P have been involved, members of my
team have worked closely with members of theirs, both in terms of agreeing the
general approach on how to respond to the requests and in advising POL as
our mutual client. Until very recently, my firm has not generally been involved
with P&P’s conduct of reviews, including their devising of search terms, their
Page 9 of 63
WITN09950100
consideration of whether or not to review family documents, or their instructions
to KPMG (and their predecessor eDiscovery provider) in respect of
deduplication when running searches. Exceptionally, however, my firm was
involved in agreeing with P&P the search that was run in relation to Legal
Advices (Question 19 of Rule 9 Request (14) (INQ00002008)) as part of the
Policy Review, and proposing search terms that were run as part of the data
scoping analysis set out in detail in my firm's letter to the Inquiry dated 6
February 2023 (WITN09950109) (the "Data Scoping Analysis" described in
more detail at paragraph 136 below), and in particular the first part of the Data
Scoping Analysis relating to the extent to which PCDE search terms covered
the issues raised in CLI 49 (Knowledge).
(dq) Burges Salmon and Fieldfisher
32. Burges Salmon LLP, who are assisted by Fieldfisher LLP, are now also
instructed by POL and are assisting POL on various matters. They will have
conduct of POL's participation in the Inquiry from 1 September 2023 onwards,
subject to Mr Jackson being designated as RLR. Burges Salmon LLP and
Fieldfisher LLP have been involved in aspects of the disclosure remediation
and the disclosure assurance work.
PART 3 — THE APPROACH TO DISCLOSURE BY POL
Introduction and general observation on POL's approach
33. Paragraph 9 of the July Directions raises the question of the "systems and
processes" that the Chair has said should be in place to facilitate disclosure. I
therefore address at a high level how my firm, together with POL and those
Page 10 of 63
WITN09950100
others who are involved in the disclosure process, have typically approached
disclosure on receipt of a Rule 9 Request.
34. The detail of POL's approach to previous Rule 9 Requests has been explained
in four disclosure statements (dated 27 May 2022 (POL00114170ds), 18
October 2022 (POL00114173ds), 30 November 2022 (POL00114176ds) and
12 January 2023 (POL00114177ds)), supplemented by two witness statements
from Mr Foat (POL00114188ds, POL00118164ds).
35. I wish to begin by making a point which I believe it is important for the Inquiry
to hear from me in the present context. Neither I, nor any member of my team,
have been asked, encouraged or placed under pressure at any time to withhold
any relevant documents from the Inquiry, and nor are we aware of any instance
in which POL has done so itself. It is clear to me from my interactions with
employees at POL that POL genuinely wants to achieve transparency with the
Inquiry and there has never been any discussion of which I am aware about
finding ways to keep problematic documents back — whether the racist and
offensive ‘Appendix 6' (POL00115674), to which I return below, or any other
document. Insofar as I refer below to challenges in disclosure, they are all to do
with process, complexity, cost, expertise and timing: nothing whatsoever to do
with the content of any documents.
Volume and complexity of disclosure
36. I have worked on a number of different disclosure exercises throughout my
career, in a number of different contexts, particularly civil litigation and
international arbitration. All large disclosure exercises have their challenges.
Page 11 of 63
37.
38.
39.
40.
WITN09950100
However, based on my prior experience, this has been a particularly complex
exercise.
To date, the Inquiry has issued 47 Rule 9 Requests and three Section 21
Notices to POL, which in total contain over 365 discrete questions (many of
which contain sub-questions). The Inquiry has also made requests for
documents and information other than by way of Rule 9 Request or Section 21
Notice.
There are 218 issues in the CLI. The Rule 9 Requests made to support the
Inquiry's work in considering these issues have often overlapped (in that POL
has been responding to more than one Rule 9 Request at once — sometimes
several).
There are over 60 million electronic documents on the Relativity platform
maintained for POL by KPMG across five principal databases. Due to the way
in which these databases originated and how documents have been uploaded
to them (and indeed how the documents concerned were held within POL's
business originally), a significant amount of duplicative material exists both
across the principal databases and within them. In addition, POL holds a vast
volume of electronic data on its IT systems which has not been harvested, and
hard copy documents stored in various different locations across the United
Kingdom.
Generally, the hard copy and electronic documents held by POL are not well
known to the individuals on the POL Inquiry team: they often have limited
knowledge of POL's document repositories and where potentially relevant
documents might be found. Those who do understand where documents might
Page 12 of 63
WITN09950100
be located are diverse in location and function, and in many cases are no longer
employed by POL.
41. A significant proportion of my firm's work on this matter has involved assisting
POL with disclosure. To date, we have reviewed over 900,000 electronic
documents on POL's Inquiry Database (often more than once, either as part of
one multilevel review for a Rule 9 Request or across multiple reviews) and, I
estimate, many hundreds of thousands of hard copy documents, originally
contained in boxes in POL's various storage sites. Through that process, POL
has produced over 127,000 documents to the Inquiry. The documents are
diverse in context, format and complexity, and they cover a time period of nearly
30 years.
The standard for searches
42. The Inquiry’s Disclosure Protocol sets out a standard for searches, which is that
searches should be "reasonable in all of the circumstances" and
“comprehensive, thorough and rigorous" (paragraphs 9 and 18). Section
21(4)(b) of the Inquires Act 2005 also refers to the standard of "reasonable in
all the circumstances" in the context of a claim brought by a person under that
sub-section.
43. What is reasonable in all the circumstances, and what is demanded by a
comprehensive, thorough and rigorous search, must be addressed within the
context of each Rule 9 Request: there is no ‘one size fits all’. As well as the over
60 million electronic documents held on the electronic disclosure platform
managed by KPMG, each Rule 9 Request can result in new documents being
uploaded to Relativity from the very substantial body of electronic and hard copy
Page 13 of 63
WITN09950100
documents held by POL. To state the obvious, it would be impossible for all of
these documents to be reviewed for each Rule 9 Request, so, as I describe
below, systems are used for each Rule 9 Request to produce a set of
documents which is then to be reviewed for relevance to that request.
44. In considering reasonable search parameters, there are also practical
considerations. The Rule 9 Requests set tight deadlines: typically 7, 14 or 28
days, although sometimes longer and sometimes shorter. Extensions of time
have often been sought from the Inquiry and granted, albeit not always for the
full period requested, and until this year POL generally sought to avoid seeking
extensions of time. In addition, POL's costs are funded from the public purse.
Limiting costs has also been a factor in decisions made about disclosure and
the scope of work which POL requests from us more broadly. The best way to
exemplify this is in the context of the choice of search terms, which I address
below.
The process for each Rule 9 Request
45. The Chair has expressed concern that there might have been a "mechanistic"
approach to disclosure by POL, and in paragraph 9 of the July Directions, he
noted that responding to a request for documents or information is about more
than just devising search terms.
46. I donot believe that the approach taken by POL to the Rule 9 Requests that my
firm has assisted with was mechanistic. As I hope to demonstrate here, it is
about more than search terms. They are only one of several tools that we and
others deploy on behalf of POL, albeit an important tool when dealing with large
volumes of diverse electronic data.
Page 14 of 63
47.
WITN09950100
My firm has approached Rule 9 Requests (and Section 21 Notices) from a
number of different angles and used a range of methods to identify documents
that are potentially responsive. These methods are typically selected by our
senior associates, with whom I have regular discussions: there are usually at
least a few such discussions per week and, at busy times, many more than that;
and there are also many emails exchanged on the subject as well. The methods
we use include:
a. requesting information and/or specific documents from members of the
POL Inquiry team, who in turn would make enquiries in the business at
POL;
b. members of my team and/or the POL Inquiry team speaking to
individuals in the business at POL to understand who might know
relevant information, where relevant documents might be held, and what
steps might be taken to respond to the requests / notices;
c. reviewing known relevant documents to identify possible lines of enquiry
such as potential custodians and search terms;
d. drawing upon work previously undertaken by POL or my firm that is
relevant to the request / notice;
e. as the Inquiry has progressed, having regard to documentary and
witness evidence and other information that has emerged during the
process;
f. liaising with POL's former professional advisors, such as Womble Bond
Dickinson (UK) LLP;
Page 15 of 63
WITN09950100
working with POL's other current advisers, such as P&P;
conducting searches across previously harvested documents hosted on
Relativity (across the databases described in paragraph 39 above),
using date ranges, custodians and/or parties, keywords and/or any other
relevant search parameters, as appropriate;
searching across POL's Mimecast database (POL's email archiving
system);
harvesting any new repositories and custodian mailboxes;
searching hard copy archives, such as those operated by Oasis Group
and at the Postal Museum;
searching sites across the POL network at which hard copy documents
are or were stored;
considering metadata (e.g. coding applied in respect of the Group
Litigation) and secondary records (e.g. database summaries shared by
POL);
using functionality which is built into Relativity (the eDiscovery platform
on which we review documents), such as its ability to detect similar
documents so that they may be grouped together and ranked by
similarity; and
‘dip’ samples of repositories and/or investigative search results (this
could be used, for example, to target searches and/or refine search
parameters).
Page 16 of 63
48.
WITN09950100
The exact system and process we use on behalf of POL may differ for specific
Rule 9 Requests, depending on the circumstances and terms of the request or
notice, the documents or types of documents sought, whether POL has
previously directly identified and provided relevant documents (for example, in
the context of P&P’s work in relation to the PCDE), the nature of the repositories
to be searched and the guidance and instructions provided by POL. However,
in broad terms, the approach to the Rule 9 Requests in respect of which my
firm had conduct is as follows:
a. We send each request to POL and discuss with POL what is likely to be
required, including which of the relevant methods above are likely to yield
relevant documents, identifying the repositories that need to be searched
(hard copy, electronic or both), the extent to which harvests of potentially
relevant documents have already been undertaken, formulating keyword
search terms, and the harvesting, processing and uploading to Relativity
of any additional materials.
b. We instruct KPMG to process and upload any additional materials, then
to apply keyword search terms and/or any other relevant parameters
(such as date ranges) as appropriate across Relativity or specific
contents within it.
c. KPMG identify the documents that are responsive to the search terms
and/or other parameters, apply any exclusions (e.g. documents that
have already been produced to the Inquiry) and batch documents in
order for them to be reviewed.
Page 17 of 63
WITN09950100
We review the documents and tag them for relevance and privilege. We
typically utilise a sequence of first-level review, second-level review and
sometimes third-level review. Alongside the review, there is also typically
a system of quality control ("QC") checks in place, which assesses (i) a
sample of relevant documents to ensure the correct documents are
being tagged as relevant at first level, and (ii) a sample of irrelevant
documents to ensure relevant documents are not being incorrectly
tagged as irrelevant. As I return to at paragraph 61 below, frequently, the
review process will cause us to revisit the scope and nature of the search
terms and issue further instructions to KPMG.
The first level review is typically conducted by reviewers in our ALT
(Alternative Legal Services) team or an associate team in London and/or
elsewhere in my firm's global network. The first level reviewers are
briefed (usually in writing and orally) on the terms of the relevant request,
the context and anticipated purpose of the request, the approach to be
taken to relevance and privilege, and how to raise queries.
Where first level review is conducted by members of the ALT first level
review team, QC checks are typically carried out by more senior ALT
team members before the documents are reviewed in London.
Second level review and, where appropriate, QC and/or third level
review, is carried out by the London team, by individuals with appropriate
seniority and familiarity with the subject matter.
Page 18 of 63
WITN09950100
h. Throughout the process, supervision is provided by, and queries are
escalated to, senior members of the London team and ultimately to POL
as appropriate.
i. Once the pool of relevant and disclosable documents is finalised,
production instructions are sent to KPMG who (following completion of
pre-production checks) upload relevant documents to the Inquiry's file
sharing platform.
iF My firm sends a covering letter to the Inquiry, explaining that the
documents have been uploaded and the approach that has been used
to collate them.
The broadening of the Inquiry
49. The nature of the Inquiry has changed over time. When my firm was retained
by POL to act as its legal representative in the Inquiry, it was understood that
the Inquiry hoped to report its findings by autumn 2022. Most recently, its
published timetable indicates that Phase 4 will conclude by the end of 2023,
with the remaining evidentiary hearings for Phases 5 to 7 taking place in 2024.
50. The way that the Inquiry has developed has meant that the Inquiry is
considering a very substantial, and increasing, volume of documentation, and
a large, and increasing, number of witnesses have been required to give
evidence. It is fair to say that POL's initial planning for disclosure (and the
Inquiry generally) did not anticipate this evolution, and POL had undertaken
limited preparation by the time the Inquiry began to issue Rule 9 Requests.
Page 19 of 63
WITN09950100
51. Once we were retained, we undertook preparatory work at pace with a view to
enabling POL to respond promptly to Rule 9 Requests. Work began in earnest
to understand POL's electronic and hard copy data universe, which it is fair to
say is vast and highly complex, and the attendant practical challenges.
52. We wrote to the Inquiry on 15 October 2021 (WITN09950104) in relation to the
sources of documents that we had at that point identified as containing
potentially relevant documents to the Inquiry's Provisional List of Issues and
identified known limitations on the data held by POL (for example, that POL's
email data for the period prior to its separation from Royal Mail Group in 2012
was substantially incomplete due to the way in which email data was transferred
as part of the separation).
PART 4-— SEARCH TERMS, FAMILY REVIEW, AND DEDUPLICATION
53. The relevant background to the hearing on 4 July 2023 and the Chair's July
Directions is set out in the second witness statement of Mr Foat dated 21 June
2023 (POL00118164ds). The issue was that a suite of eight historical
documents used by POL's security and investigations team had been provided
by POL, in an apparently quick and straightforward fashion, in response to a
Freedom of Information Act request in May 2023, but the full suite had not been
provided to the Inquiry in response to any previous Rule 9 Request.
54. Appendix 3, which was part of the suite of documents, had been produced to
the Inquiry on 14 April 2022 (POL00038452) in response to Rule 9 Request
(11) (INQ00002007). A near final version of Appendix 3 (POL00105216) was
also disclosed in response to Question 18 of Rule 9 Request (14)
(INQ00002008). One of the other documents in the suite was Appendix 6 (also
Page 20 of 63
WITN09950100
labelled "Identification Codes") (POL00115674), which contained racist and
offensive language. This document was not disclosed to the Inquiry until after
POL was alerted to it in responding to the Freedom of Information Act request
referred to at paragraph 53 above. The suite of documents, including both
Appendices 3 and 6, was not responsive to Rule 9 Request (11)
(INQ00002007), although the suite was responsive to Rule 9 Request (14)
(INQ00002008) and in any event ought to have been disclosed to the Inquiry in
2022, as the Chair noted in his July Directions.
55. In the July Directions, the Chair refers to three issues under the headings
"Search terms", "Family documents" and "Deduplication". Those are the terms
that I use in this statement: when I refer to the "three disclosure issues", I am
referring to those three issues.
The use of search terms in principle
56. As described at paragraphs 47 and 48 above, various methods are used to
produce a set of documents to review for each Rule 9 Request so as to ensure
that a reasonable response to the request can be provided within the deadline
imposed (subject to any extensions being sought from and granted by the
Inquiry). Search terms are not always used, and when they are, they are usually
only one of a variety of tools used.
57. It is fair to say that for those Rule 9 Requests which sought documents
originating from the 1990s and the 2000s, which tended to relate to Phase 2
issues, my team were particularly dependent on using search terms to identify
documents (although members of my team did also visit the Postal Museum
and POL's various hard copy storage sites on many occasions to search for
Page 21 of 63
58.
59.
WITN09950100
hard copy documents). There were very few people still employed by POL from
that early period who could help us identify documents through other methods.
In my view, the use of search terms is inevitable and indeed desirable if a
particular Rule 9 Request requires scrutiny of POL's databases of electronic
documents assembled for the Inquiry, which as referred to above, contain over
60 million documents. Full manual review of all documents would be impossible
and unhelpful: the number varies significantly depending on the length and
density of the documents, but typically, an ambitious target for a well-briefed
lawyer reviewing documents for relevance and privilege could be 250 per day,
but with a dense and complex document set it might be 200 or lower; each Rule
9 Request must be completed in a matter of days or weeks, with time being
allowed within that period for second (and where appropriate, third) level review;
and, many documents held in the databases are not relevant to the Inquiry. In
addition, KPMG's default protocol is to require the documents identified for
disclosure 48 hours before the deadline for uploading to the Inquiry's platform.
We use a number of techniques to try to increase the effectiveness of search
terms. These include (i) searching for all forms of words that begin with a certain
root (e.g. investigat* - which would capture investigate, investigation,
investigating, investigated etc); (ii) searching for multiple words in a document
(e.g. "Policy" AND "investigat*" — which would capture a document where both
those words appeared but not only one of them); (iii) searching for words that
are used in close proximity to one another (e.g. "Investigat*" w/10 "Policy" —
which would capture a document where those two words appear within ten
words of each other in a document); and (iv) testing the results (as discussed
Page 22 of 63
60.
61.
62.
WITN09950100
below). We might also specify other criteria against which the search terms
would be run, such as document date, file type and/or custodian.
In making decisions around search criteria (including search terms), there is a
balance between seeking to maximise the breadth of the results on the one
hand, and, on the other hand, ensuring that the review set is manageable and
therefore that the resulting review for relevance and privilege can be achieved
in the time available and relevant documents can be produced to the Inquiry
within the timeframe set. The choice as to the breadth of search terms is also
informed by the degree of confidence held about other techniques which have
already been utilised by POL, HSF and others including P&P (such as, for
example, obtaining the documents directly from a discrete source, where
available).
The search terms used for each Rule 9 Request are not necessarily fixed: a set
of search criteria (including search terms) will be sent to KPMG, who will
implement them over one or more databases and send the results to my team.
We will examine how many ‘hits' are returned per term, how many items have
been identified by KPMG for review cumulatively and whether the results might
reflect any issues with the criteria (for example, if the terms are returning a
significant volume of ‘false positives'). If necessary, the terms will be refined
further, to arrive at a sensible balance. The search terms are generally shared
with POL, although I do not believe that we received specific input or feedback
from POL in relation to the search terms now under discussion.
The devising of search terms was usually done by senior associates in my
team, working in an iterative way with the search results provided by KPMG as
Page 23 of 63
WITN09950100
each search is conducted. I or another partner would be consulted as the senior
associates considered appropriate. In any event, that exercise is, in my
experience, usually (and better) done by those with a closer understanding of
the documents and databases. If a search was proving particularly difficult or
onerous, we would usually discuss that, not least because that would impact on
the time and cost of the exercise, which were points I or another partner would
need to be involved in communicating to POL and ultimately the Inquiry if an
extension was required. I take full responsibility for the search terms that my
team devised, which I believe were carefully considered.
Using search terms in relation to the relevant Rule 9 Requests
63. There are two potentially relevant Rule 9 Requests in relation to Appendices 3
and 6: Rule 9 Requests (11) and (14).
(a) Rule 9 Request (11)
64. Rule 9 Request (11) (INQ00002007) contained 68 sub-questions. Some
responsive documents were identified without needing to conduct searches
using search terms (for example, some documents were obtained directly from
POL's Company Secretary). Where searches were required in order to respond
to the sub-questions of Rule 9 Request (11) (INQ00002007), the specific search
terms that were used are set out in the Interim Disclosure Statement dated 27
May 2022 (POL00114170ds). They involved 19 different formulations of
detailed search terms.
65. The particular sub-questions relating to prosecutions or investigations policies
were 15 and 46(a), which sought:
Page 24 of 63
WITN09950100
Question 15: "The Minutes of the Audit, Risk and Compliance Sub-
Committee of 11 February 2014 [your ref POL0018054] refer to a report
which outlined the proposed changes to the prosecutions policy and a
paper to explain the most appropriate way to communicate the
prosecutions policy. Please provide copies of the same and copies of all
iterations of the prosecutions policy since 1999 that are in POL’s custody
or control."
Question _46(a): "In reference to the Minutes of the Audit, Risk and
Compliance Committee of 26 January 2021 [your ref POL-0018094]: a.
These refer to an investigations policy that was circulated. Please
provide copies of the same and copies of all iterations of the
investigations policy since 1999 that are in POL’s custody or control."
66. Issue 109 in the Inquiry's CLI is:
“What policies and guidelines were adopted during the relevant period
regarding the bringing of private prosecutions against SPMs, managers
and assistants alleged to be responsible for shortfalls shown by the
Horizon IT System?".
67. We used three principal methods to identify responsive documents.
68. First, it was clear that these documents would be sought by the Inquiry, so in
January 2022, a workstream (the "Policy Review") was scoped by P&P,
together with POL, and on which my team was asked to comment, one aspect
of which was targeting the documents referred to in CLI 109. This was part of
the preparatory work relating to issues 109 to 161 in the CLI that POL instructed
Page 25 of 63
69.
70.
WITN09950100
P&P to undertake, and one of a number of workstreams that POL has instructed
P&P to carry out in connection with criminal matters that are relevant to Phase
4.
As part of the Policy Review, P&P utilised the work that they had undertaken
as part of the PCDE, including searches that had been run in 2020 to find
prosecution and criminal investigation policies and procedures. I understand
that the approach to the Policy Review was adopted in anticipation of POL
receiving Rule 9 Requests, and because POL recognised that the exercise
would be complex and take considerable time. When Rule 9 Request (11)
(INQ00002007) was issued by the Inquiry on 28 February 2022, P&P were in
the early stages of the Policy Review, and we liaised with them, as well as with
POL, in order to draw upon their work when responding to Questions 15 and
46(a).
Second, we liaised directly with POL to try to find relevant documents. As part
of that, we considered POL's response to a potentially relevant request that
POL had received under the Freedom of Information Act in 2020, which sought
the "current guidance and/or rules provided to POL prosecutors and
investigators". (For the avoidance of doubt, this was not the Freedom of
Information Act request which led to the production of Appendix 6
(POL00115674) in 2023: that Freedom of Information Act request was made
over two years later). As part of our work at that time, we had liaised extensively
with POL and others to try to understand the policies that were in place at
various times. We had been in direct contact with POL's Head of Security
Operations, and we also sought information directly from P&P and Cartwright
Page 26 of 63
71.
WITN09950100
King (who had historically been involved in reviewing POL's Prosecutions
Policy). Members of POL's Freedom of Information Act request team also spoke
directly with various individuals in POL's business, including the team lead
investigator, the manager of the intelligence and administration team, and the
individual who subsequently identified the suite of appendices containing
Appendix 6 (POL00115674) when POL responded to the Freedom of
Information Act request in 2023. At that time, we were provided with various
iterations of the prosecutions policy and versions of a ‘conduct of criminal
investigations’ policy.
Third, my team searched electronic documents, including using search terms
and other criteria that my firm devised, as follows:
The following search was conducted over all of the materials which POL
had at that time harvested from the Postal Museum and Oasis archives:
- Search Terms: "Policy" AND ("Investigat*" or "Prosecut*" or
"Whistle")
- Date Range: N/A
In addition to conducting searches across the materials collated from the
Postal Museum and Oasis archives, POL also arranged for the following
search to be conducted across the entire POHIT and CCRC databases.
- Search Terms: ("Investigat*" or "Prosecut*") w/10 ("Policy")
- Date Range: N/A
Page 27 of 63
72.
73.
WITN09950100
- Other parameters: Review limited to Microsoft Word and PDF
documents only.
All responsive documents were manually reviewed.
Looking back at these search terms and search criteria, whilst I appreciate that
Appendix 6 was not responsive to them, even with the benefit of hindsight and
knowing what I do now about Appendix 6 and the suite of documents to which
it belonged, and doing my best to be objective, I consider these to be a
reasonable set of search terms (and search parameters / criteria) for the team
to have used having regard to Questions 15 and 46(a). The Inquiry's request
was for iterations of two specific documents: the prosecutions policy and the
investigations policy; and the search terms sought to home in specifically on
prosecution or investigation policies. Appendix 6 is in many ways an
extraordinary document. It contains no words about policy, procedures or
guidance. Using search terms targeting policies and procedures, it would only
be possible to generate search terms to which Appendix 6 responded if one
knew about the existence of the document first.
My firm first conducted these searches in March 2022, and on 31 March 2022
POL produced to the Inquiry 307 documents in response to Questions 15 and
46(a): 179 were identified as being responsive to Question 15, 102 as
responsive to Question 46(a), and 26 as responsive to both. Six further
documents responsive to Questions 15 and 46(a) were produced on 6 April
2022 after an error had been identified in the production set provided on 31
March 2022.
Page 28 of 63
74.
75.
76.
WITN09950100
To assist the Inquiry, we set out in our letter accompanying the production dated
31 March 2022 (WITN09950105) two tables listing dated versions of the
prosecution policy and the investigations policy which POL had been able to
locate. We also listed in a separate annex a series of other documents which
P&P had identified which related to the investigative and prosecutorial
processes. Our letter explained that whilst these documents were separate from
the prosecution policy and the investigation policy that the Inquiry sought in
Questions 15 and 46(a), POL would be willing to produce them, if the Inquiry
wished. One of these documents was a version of Appendix 3 that P&P had
identified (described as ‘Post Office Ltd Security Operations Team —
Compliance — Guide to the Preparation and Layout of Investigation Red Label
Case Files: Offender reports and Discipline Reports’).
My firm then undertook further searches in April 2022, after being informed by
KPMG that a substantial volume of documents stored on the Criminal Cases
Review Commission ("CCRC") database had been uploaded onto Relativity
without any extracted text (which is the text used for search terms and retrieval
purposes), and therefore they had been missed by the searches that had been
run in March 2022. The first tranche of documents identified following that
exercise was produced on 14 April 2022: 168 documents were produced in
response to Question 15; 101 in response to Question 46(a); and 27 in
response to both. One of those documents was Appendix 3 (POL00038452).
The balance of the documents was produced on 30 May 2022. The delayed
production of these documents was due to a technical error, which KPMG
brought to our attention on 13 May 2022, that had occurred when they reran
Page 29 of 63
77.
78.
79.
80.
WITN09950100
the searches in April 2022. This had caused 2,000 documents to be omitted
from the documents provided by KPMG to my team to review at that time. These
additional documents were reviewed by my team in late May 2022, resulting in
10 documents being produced in response to Question 15; 38 in response to
Question 46(a); and 2 in response to both.
Finally, one document was produced on 19 October 2022 as being responsive
to Question 46(a); and a standalone version of Appendix 7 (POL00094200 ) was
produced on 2 December 2022, which had been identified in a review of hard
copy documents.
In total, across these productions, 660 documents had been produced in
response to Rule 9 Request (11) (INQ00002007), Questions 15 and 46(a), plus
the standalone Appendix 7 (POL00094200), before the production on 30 May
2023 of the suite of documents including Appendix 6 (POL00115674).
I have sought to understand the circumstances which led to Appendix 3 but not
Appendix 6 being disclosed to the Inquiry in response to Rule 9 Request (11)
(INQ00002007), and my firm's involvement. I have done so having particular
regard to the fact that my firm reviewed Appendix 3 in the context of our work
on this Request and therefore, on the face of things, Appendix 6 was within our
reach.
I understand the position to be as follows. A number of versions of Appendix 3
were reviewed by my firm's ALT team and marked irrelevant as part of the initial
review of search-term responsive documents in March 2022. It is apparent from
data on Relativity that none of these were in families containing Appendix 6.
Page 30 of 63
81.
82.
83.
WITN09950100
Subsequently, as part of the further review of materials in April 2022 that I
describe at paragraph 73 above, a senior associate in my team reviewed a
number of versions of Appendix 3, some of which were in families containing
Appendix 6. The reviewer tagged one version of Appendix 3 as 'relevant' and
that document was produced to the Inquiry on 30 May 2022 (POL00038452).
I have spoken with the reviewer who in April 2022 reviewed the versions of
Appendix 3 which had Appendix 6 in their family. He explained that he was
familiar with both the Prosecutions Policy and the Investigations Policy, having
been closely involved in preparing POL's original response to Rule 9 Request
(11) (INQ00002007) in March 2022, and having reviewed iterations of both
documents in that context. His assessment, so far as he can recall now, was
that he considered that Appendix 3 was not responsive to Rule 9 Request (11),
Questions 15 or 46(a), because it was not an iteration of either of those policies.
However, he considered that the document might be of interest to the Inquiry,
and he tagged the document as relevant on that basis. I believe that that
approach to Appendix 3 was correct. He did not look at the family documents
of Appendix 3, and did not review Appendix 6. In fact, as far as I am aware, no
member of my team reviewed Appendix 6 until May 2023.
(b) Rule 9 Request (14
Rule 9 Request (14) (INQ00002008) contained 42 sub-requests (or questions).
Question 18 requested:
"Policies and guidelines (applicable in the four countries of the United
Kingdom) relating to the bringing of private prosecutions against
subpostmasters and other end users (including Crown Office
Page 31 of 63
84.
85.
86.
WITN09950100
employees) alleged to be responsible for shortfalls shown by Horizon
during the relevant period."
My firm was responsible for preparing responses to the questions relating to
civil recoveries (Questions 1 to 17 inclusive). For the questions relating to
criminal prosecutions (Questions 18 to 42 inclusive), which include the question
that I have just cited, P&P provided documents and information to my team,
which we used to prepare the narrative responses to the Inquiry with P&P's
assistance; and POL produced the documents to the Inquiry. POL responded
to the Inquiry in relation to Question 18 on 12 August 2022.
My understanding is that in considering Question 18, P&P utilised the Policy
Review workstream that I explained at paragraph 68 above, and also the work
they had undertaken as part of the PCDE, including searches that I understand
from P&P had been run in 2020 (before my firm's involvement in the Inquiry and
before the Inquiry had been established) to identify POL's policies and
procedures relating to prosecutions and criminal investigations. I understand
that that exercise had involved P&P running search criteria that they had
designed.
P&P conducted interviews with current employees in POL (formerly in the
Security Team) to ask questions about (amongst other things) policies and
procedures. This included an interview with the same employee in the Security
Team who provided the documents to POL's Freedom of Information request
team in May 2023. The responses given by the employees that related to
policies and procedures fed into the Policy Review. P&P also engaged
Page 32 of 63
WITN09950100
extensively with POL in relation to identifying the current versions of the
prosecution and investigation policies.
87. Forresponding to Rule 9 Request (11) (INQ00002007) Questions 15 and 46(a),
my firm instructed KPMG to conduct some additional searches to build upon
the information that had been provided by P&P. For responding to Rule 9
Request (14) (INQ00002008) Questions 18 to 42, we understand that P&P
instructed KPMG to conduct further searches, but we did not duplicate that work
given their expertise and history with the criminal procedure documents.
Reviewing family documents
88. A ‘document family’ is a group of connected documents — a common example
of a document family is an email, its attachments and any embedded
documents. The covering email will often be referred to as the ‘parent’ and the
attachments referred to as the ‘children’. The attachments may be other emails
(which in turn, have their own attachments) or they may be other types of files
such as pdf documents, spreadsheets, or zip files. The family members may in
turn contain further documents. In this way, a document which responds to a
search may have 2 family members; but it might also have 20, or 200 or more.
89. In general, POL produces only relevant documents to the Inquiry — not, as a
default, the entire family of each relevant document. I understand that there is
no express requirement (whether in statute or guidance) to produce full families
of documents to the Inquiry; and it would be positively unhelpful to do so: it
would add to the Inquiry's own document review burden by an order of
magnitude.
Page 33 of 63
90.
91.
1
WITN09950100
I believe that this is consistent with the standard for searches set out in the
Inquiry's Disclosure Protocol, which is that searches should be "reasonable in
all of the circumstances". It is also consistent with the communication from the
Inquiry of 11 July 2023" (INQ00002010) which confirmed that it wanted relevant
documents (in that specific case, covering emails), not all covering emails, and
that "everything would depend on context".
Our approach to reviewing family documents has varied between Rule 9
Requests, according to the scope of the request and relevant circumstances.
By way of illustration:
a. In Rule 9 Request (5), the Inquiry made wide-ranging requests for legal
advice on certain subjects. In circumstances where the Inquiry was
making a request for any written external legal advice, a decision was
taken to review full families of keyword responsive documents tagged as
relevant, because it was anticipated that both emails between POL and
its legal advisors, as well as their attachments, could contain legal
advice.
b. In the context of Rule 9 Request (7), it was clear that a review of full
families was not required, because the request sought copies of specific,
named historical documents. When searches were undertaken to identify
documents that were responsive to this Rule 9 Request, the primary
focus was on reviewing hard copy materials identified at the Postal
Email from the Inquiry dated 11 July 2023 timed at 09:56 (INQ00002010), in response to a letter from my
firm dated 7 July 2023 (WITN09950106).
Page 34 of 63
WITN09950100
Museum and Oasis. Such documents do not generally have families in
the same way as electronic documents, such as emails.
For Rule 9 Request (10) (INQ00002002), my team reviewed full families
with keyword responsive documents in some instances and we only
reviewed keyword responsive documents for other aspects of the
request. The approach to reviewing families was an operational decision
taken in the overall context of the review exercises and was conducted
under our general authority from POL to progress disclosure. Where full
families were not reviewed, it was considered at the time a sound
approach, noting the various factors mentioned above in Part 3 above
including, relevantly, the difficulty that would arise from reviewing family
documents in the timescales set by the Inquiry (as extended on a number
of occasions) and the cost of doing so.
For Question 27 of Rule 9 Request (11) (INQ00002007), we were looking
for one specific document, namely a note that the General Counsel was
asked to circulate at a particular meeting of POL's Audit, Risk and
Compliance Committee. In order to identify this specific document, we
asked KPMG to search the mailbox of the General Counsel with a date
range of roughly four months from the date of the meeting, using suitable
search terms. This returned 81 documents (674 when including full
families). Our approach was to first review the 81 documents — but to
review the family documents if the specific document could not be found
amongst them. Shortly after we started the review, the document was
Page 35 of 63
92.
93.
94.
WITN09950100
found and so the review of family documents was not required (though
the set of 81 documents were reviewed in full for completeness).
For those questions within Rule 9 Requests (11) and (14) that were not related
to private prosecutions, generally my firm’s approach was not to review families,
although there were certain questions where we considered family documents
should be reviewed. That did not include Questions 15 and 46(a) of Rule 9
Request (11) (INQ00002007). Those Questions sought a specific report in
respect of POL's prosecution policy (referred to in minutes of a particular
meeting of POL's Audit, Risk and Compliance Committee), all iterations of "the
prosecution policy" since 1999 and all iterations of "the investigations policy"
since 1999 (which had been referred to in minutes of a particular meeting of
POL's Audit, Risk and Compliance Committee).
In respect of Questions 18 to 42 in Rule 9 Request (14) (INQ00002008), which
concerned criminal prosecutions, I understand that at the time of P&P's PCDE
review in 2020, they had within their data set a version of Appendix 3 which was
a standalone document and other versions which included Appendix 6 within
the same family. I understand P&P only reviewed the standalone version of
Appendix 3 as part of the PCDE, and this was the version that was disclosed to
the Inquiry in response to Rule 9 Request (14) in August 2022 (POL00105216).
P&P have confirmed to me that, as far as they are aware, they did not review a
version of Appendix 6 until May 2023.
I understand from P&P that they do not know why only the standalone version
of Appendix 3 was reviewed in 2020 and not the other versions. It seems at
least possible that this was caused by the approach to deduplication by POL's
Page 36 of 63
95.
96.
97.
WITN09950100
predecessor eDiscovery provider; but we are unable to say that with any
certainty, and therefore unable to say that deduplication was a reason why P&P
did not review Appendix 6.
It should be noted that when a decision is taken that family documents need
not be reviewed, that does not mean that they cannot be considered: unless
they have been removed in the deduplication process (or some other technical
process) the entire family of each document to be reviewed is still available to
our reviewers. Whilst not reviewing family members will be the default when
that decision has been made, if the reviewer considers that the content of a
document within a family is such that there might be other documents within
that family which should be reviewed to understand the context of a responsive
document, then they will proceed to conduct that further review.
On every occasion where family documents were not reviewed, this approach
was adopted having regard to the over-arching purpose of enabling POL's
disclosure obligations to the Inquiry to be fulfilled. To review family documents
would involve the review of a material number of additional documents, and it
was anticipated that a very large proportion of these documents would be
irrelevant. This was on the basis that the relevance rate generally drops away
considerably when non-keyword responsive documents are reviewed.
In respect of the specific suite of documents which has led to the hearing for
which I am providing this statement, Appendix 3 responded to the search terms
devised for Rule 9 Request (11) (INQ00002007), it was reviewed, and it was
disclosed to the Inquiry (POL00105216). Appendix 6 did not respond to the
search terms. It did occur on more than one occasion in the same family as
Page 37 of 63
WITN09950100
Appendix 3; but because we were not reviewing family documents for the
purposes of responding to Questions 15 and 46(a) of Rule 9 Request (11),
Appendix 6 was not reviewed, and was not therefore disclosed to the Inquiry.
98. On4 July 2023, Mr Foat gave evidence that the deduplication process may
have been a factor in why Appendix 6 was not reviewed by my firm. That was
based on the understanding at the time Mr Foat gave evidence that only one
version of Appendix 3 had been reviewed by us and that others had been
removed in the deduplication process. We have been informed by KPMG that
they applied item level deduplication on the first occasion that the search was
run. However, multiple versions of Appendix 3 in addition to the one that was
produced (and including versions that had Appendix 6 within the same family)
were still promoted to review because not all versions of Appendix 3 were MD5#
duplicates. I understand that an "MD5#" value is a sequence of letters and
numbers that make up a document's digital fingerprint and/or text, and can be
used to exclude duplicates irrespective of their families.
99. The search was later re-run in April 2022 and KPMG have informed us that item
level deduplication was not applied on that occasion. It therefore appears that
item level deduplication did not play a material role in the rest of the suite of
documents not being identified at this stage, and the approach to reviewing
family documents was the operative reason these documents were not
reviewed by my firm in relation to Rule 9 Request (11) (INQ00002007).
100. This approach of not reviewing full families for relevance was the preponderant
one adopted for subsequent Rule 9 Requests that required the production of
documents. For context, all the Rule 9 Requests received by POL after Rule 9
Page 38 of 63
WITN09950100
Request (14) (INQ00002008) had deadlines of one month or less (save for Rule
9 Request (31)). Many had deadlines of a few days.
101. The approach was not intended or thought to detract from the fulfilment of POL's
disclosure obligations. To the contrary, it meant that the search terms could be
calibrated in a way that ultimately enabled more keyword responsive
documents to be reviewed, which was thought conducive to the fulfilment of
POL's disclosure obligations. In other words, focusing on keyword responsive
documents would be more likely to result in the identification of relevant
documents than reviewing non-keyword responsive family documents.
102. It was considered not to be feasible in the time allowed by the Inquiry for
responding to the Rule 9 Requests to review family documents, at least without
recalibrating the search terms in order to generate fewer keyword responsive
documents. As I have noted above, it is not unusual for a single document to
have dozens of family documents, and some can have hundreds of family
documents.
103. To place this discussion in the particular context of Rule 9 Request (11)
(INQ00002007) Questions 15 and 46(a), when my firm ran the search terms I
describe in paragraph 71 across the databases, there were around 10,000
responsive ‘hit! documents, which expanded to approximately 75,000
documents when family members were included. The number of ‘hits' that were
reviewed by my firm, after KPMG applied item level deduplication, was around
5,000. Looking at the data on Relativity, there were around 35,000 documents
in that set when families are included, i.e. if all families were reviewed, the size
of the review would have increased by a multiple of seven. This was the context
Page 39 of 63
WITN09950100
in which the decision was taken to review the ‘hit’ documents only and not the
families, and in circumstances where there was only a short period of time in
which to complete the review and to instruct KPMG to produce the relevant
documents identified.
104. The volume of family documents may mean that reviewing all of them in the
time allowed — as in the case of Rule 9 Request (11) (INQ00002007) Questions
14 and 46(a) — would simply be impossible, particularly in circumstances where
my team were often working on multiple requests at the same time. For
example, in parallel with working on Rule 9 Request (11), we were also
undertaking significant work on Rule 9 Request (10) (INQ00002002) which
contained 29 Questions, some of which requested documents or information on
a number of sub-issues. In addition, until this year, POL had a preference not
to seek extensions of time from the Inquiry and, where there was no alternative,
wished to keep them as short as possible. To review family documents would
also have given rise to significant additional cost, which at that time POL
generally sought to avoid.
105. I recognise that reviewing all the family documents was not the only possibility,
and that one available option would have been to review only the families of
documents marked as relevant. We did take this approach sometimes, as I
describe above in relation to Rule 9 Request (10) (INQ00002002). Even this
approach, however, could materially increase the number of documents to
review and extend the timeframe in which relevant documents might be found.
Focussing again on the specific example of the search for Rule 9 Request (11)
(INQ00002007) Questions 15 and 46(a), I estimate from the data on Relativity
Page 40 of 63
WITN09950100
that an additional 1,200 documents would have had to be reviewed for those
Questions alone had that methodology been adopted; and of course it is only
hindsight that suggests those particular Questions might have been treated
differently — adopting that methodology across all the Questions in Rule 9(11),
or indeed as a blanket policy across all Rule 9 Requests, would have increased
the review work by an order of magnitude.
The use of deduplication
106. Deduplication is a conventional and common step in disclosure: to state the
obvious, there is no utility in my team reviewing the very same documents
multiple times in response to a particular request from the Inquiry and there is
no evidentiary value in providing to the Inquiry the very same document multiple
times in response to the same request.
107. I understand that there are different forms of deduplication. "Family level" (or
“parent level") deduplication analyses duplication as between complete families
of documents. It can be used to exclude duplicate families, ensuring that only
one copy of a family survives and is identified for review (and, if applicable,
produced) based on MD5# value. "Item level" deduplication analyses
duplication between copies of one singular document — it can be based on a
document's MD5# value and/or extracted text.
108. If used to exclude documents from review, item level deduplication has the
effect that the reviewer will only see one version of a search criteria-responsive
document, with other duplicates being excluded from the review pool,
irrespective of whether they appear in a different family context. This 'chosen' /
‘primary’ version of the search criteria-responsive document will be identified by
Page 41 of 63
WITN09950100
KPMG to a review team for review along with its family (if applicable). All other
duplicates will not be identified for review, including their families which may be
different to the family of the 'chosen' version.
109. So if a particular attachment (say, a pdf note of a meeting) is included in the
data set 27 times; and each time it occurs, it is an attachment to a unique,
different email (say, between different senders and recipients, sent and
received on different dates over a 5 year period) and none of the other emails
or their attachments respond independently to the search terms — the effect of
item level deduplication will be that 26 of those emails, and any other
attachments they may have, will be withheld from the review teams. It therefore
excludes not only duplicate documents but also non-duplicative documents.
From a practical point of view, this has a number of consequences. For
example, reviewers will only be able to consider the context set out in the one
parent email (and any other family members) related to the "chosen" duplicate
in their "review batch". We understand from KPMG that the instance of the
attachment ultimately promoted for review is selected based only on the fact
that it was the first iteration of that document ingested into the database.
110. Where the review team only wishes to consider one iteration of a particular
document in their review batch in the first instance, it is conceivable that they
may wish to change their approach after the review commences, or has
completed, so that they can review multiple iterations (including duplicates) of
particular documents with their family members. It is therefore important that,
to the extent that item level deduplication is applied, it is done in a way that the
Page 42 of 63
111.
112.
113.
WITN09950100
review team can, if they wish, conveniently review those iterations and their
families at a later point.
It became clear in June 2023 that KPMG had regularly been applying a form of
item level deduplication in returning documents to my team for review where
my team had requested that documents be deduplicated. My team's requests
for 'deduplication' appear often to have been understood by KPMG to mean
‘item level deduplication': that is, an instruction to exclude all duplicates of the
keyword responsive document and their families, whether duplicative or not.
On many occasions members of my team did ask for duplicates of keyword
responsive documents to be deduplicated, in the sense of not being selected
for review for the purpose of the particular review that they were then
undertaking. However, I am not aware of any instruction from my team to
exclude non-duplicative families in the manner that occurred, that is, in a way
that meant that they were not only excluded from review at that particular time,
but also that they could only be retrieved through a complex, time-consuming
and labour-intensive process. I am also not aware of KPMG advising my team
or POL of these implications, which only became apparent over the period from
June to August 2023, nor of clear advice from KPMG (until late June 2023) as
to the implications of item level deduplication. In my view, this advice should
have been given at the outset and before item level deduplication was applied.
As can be understood from my explanation above, there is an obvious
connection between the family review issue and the deduplication issue. I return
to this below in addressing the Inquiry's specific questions.
Page 43 of 63
WITN09950100
114. The relevant effect for present purposes of deduplication has been the delay
caused to the Inquiry and the significant resource required to put it right. In
particular, the way in which deduplication was applied by KPMG and also
prolonged uncertainty as to which Rule 9 Requests, Questions and searches
were impacted, caused significant delay to the commencement of the
remediation process and extended the time that the process has taken.
115. Since this issue came to light my team and I have worked with KPMG, together
with POL, to ensure better understanding and communication, including in
relation to the remediation activities described in this statement.
PART 5 — THE INQUIRY'S QUESTIONS AND MY ANSWERS
116. I have been, in summary, asked by the Inquiry to explain the events which led
to the three disclosure issues, where responsibility for them lies, what steps are
being taken to remediate them, and the systems and processes in place to
avoid similar issues occurring.
Question 1: Decisions and communications
117. Ihope that I have addressed this aspect adequately above.
Question 2 (a): discovery of issues
118. The approach by my firm to the review of family documents, and to the use of
search terms, were known about within the team when the review was taking
place: indeed they were both intentional exercises of judgement, for which I
have explained our rationale in the preceding sections of my statement.
Page 44 of 63
WITN09950100
119. I began to consider these issues in June 2023, when we were investigating the
reasons why Appendix 6 (POL001 15674) had not been disclosed to the Inquiry.
Appendix 6 (POL00115674) was produced on 19 May 2023 in response to a
request made under the Freedom of Information Act 2000, and disclosed to the
Inquiry on 30 May 2023.
120. As to the discovery of the deduplication issue, on 15 June 2023, in the process
of assisting POL with the preparation of the Second Witness Statement of Ben
Foat, my team asked KPMG about its approach to applying deduplication. On
16 June 2023, KPMG informed us that, in short, item level deduplication had
been applied broadly on the matter for some time. Following this exchange, my
team liaised urgently with KPMG to seek to understand when item level
deduplication had been applied to reviews on this matter. KPMG are providing
the information to us on an iterative basis, starting on 13 July 2023. On 4 August
2023, KPMG updated my team regarding the searches affected by item level
deduplication, with additional information being provided on 11 August 2023.
Question 2(b): responsibility
121. To state the obvious, my firm and I have an interest in this question.
Nonetheless I shall seek to be as objective as I can and to confine myself as
closely as I can to the facts. I would comment as follows:
a. These issues have come to be considered against the background that
a racist and offensive document, Appendix 6 (POL00115674), was
disclosed late to the Inquiry. Appendix 6 was not responsive to search
terms, and so it had not been reviewed. There had been no positive
decision taken that it was not disclosable.
Page 45 of 63
WITN09950100
Search terms
b. Search terms devised by P&P as part of the PCDE to identify criminal
policies and procedures were relied on as part of the Policy Review. My
firm was not involved in devising these search terms. The work
undertaken by P&P as part of the Policy Review was utilised when
responding to Rule 9 Requests (11) and (14).
c. As regards the search terms that were devised by my team in relation to
Rule 9 Request (11) (INQ00002007), I have explained that those search
terms targeted the type of documents sought by the Inquiry by Questions
15 and 46(a) and I hope that it can now be seen why the approach
adopted was considered appropriate.
d. As I have explained at paragraph 60, the choice of particular search
terms in responding to each Rule 9 Request is a balancing exercise. In
addition, as I have said at paragraph 72 above, Appendix 6 is an
extraordinary document, and there is a real question as to whether it
would have responded to any reasonably devised search terms.
Reviewing family documents
e. Versions of Appendix 3 did respond positively to the search terms
devised respectively by P&P and my team, and versions of that
document were disclosed to the Inquiry in response to Rule 9 Requests
(11) and (14).
fi In relation to the versions of Appendix 3 reviewed by my team when
responding to Rule 9 Request (11) (INQ00002007), a decision had been
Page 46 of 63
WITN09950100
made not to review family members of documents responsive to search
terms to which Appendix 3 responded. Therefore, only Appendix 3 was
reviewed, not any of the other documents in the suite. We know with
hindsight that there are multiple versions of Appendix 3 that contain
Appendix 6 within the same family.
I would characterise the decision not to review family documents as an
operational one, or a ‘judgement call’ made by my team; and with no
intention on the part of POL or my team to withhold Appendix 6. Giving
disclosure of only relevant documents from such a vast repository is not
a perfect science, and searches that with hindsight can be seen would
have identified an important document are often not apparent at the time
that the searches are devised. I do not believe that it can be said,
objectively, that the approach was not a reasonable one in the context.
However, I recognise that Appendix 6 should have been disclosed earlier
to the Inquiry; and that our decision concerning family review — for which
I take full responsibility - caused the loss of an opportunity to identify
Appendix 6 in the context of Rule 9 Request 11. I wish to express my
regret to the Inquiry and to the other Core Participants for the role this
has played in causing a delay to the Inquiry's work. I would ask that this
is considered in the context as I have explained it above, and to note that
I believe that there are other factors involved in causing the delay, as
also explained above. I would also highlight the prompt and extensive
remediation exercise which I explain below, which was swiftly completed
Page 47 of 63
WITN09950100
in respect of families, and to which very considerable resource continues
to be committed.
The use of deduplication
ie As explained at paragraphs 93 and 98, on the basis of the information
now to hand, we would not invite the conclusion that KPMG's approach
to deduplication played a causative role specifically in the late disclosure
of Appendix 6 (POL00115674).
J: As to the impact of deduplication more generally, I have noted at
paragraphs 111 and 112 above that there appears to have been a
misalignment between my team and KPMG as to the approach to
deduplication being taken and its implications.
k. The approach taken to deduplication has undoubtedly been a cause of
delay to the Phase 4 hearings. I also recognise that the remediation task
of looking at the family documents produced against all Rule 9 Requests
has resulted in documents that are relevant or potentially of interest
being produced to the Inquiry at a late stage.
Questions 2(c) and 3: Remediation and future systems and processes
"Policy Review" searches remediation
122. Inrelation to search terms and search criteria remediation, an exercise is being
undertaken jointly by P&P and my firm to consider the suitability of the PCDE
search terms that were relied upon as part of the Policy Review exercise (see
paragraph 68 above).
Page 48 of 63
WITN09950100
123. This was initiated as a result of specific gaps being identified in the PCDE
search terms relied on as part of the Policy Review, such as not using the term
"guid*" as a search term when Question 18 sought, in terms, "guidelines". The
gaps in the search terms used were identified in the process of responding to
Rule 9 Request (44) and preparing the Second Witness Statement of Ben Foat.
This has resulted in a number of additional search terms being run across the
CCRC, POHIT and Group Litigation Order ("GLO") databases as well as
custodian mailboxes. I understand that it is likely to entail the production of a
material number of further documents, although at this stage POL cannot form
a view on the forensic value of those documents to the Inquiry.
124. The work is not limited just to re-considering the search terms used. A specific
search has been run against the mailboxes of 21 individuals who have been
identified as potentially having formed part of a Working Group set up to
transition RMG policies into POL policies following separation. In addition, P&P.
and my firm have sought input from the POL Security Team to assist in
capturing relevant documents. The POL Security Team provided suggestions
of further lines of enquiry, and P&P have ensured that documents highlighted
by the POL Security Team have been captured by the searches run, and that
further search terms have otherwise been incorporated as necessary.
125. As I explain in the Annex, POL has completed the review of more than 50,000
documents in the CCRC, POHIT Inquiry and GLO databases that were
identified by the updated searches. This resulted in the production to the Inquiry
of 172 documents on 14 August 2023 (WITN09950110) and 2,740 documents
on 15 August 2023 (WITN09950111). Of the 172 documents that were
Page 49 of 63
WITN09950100
produced on 14 August 2023 from the POHIT Inquiry database, 61 were
duplicates or near duplicates of documents previously produced by POL to the
Inquiry. Of the 2,740 documents that were produced on 15 August 2023, at
least 602 documents were duplicates or near duplicates of previously produced
documents to the Inquiry (it is possible that there are more exact and near
duplicates which cannot be identified due to limitations in the CCRC database
(which KPMG are seeking to resolve)). The review is continuing and I currently
anticipate that a further production will be made on 25 August 2023.
Families remediation
126. The family documents issue has been swiftly remediated. In summary:
a. On 23 June 2023, my firm commenced an urgent review of all unique
(that is, non-duplicative) documents which had not previously been
reviewed or produced to the Inquiry and which are family members of
relevant documents that have been produced to the Inquiry.
b. In that process, approximately 27,300 unique unreviewed family
documents from the POHIT Inquiry database were reviewed by my firm,
by deploying significant resource, intensively and urgently. In parallel,
P&P reviewed 3,230 documents from the CCRC database.
c. As a result, 1,516 family documents from the POHIT Inquiry database
and 360 family documents from the CCRC database were identified as
relevant or potentially of interest to the Inquiry. The documents from the
POHIT Inquiry database were produced to the Inquiry on 3 July 2023,
and the documents from the CCRC database were produced on 6, 11,
Page 50 of 63
WITN09950100
and 12 July 2023 (following technical issues that arose when KPMG
attempted to upload the CCRC documents to the Inquiry's platform on 3
July 2023).
Of the documents disclosed from the POHIT Inquiry database, 904 were
near duplicates of documents already produced to the Inquiry, and fewer
than 730 related to Rule 9 Requests which POL understands were relate
to Phase 4 of the Inquiry (many of which were similar to and/or near
duplicates of previously produced documents). By ‘near duplicate’, I
mean a document that is 95-99% similar to another document based on
a ‘textual near duplicate’ analysis performed by KPMG.
Of the documents disclosed from the CCRC database, at least 8 were
duplicates and at least 42 were near duplicates of documents already
produced to the Inquiry. As mentioned above, it is possible that there are
more exact and near duplicates which cannot be identified due to
limitations in the CCRC database (which KPMG are seeking to resolve).
Thus, in a context where over 127,000 documents have been disclosed
to the Inquiry overall, at most 922 technically new documents were
disclosed as part of the families remediation exercise described in this
section, which is approximately 3% of the family documents reviewed. It
is possible that the true relevancy rate is materially lower because of the
limitations on identifying exact and near duplicates in the CCRC
database mentioned above and also because POL has erred on the side
of inclusion in giving production.
Page 51 of 63
127.
WITN09950100
I would observe at this point that the number of family documents reviewed as
part of this exercise — 27,300 — does not accurately indicate the number of
family documents that would have been reviewed if POL had adopted a different
approach to family review in relation to each of the affected Rule 9 Requests
during the course of the disclosure exercise. It would have been necessary to
review more — almost certainly far more — than 27,300 family documents:
a. Since the family review was conducted once and retrospectively (that is,
across all productions over time) we were able to focus on unique
documents rather than the full suite of duplicative material (that is,
duplicates of documents which had already been reviewed and produced
to the Inquiry and internal duplicates within the families).
b. Reviewers could review each identified document once against all past
Rule 9 Requests rather than potentially reviewing a document family as
against multiple Rule 9 Requests over time. To take a hypothetical
example, the family documents to a relevant ‘hit! may have been
reviewed and determined in March 2022 to have been irrelevant to
Rule 9 Request (11) (INQ00002007) and not likely to be of interest to the
Inquiry; but if another document in the family was a relevant ‘hit’ to a later
Rule 9 Request, the family documents would have been re-reviewed for
relevance against all requests that had been issued as at that date, and
a different decision as to relevance might have been taken. In this way,
individual family documents might have ended up being reviewed
multiple times.
Page 52 of 63
WITN09950100
c. It is not uncommon for first level reviewers to conservatively tag
documents as ‘relevant’ which are found not to be relevant at second
level review and/or during QC. When those ‘hits’ would have been
reviewed at second level review and during QC, the family documents
would be brought with them for review; and those families would have
been reviewed for relevance, even if the ‘hit' itself (and/or the family
documents) was subsequently downgraded to irrelevant. During the
family review, in contrast, we were focussing only on documents
confirmed to be relevant and produced, so did not encounter this
difficulty.
128. POL has confirmed to the Inquiry (in my firm's letter of 1 July 2023
(WITN09950107)) that it will review family documents wherever this is both
technically possible and there is a reasonable prospect that the family
documents may contain relevant information and/or documents that are likely
to be of interest to the Inquiry. The Annex to this statement describes in detail
the remediation steps that my firm has been undertaking on behalf of POL in
respect of the deduplication and search terms issues.
Deduplication remediation
129. As regard deduplication, given the uncertainties as to the scope of impact and
complexities that arise from variables such as the number of affected
duplicates, the sizes of affected document families and relevance rates, the
process has been very challenging. My team has been working with POL and
its other advisers intensively and around the clock to try to mitigate this issue.
A threshold issue which has delayed the commencement of remediation, and
Page 53 of 63
WITN09950100
which extends the timeframe for resolution significantly is the uncertainty as to
which of the Requests, Questions and searches were impacted by item level
deduplication. KPMG have had to undertake complex, time-consuming and
labour-intensive work in order to provide this information, which they have
continued to provide until very recently.
130. Once KPMG inform my team which Requests, Questions and searches are
impacted, additional time is required to re-run searches (sometimes across
multiple databases, in which case it is also necessary to migrate documents),
exclude materials which have already been reviewed and in some instances
produced to the Inquiry, and batch the remaining documents for review. My
understanding based on discussions between members of my team and KPMG
is that these steps in the deduplication remediation are also time-consuming
and manual, and involve machine processing time. By way of illustration, we
understand that one step takes (on average) a full day to run.
131. All ofthese steps are required before my team can begin reviewing documents,
which (depending on numbers) can itself be a time-consuming process.
However, the resource committed by my firm and POL has been very
significant. As I mentioned at paragraph 15, there are approximately 150 people
working on the matter at present, including ALT staff, lawyers, trainees,
paralegals and legal operations staff, and approximately 280 people have
worked on this matter across its lifetime. Since early July 2023, approximately
40 people have been brought on to the matter to assist with disclosure work.
132. We have sought to prioritise the review of documents in accordance with the
order in which topics within Phase 4 are due to be heard when hearings resume
Page 54 of 63
WITN09950100
in September 2023 and to produce any relevant materials to the Inquiry on a
rolling basis expeditiously within the constraints of this exercise.
133. As at 18 August 2023, as a result of this work, 76 documents had been
produced to the Inquiry as being responsive to a Rule 9 Request and/or likely
to be of interest to the Inquiry. These 76 documents were produced following a
review of approximately 35,000 documents.
134. Prior to the development of the methodology for remediating the deduplication
issue set out above, and as explained in my firm's letter of 5 July 2023
(WITN09950108), POL sought to mitigate the implications of deduplication in
respect of the evidence of Gareth Jenkins. My team therefore undertook a
review of 1,116 documents and produced 190 documents from this review in
respect of Gareth Jenkins on 13 July 2023. This pool of documents was derived
by focusing on family documents to duplicates of the 13 documents produced
by POL which existed at that time on the Inquiry's Core Participant Platform on
Relativity and which were referred to by Gareth Jenkins in his Second Witness
Statement (being materials that item level deduplication might have excluded).
135. Thus, in a context where over 127,000 documents have been disclosed to the
Inquiry, the output of POL's deduplication remediation exercise described at
paragraphs 130 and 131 above has, to date, increased the number of produced
documents by less than 0.06%. The relevancy rate of the documents reviewed
in the exercise was approximately 0.22%.
136. My firm is also investigating (with KPMG's assistance) the extent to which item
level deduplication has impacted on searches and document reviews carried
out in connection with Rule 9 Requests relating to Phases 2 and 3. The
Page 55 of 63
WITN09950100
information we have received from KPMG indicates that POL's disclosure in
respect of Phase 2 requests (and the aspects of Rule 9 Requests 10 and 38
that relate to Phase 2) have not been impacted by item level deduplication. Our
work in respect of Phase 3 requests is ongoing.
Appendix 6
137. In relation to the suite of documents containing Appendix 6 (POL00115674),
POL has carried out targeted searches to identify all versions of Appendix 6 on
its eDiscovery platform. On 2 August 2023, POL produced to the Inquiry all
copies of Appendix 6 plus family documents that were on POL's Relativity
databases and three other copies that POL had independently identified,
together with four other documents which included the same language used in
the identification codes in Appendix 6. POL is continuing to search for further
versions of this document.
Assurance
138. As part of our general conduct of this matter, and as a matter of good practice,
my team considers on an ongoing basis whether additional work ought to be
conducted following POL's initial production of documents in response to a
particular Rule 9 Request. I would describe this work as ‘business as usual’. By
this I mean, they are matters that POL and my team undertake in the ordinary
course, quite apart from the specific issues recently identified in the July
Directions.
139. Where our understanding of the subject matter of previous Rule 9 Requests
has developed since the original response, we have considered whether further
Page 56 of 63
WITN09950100
searches, harvests and reviews should now sensibly be done. As part of this,
we have also identified some further repositories that we consider are
appropriate to search. Documents identified as relevant to previous Rule 9
Requests have been produced to the Inquiry as a result of this work. This is as
set out in the various interim disclosure statements that POL has provided to
the Inquiry and my firm's correspondence at the time of these further
productions.
140. As part of this ongoing work, POL is presently undertaking an assurance
exercise. This is being conducted by my firm, with input from Burges Salmon
LLP and Fieldfisher LLP. It is separate to the Policy Review searches
remediation work that is now being undertaken by P&P with my team (which is
dealt with at paragraphs 122 to 125 above). It was not prompted by the
identification of any specific issues with the search terms.
141. The rationale for the assurance exercise was to take the opportunity to reflect
on the approach, and to consider, with the benefit of the additional knowledge
we have now, whether there was anything further which should be done in order
to assist the Inquiry. Such work is also done with the benefit of input from
members of the counsel team representing POL in the Inquiry, given their
detailed understanding of the underlying issues, including from their work on
review of documents and witness statements on the Inquiry's "Core Participant"
platform.
142. As I describe in detail in the Annex to my statement, the assurance exercise
involves undertaking an analysis of the approach in responding to certain Rule
9 Requests. It involves analysing, on a question-by-question basis, whether in
Page 57 of 63
WITN09950100
the light of the information that is now available to us there may have been any
gaps in the disclosure previously provided or whether any supplementary
searches might now be appropriate. We have focussed principally on the Rule
9 Requests relating to Phase 4 given their obvious relevance to the current
hearings.
143. So far, the assurance exercise undertaken by my team in respect of the civil
recoveries element of Phase 4 has not identified any gaps or issues in relation
to the exercise of responding to the Rule 9 Requests as this was done at the
time. It is relevant to note here that our approach to Rule 9 Requests was multi-
faceted, and there were several workstreams separate from the searches that
were run on Relativity when responding to the Requests initially. For example,
in responding to Rule 9 Request (14) (INQ00002008) Questions 1-17 (which
related to civil recoveries), with POL's assistance my team conducted fact-
finding sessions with 14 current POL employees who had been identified as
having potentially relevant information on the operation of these areas within
POL. In these sessions, we specifically asked questions about and discussed
documents that might be relevant to the Request.
144. Looking at the recent output from this assurance work since 6 July 2023, 124
documents have been produced as being responsive to a Rule 9 Request
and/or likely to be of interest to the Inquiry. 91 of these documents were
produced following a review of a pool of documents resulting from searches
conducted over repositories totalling approximately 10,000. The remaining 33
documents were produced on 14 August 2023 following a review of specific
subfolders in a particular SharePoint that had been harvested.
Page 58 of 63
WITN09950100
145. Again, where over 127,000 documents have been disclosed to the Inquiry, the
output of POL's intensive civil recoveries assurance exercise has, to date,
increased the number of produced documents by less than 0.098%. The
relevancy rate of the documents reviewed as a result of applying search
parameters in the exercise was approximately 0.89%.
146. Together with my firm, P&P have also been undertaking additional work in
relation to certain of the Rule 9 Requests with which they assisted:
a. This has included additional work in relation to a search that was undertaken
specifically for the purpose of identifying documents relevant to Rule 9 Request
(14) (INQ00002008) Questions 22, 26, 30, 34 and 38 (the "Training Search"),
including running the original search terms directly over certain custodian
mailboxes, and running additional search terms; and
b. My firm's letter to the Inquiry dated 6 February 2023 (WITN09950109)
outlined the Data Scoping Analysis that was being undertaken to
consider the extent to which: (i) the PCDE search terms covered the
issues raised in CLI 49 (Knowledge); and (ii) there were any further
repositories or custodian data that could be collected (in addition to that
already collected for the purposes of the PCDE). The Data Scoping
Analysis was undertaken having regard to the fact that POL has sought
to utilise the work undertaken by P&P in relation to the PCDE to the
fullest extent possible for the purposes of responding to Rule 9 Requests
related to private prosecutions.
Page 59 of 63
WITN09950100
14 August deadline
147. POL and my team are acutely conscious of the importance and significance of
the July Directions. We have devoted considerable resources and senior
management focus in an effort to produce documents to the Inquiry by the
deadline set by the Inquiry of 14 August 2023. As I explain in the Annex and
elsewhere in this statement, addressing the matters raised in the directions has
involved conducting several complex workstreams and disclosure exercises
simultaneously. This work been conducted alongside additional assurance
work so as to ensure that the Inquiry receives all documents relevant to Phase
4 witnesses as promptly as possible.
148. Despite these intensive efforts, and due to several factors, it was not possible
to complete all workstreams before the deadline of 14 August 2023. These
factors included the need to work with KPMG to understand the scope of impact
of item level deduplication (which itself is a manual and time-consuming
process) and the implementation of similarly time-consuming, labour-intensive
and technically complicated remedial workflows in order to avoid re-reviewing
the same documents or producing significant quantities of duplicative
documents to the Inquiry; and substantial further document review.
149. POL was able to produce all documents relevant to the witnesses who gave
evidence in the week commencing 24 July 2023, and POL does not anticipate
making any further disclosure in response to paragraphs 1 and 3 of the July
Directions insofar as they relate to the Marine Drive Post Office Case Study,
which is scheduled to be heard by the Inquiry in the week commencing 19
September 2023.
Page 60 of 63
WITN09950100
150. Further, in relation to the workstreams relating to Phase 4 which were not able
to be completed by 14 August 2023, considerable progress has been made. By
way of illustration, as at 16 August 2023, my firm had reviewed at least 183,225
documents. I say ‘at least' because quantifying this involves an analysis across
multiple databases, coding layouts and workflows, which is not straightforward.
A significant proportion of the reviewed documents are 'near complete’, in the
sense that they have either been confirmed not to be relevant, or they have
been identified as potentially relevant and are ongoing further review and QC.
POL will continue to progress this work urgently and produce documents to the
Inquiry as promptly as possible. As at the date of this statement, POL intends
to complete substantive workstreams relevant to Phase 4 in time to produce
responsive documents on or around 25 August 2023 and 31 August 2023, and
will continue to complete production of documents that relate to Phase 4
witnesses in good time for each witness.
Systems and processes
151. As to the systems and processes in place to ensure that POL's disclosure will
be provided and to avoid future issues of a similar nature, I must begin my
explanation by noting that, as I explained at paragraphs 18 to 20 above, the
role of RLR will soon transfer to Chris Jackson of Burges Salmon LLP. Whilst
I anticipate that my firm will continue to have a role in assisting POL with Phase
4 work until its conclusion, Burges Salmon LLP will have conduct of the
remainder of the Inquiry. This is therefore an issue on which the Inquiry will wish
primarily to hear from others.
Page 61 of 63
WITN09950100
152. In my experience, large-scale disclosure exercises are evolving processes, that
are developed and refined over time as matters progress and additional
information becomes available. Throughout the course of our work assisting
POL in the preparation of its responses to Rule 9 Requests and Section 21
Notices, my team has sought to identify and incorporate learning points (factual,
procedural and process-based) into our approach, on an ongoing and
incremental basis. We will continue to do this in the disclosure work we are
currently carrying out on POL's behalf and in respect of any future requests
from the Inquiry to the extent we are involved.
153. Insofar as we are involved in preparing POL's responses to requests from the
Inquiry going forward, we will:
a. Continue to carefully consider, develop, revise and refine our use of
search terms and other parameters to fit the unique demands of each
Rule 9 Request.
b. Continue to ensure that the review workflow undertaken in response to
Rule 9 Requests or Section 21 Notices is tailored to each individual
request, based on available information and in consultation with POL.
6. Review family documents wherever this is both technically possible and
there is a reasonable prospect that the family documents may contain
relevant information and/or be of interest to the Inquiry.
d. Insofar as it is within our control, ensure that our instructions to KPMG to
perform family level deduplication are clear, understood and consistently
and accurately executed by KPMG.
Page 62 of 63
WITN09950100
Statement of truth
I believe the content of this statement to be true.
Dated: 23 August 2023
Page 63 of 63