Practitioners involved in large-scale discovery should pay attention to version 2.0 of the Good Practice Discovery Guide, published by the Commercial Litigation Association of Ireland (CLAI).
Regard should also be had to the recent CLAI discussion document entitled Possible Reforms to the Rules of the Superior Courts to Reduce the Time and Cost of Discovery.
These documents will bring any practitioner up to speed with the various issues facing the courts in the context of discovery projects involving large amounts of electronically stored information (ESI).
Back to basics
To understand the practical issues addressed below, it is necessary to understand the fundamentals of electronic document review (EDR) and technology assisted review (TAR). For this, practitioners should refer to the April 2017 issue of the Gazette (p44).
The increased use of electronic devices has led to a corresponding proliferation of ESI. Today, even relatively standard cases might necessitate a document review encompassing electronic documents numbering in the tens or hundreds of thousands.
This process will involve a team of reviewers working on an electronic review platform. This is a program that presents the reviewer with documents that the reviewer codes by clicking the appropriate boxes in respect of applicable categories, privilege, redaction, etc, and is called manual review.
Manual review
Before manual review commences, the party making discovery must gather the data from the relevant custodians and sources. In large cases, it is increasingly common that this process can return over a million documents.
This body of documents is referred to as the ‘universe’ of documents. This raw ESI must then be processed. Documents may then be filtered by date range and keywords (or search terms), to identify potentially relevant documents.
The quantity of documents returned through this process will depend on factors such as the temporal scope of the discovery order, and the precision of the keywords applied.
After initial processing, it may be necessary to further narrow the universe of documents through TAR – a method aimed at reducing the universe of documents to a more manageable size.
In broad terms, TAR can be described as a system whereby algorithms are used so as to establish the likelihood of the relevance of a document, based on human interaction in the form of manual review of sample sets by a senior lawyer.
Eventually the system’s predictions will be sufficiently consistent with the manual decisions, so that it will be able to make reliable and accurate predictions for the remaining documents.
Keeping it relevant
TAR involves a number of stages, and although there has been debate as to the precise stage at which it should be carried out, the process will involve the preliminary identification of potentially relevant documents by applying keywords to the electronic documents held by a party.
Even where TAR isn’t applied, keywords may still be used in the processing stage to remove irrelevant data.
In either case, the categories are already set. The specificity of keywords should not be perceived as an attempt by the discovering party to avoid their obligations.
The application of appropriately constructed keywords is a device enabling parties to separate the wheat from the chaff.
While a balance must be struck, it goes without saying that overly broad keywords are not helpful in this process.
Practitioners who are less familiar with the process will often push for the inclusion of inappropriate terms, based on the fear that what they perceive to be overly precise terms will lead to relevant documents being missed. This issue was addressed in Thema and, in Britain, in Goodale.
Section 10.7.2 of the Good Practice Discovery Guide provides guidance on the process of developing filtering criteria and search terms, which will require collaborative input from legal professionals and IT specialists.
When countering with further proposed search terms, practitioners should ensure they are adding value. Although methods such as Boolean operators may assist, non-specific keywords like ‘bank’, ‘central bank’, ‘guarantee’, ‘contract’, or ‘interest’ will not assist either side.
While a party may feel that they have achieved a victory through the inclusion of additional broad keywords, such a victory may later prove pyrrhic, particularly when costs are being addressed.
Who’s yo’ daddy?
The manner in which families of documents are treated is a central issue in any review. In both hard and soft-copy review, it is vital that the context and familial relationship of documents is maintained.
‘Family’ in this context refers to the hierarchical relationship between groups of documents. The best example of this in the context of ESI is the relationship between an email – the ‘parent’ document – and its attachments, ‘the children’.
Irrelevant documents should not be discovered, but attachments should never be ‘orphaned’. The Good Practice Discover Guide covers this issue in detail at 15.1 (see p57); and F.2.3 (pp78-80).
In short, where an email is irrelevant, save for the fact that affixed to it is a relevant attachment, the parent should be discovered along with the relevant attachment. It is not necessary to discover any other irrelevant sibling attachments.
One consequence of this approach is that a great many documents, relevant only by virtue of their attachment, will be marked as relevant, thereby necessitating review by the receiving party.
A solution to this problem is the inclusion of ‘parent for context’ (PFC) as either a distinct option in the category field; as an alternative option in the relevancy ‘yes/no’ field; or both.
I first encountered this approach within the data investigation group of a large firm, and has consistently recommended, requested or applied it ever since. The fact that a large firm utilised this approach should be of comfort to smaller firms. PFC benefits all parties.
At the front end, it means that those conducting the review have a more manageable data set, while at the back end, it enables the receiving party to identify those documents to which primary attention should be paid – all while maintaining compliance with the Good Practice Discovery Guide.
The use of PFC in the manner set out above should be regarded as the ‘gold standard’, and should be incorporated into the next edition of the guide.
Smaller parties should, at the outset, request that the party making discovery utilise this approach. If this cannot be agreed, it should be addressed in case management.
Don’t take the bait
Despite the guide’s clear statement that irrelevant documents should not be produced, some firms adopt a policy of discovering the entire family of all relevant documents.
This could be misinterpreted as some manner of concession, or as being advantageous to the receiving party.
Most large firms have review teams working on a semi-permanent basis, and this approach merely enables such a firm to ensure its review team processes more matters in less time.
By adopting such a policy, once one family member is deemed relevant, it becomes unnecessary to consider each additional familial document for relevancy. Each document will still need to be considered for privilege, redaction, and other such issues but, in most cases, it will simply be a case of applying the same coding to each document.
A practical example of the consequences of such a policy is the email sent by a party whose pro forma email includes unique images or logos in the signature section.
Even after de-duplication, a long email thread will often generate tens or hundreds of iterations, all requiring manual review in their own right. However, despite the relevancy of the email thread itself, many such iterations could have a large number of entirely irrelevant attachments.
In theory, this specific problem ought to be solved at the front end by the service provider. In the author’s experience, however, this does not always happen and, in any event, is only one example of the overall problem.
One hundred relevant iterations of an email thread, half of which attach 15 irrelevant attachments, would generate 750 more documents for production. This equation could repeat itself multiple times throughout a day of review.
For the party making discovery, the end result is increased efficiency within its discovery unit. In contrast, the receiving party, who is potentially at a disadvantage in terms of finances, technology and manpower, receives what is, in effect a ‘data-dump’.
Swamping tactics like this can lead to the obfuscation of highly relevant documents, and can also rapidly consume the finances of the smaller party, perhaps pricing them out of the litigation.
Smaller parties should, at the outset, ensure that documents are not to be discovered merely due to the relevance of their sibling.
Where a document is necessary so as to provide context for its relevant sibling, then that document should be discovered and coded accordingly