1 Introduction

There have been many cases in the US1 , and more recently in Ireland2 and New Zealand3 endorsing the use of 'technology assisted review' in discovery. Australian lawyers and discovery professionals have long anticipated a definitive case in Australia. The first English case on this topic – Pyrrho Investments Ltd v MWB Property Ltd4 – brings us one step closer to the formal acceptance of TAR in Australian jurisdictions. It helpfully sets out ten factors in favour of TAR, which all seem to equally apply in Australia.

2 What is Technology Assisted Review ('TAR')?

The following extract from the Pyrrho judgment describes TAR as:

"...the review of the documents concerned being undertaken by proprietary computer software rather than human beings. The software analyses documents and scores them for relevance to the issues in the case ... a representative sample of the 'included' documents is used to 'train' the software."

This extract focuses on the use of technology in the process. However, human involvement – in the reviewing of documents to train the software – is vital to the success of TAR.

While the judge in Pyrrho uses the terms 'TAR' and 'predictive coding' interchangeably, TAR actually covers a number of different technologies and processes to improve the review and disclosure process; predictive coding is only one. Other examples include concept clustering, email threading and document categorisation analytics.

At KordaMentha we use one or more aspects of TAR in almost all of our review and disclosure engagements – from investigations through to electronic discovery – even if discovery is also run by more traditional review processes. Increasingly, the process includes elements of predictive coding.

3 Ten factors

The ten factors in favour of approving the use of predictive coding, usefully set out in Pyrrho, provide a good summary of the key issues. These factors will likely prove useful in assessing whether or not to approve TAR in Commonwealth law jurisdictions which have no specific case law on this issue (as was the case in England prior to Pyrrho):


Experience from other jurisdictions – predictive coding "can be useful in appropriate cases"

Whilst potentially very powerful and useful, TAR is not a panacea for all discovery issues. However, it can – for appropriate cases – provide significant time and cost savings, as well as allowing for improved consistency.

In our experience one of the main drivers is not the number of documents, but how intimately the client understands its electronic data. For example, what information is stored where? Where a client is not familiar with its data we have found TAR to be beneficial with as few as 20,000 documents.


No evidence that TAR is less accurate – "indeed some evidence to the contrary"

There are a number of peer review studies suggesting inherent deficiencies in the manual review process. (See KordaMentha article5 )

One of the key benefits of TAR is that it can allow the review team to consider documents which traditionally may have been excluded as being 'non-responsive' to keyword searches. The design and implementation of keyword searches have long been difficult aspects of discovery, and a source of much disagreement between parties. TAR can be undertaken either with or without the implementation of keywords.


Greater consistency

A key concern in manual reviews is duplicate or conceptually similar documents being coded inconsistently, a symptom of multiple reviewers or reviewer fatigue.

To negate this, initial review of a relatively small sample set by senior legal professionals can allow for better quality control on coding decisions; those individuals are likely to have a better appreciation of the case issues and strategy. The resulting higher quality coding is then consistently applied against the remaining documents. This consistency gives greater confidence to the legal team and, most importantly, to the individual signing affidavits, about the efficacy of the discovery processes.


Nothing in Court Rules or Practice Directions prohibits the use of such software

This is true of Australia and many other jurisdictions. The obligation on the party making discovery is to have an appropriate process which allows for the collection and review of the necessary documents: how this is achieved is largely a matter for the party making the discovery.

The authoritative US Sedona Principles6 also state that individual parties should evaluate the best processes for producing electronic documents.


Number of documents in Pyrrho – over three million – was described as "huge"

The volume of electronic documents retained by organisations continues to increase exponentially. Studies indicate that, on average, stored electronic data is growing at somewhere between 40%–60% annually. Our experience is that three million documents, whilst not small, is far from the largest document set that organisations are likely to have to contend with.


Cost of 'manual review' would have been "unreasonable"

The increasing volume of electronic documents has led to increasing discovery costs, and to calls for discovery to be limited. There have been some attempts to reduce this cost by using offshore "Legal Process Outsourcing" operations (some of which are internal to the larger international law firms). Whilst these can, under the right circumstances, be a useful cost management measure, if there is no integration with TAR then they are simply making the wrong process cheaper, rather than fundamentally reducing the costs by changing the process.


Cost of predictive coding "far less expensive"

In contrast, TAR represents a fundamental re-engineering of the review process, which can make discovery far faster and less costly. Many litigators are nervous of things which are new, but their clients are increasingly asking: "why are you not suggesting TAR to us?", and some clients are beginning to insist on using TAR.


TAR costs were " proportionate " to the scale of the claim

Because of reduced costs and improved timescales, proportionality in the cost of discovery is far more likely to be achieved when TAR is used.


Plenty of time to revisit

The judge highlighted that "if for any reason the predictive software route turned out to be unsatisfactory" the timetables put in place left slack for traditional review.

Our experience is that TAR can be particularly useful when the opposite is true – the timetable is short and therefore traditional 'linear' or 'manual' review would simply not have been possible (see sidebar on page 6).

Even where there is enough time to undertake a traditional approach, TAR can help a party gain a much faster appreciation of where the key documents are. This, in turn, can increase their knowledge of the relative strengths and weaknesses of both their own and their opponent's case. Francis Bacon said that "knowledge is power". But the extension, which could have been written for discovery, is the quote by the American sociologist Robert Staughton Lynd: "Knowledge is power only if man knows what facts not to bother with."


Parties have agreed to use of the software, and how to use it.

Whilst this may have made this case easier, decisions elsewhere have highlighted that it is up to the party making disclosure to undertake the processes it considers appropriate to the circumstances. Importantly the judgment states: "There were no factors of any weight pointing in the opposite direction."

4 Reduced Risks

Some observers consider that the use of TAR may heighten the risk of inadvertent disclosure of potentially privileged or irrelevant documents. The studies identified in factor 2 above highlight that, if appropriately robust workflow processes are put in place (essential for whichever form of review of discovery is undertaken), then TAR can provide substantially reduced risks of inappropriate disclosure of privileged documents, as well as reducing the risk of relevant documents not being disclosed. Most importantly it provides a degree of consistency, and hence reduced risk, compared with manual review by a team of reviewers.

5 Not a binary decision

Much of the discussion regarding TAR proceeds on the basis that it is an 'either/or' decision – you choose between TAR and a more traditional linear review using keywords and human reviewers. However, in our opinion, this is a false dichotomy.

TAR can – and often should – be used in conjunction with a more traditional review process. Typically TAR will require higher level legal resources at an earlier stage in order to identify issues and code the 'training set'. This can bring great benefits to the process – not least that those directing the case have a better appreciation of the data-set from which documents are to be disclosed. Furthermore, the initial TAR relevance ranking provides a useful tool for prioritisation of manual review; further manual review can continue if necessary over the whole corpus of documents, or over a substantially reduced subset of those documents.

The power of TAR

Having recently taken over the running of an arbitration matter, our client law firm was faced with a two-week deadline to file an amended defence. This required consideration of over 1.5 million documents and emails from key internal stakeholders.

In a traditional approach we helped our client to undertake a series of keyword and date-range searches which reduced the documents requiring review to approximately 75,000; still a daunting prospect to complete in two weeks, whilst also redrafting a defence.

In order to get up to speed the core legal team, including the partners on the engagement, reviewed the initial tranches of documents in the Clearwell Review Platform. Approximately 1,200 documents had been reviewed – a good start, but it became clear that there was simply not enough time to review the remaining 73,800 documents. Overnight we ran the TAR features inbuilt into Clearwell and were able to provide a ranking for every document in the corpus (not just the 75,000, but the 1.5 million). This allowed the law firm to sort the documents by relevance.

This process enabled our client to focus its key resources on the documents most likely to be relevant, and hence piece together the documents required for a robust defence. The lawyers were also pleased (but also a little disconcerted) to find that three of the most relevant documents were not responsive to the carefully selected keyword searches. Had the process followed a more traditional route in reviewing 75,000 documents, even the best human efforts would never have located these critical documents.

6 Conclusion

Although there is no judgment on the issue in Australia, we know that TAR, including predictive coding, is increasingly being used on matters here (both arbitration and litigation). We see the pace of uptake increasing as more litigators – and perhaps more importantly their in-house counsel clients – become aware of the substantial benefits of this approach to discovery.

We sense, particularly in certain jurisdictions, that the judiciary would consider the use of predictive coding very favourably. Justice Vickery of the Victorian Supreme Court discussed this in his paper to the Victorian Bar Continuing Professional Development Seminar7 :

"There is a growing body of evidence which tends to demonstrate that the traditional approach to discovery – that is a 'linear' document-by-document human review, even if aided by the use of keyword searches – is considerably less efficient, and also is likely to be less accurate, than computer-assisted review".

"The judge is needed to overcome what are endemic tendencies in an adversarial system of justice - cost and delay. Technology is a very important tool for the Court in this endeavour, but it is not the only answer."

We appreciate that a solicitor swearing that he or she has disclosed all relevant documents may be nervous in rethinking wording to describe the use of predictive coding and other aspects of TAR. But not adopting this type of technology on large disclosure matters may be detrimental.


1 Most notably Da Silva Moore v. Publicis Groupe, Case No. 1:11-cv-01279 (S.D.N.Y. April 26, 2012) and Global Aerospace v. Landow Aviation No., CL 61040 (Va. Cir. Ct. Apr. 23, 2012)
2 Irish Bank Resolution Corporation Ltd & Ors -v- Quinn & Ors [2015] IEHC 175 http://www.kordamentha.com/docs/defaultsource/for-publications/issue-15-03-predicting-the-future-of-electronic-discovery.pdf June 2015
3 Auckland Waterfront Development Agency Limited v Mobil Oil New Zealand Limited [2015] NZHC 470 (13 March 2015)
4 [2016] EWHC 256 (Cth)
5 http://www.kordamentha.com/docs/for-publications/issue-13-02-death-taxes-computer-assisted-review February 2013
6 The Sedona Principles: Second Edition, Best Practices Recommendations & Principles for Addressing Electronic Document Production
7 Peter Vickery, 'Recent Developments in Discovery in Commercial Litigation' (Paper presented at a Victorian Bar CPD Seminar, 5 February 2015).

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.