rosenblatt view

Archive for January, 2017

Predictive Coding – “TAR” very much

04/01/2017 | Louisa Hartley
Since the decision in Pyrrho Investments Limited and another v MWB Property Limited and others EWHC 256 (Ch) permitting the use of predictive coding in an electronic disclosure exercise there has be

Since the decision in Pyrrho Investments Limited and another v MWB Property Limited and others [2016] EWHC 256 (Ch) permitting the use of predictive coding in an electronic disclosure exercise there has been much commentary on predictive coding, also referred to as Technology Assisted Review [“TAR”], and its use in the disclosure process. Prior to that decision, use of predictive coding technology in the English courts had been limited, mainly due to the fact that there was no authority confirming that its use fulfilled a disclosing party’s obligations under the Civil Procedure Rules [“CPR”].

Disclosure under the CPR

The purpose of disclosure is to make available evidence which either supports or weakens the parties’ respective cases in order to assist the Court to fulfill the overriding objective of dealing with cases justly and at a proportionate cost. Parties are required to undertake a “reasonable search” for documents (CPR r31.7) which meet the test of standard disclosure. This includes electronic documents such as emails and computer files.

What is reasonable is decided by reference to a number of factors (set out in CPR r31.7(2)), specifically:

(a) the number of documents to be searched;

(b) the nature and complexity of the proceedings;

(c) the ease and expense of retrieval of any particular document; and

(d) the significance of any document which is likely to be located during the search.

There are additional factors that determine what is a reasonable search of electronic documents such as consideration as to the availability and location of electronic documents and the cost and likelihood of recovering them (Practice Direction 31B).

What is predictive coding?

Predictive coding is used in the electronic disclosure process in litigation. It is computer software which identifies electronic documents that meet the test of standard disclosure. It aims to streamline the document search and review process and, in doing so, to reduce the costs and the time which would otherwise be spent manually reviewing documents.

Predictive coding is a distinct form of TAR in that it is a continuous learning mechanism, as opposed to merely using technology to search for keywords in documents to consider whether certain document should be disclosed or not. The increasing use of TAR and now predictive coding in place of human manual review of documents in the disclosure exercise is due to the fact that it is just as, if not more, accurate than manual review and in the majority of cases costs less time and money.

How it works

Predictive coding combines human expertise with data analysis to create an algorithm by which relevant and non-relevant documents are identified, categorised and flagged.

In practice, the parties will first agree a set of guidelines to define the data size, criteria for inclusion of documents and a margin of error. Documents are then uploaded onto an electronic review platform and a representative sample is selected to train the software. A manual review of the documents produced by the software is carried out by a solicitor who categorises them as relevant or not-relevant. Predictive coding software then analyses the sample and scores the relevancy of documents to the case based on common concepts and the language included. The remaining documents are then reviewed by the software and categorised as either relevant or not relevant.


  • Avoids manual review of a large number of irrelevant documents therefore saving time and costs.
  • The number of documents in review is not directly proportionate to the costs i.e. doubling the number of documents does not double the cost of reviewing them.
  • Documents which are relevant in the case can be identified quicker than if each were reviewed manually.


  • Predictive coding does not work with non-text-searchable documents and mainly numerical content.
  • Where there is a high proportion of relevant documents the time and cost savings will not be as significant.
  • It is not always 100% accurate.

The decision in Pyrrho

Master Matthews highlighted in his judgment that whilst the CPR and its Practice Directions contemplate the search for electronic documents, neither deal with the question as to how these should be searched. The Master did note however that PD31B does mention the use of “automated methods” or “automated search techniques” in the search for electronic documents.

The number of electronic documents under review in Pyrrho amounted to over 17 million. After de-duplication that number was still 3.1 million. Drawing on the decisions of the US Federal Court case of Moore v Publicis Groupe, 11 Civ 1279 in which there were over 3 million documents to be reviewed and the Irish High Court case of Irish Bank Resolution Corporation Ltd v Quinn [2015] IEHC 175 in which there were over half a million documents after de-duplication, Master Matthews considered that the use of predictive coding was appropriate. Included in the factors he listed in his Judgment as being in favour of the use of TAR were:-

  1. The parties had agreed on the use of predictive coding software, subject to the Court’s approval.
  2. There is no evidence that predictive coding is less accurate than manual review and, in fact, there was evidence that it was more accurate in some cases.
  3. There is nothing in the CPR or its Practice Directions prohibiting its use.
  4. The number of electronic documents to be reviewed in the case was “huge” and the cost of manually searching for these would be “enormous”. The highest cost estimate given to use the technology was £469,049 plus monthly hosting costs of £20,820 which Master Matthews considered as “far less expensive than the full manual alternative”.
  5. The value of the claim was tens of millions of pounds and therefore the costs of using the software were proportionate

Although each case will turn on its own facts, the implication of the judgment in Pyrrho is that there is now judicial authority for parties to use predictive coding in e-disclosure where it is considered appropriate. The more recent case of Brown v BCA Trading Ltd [2016] EWHC 1464 (Ch) affirms the Court’s view that technology is sometimes a better alternative to traditional methods. In that case, the Court ordered the use of predictive coding against the Claimant’s wishes. Over £20,000,000 was being claimed with the cost of predictive coding estimated to be around £132,000 and the cost of traditional review £250,000.

The Court considered the factors identified by Master Matthews in Pyrrho and made an order for the use of TAR. Consequently, even where the parties do not agree to predictive coding, the Court may order its use in disclosure where the arguments in favour far outweigh those against it. It may well transpire that TAR will become the norm, unless there is a compelling reason not to use it.

If you would like any further information, please contact Simon Walton on 0207 955 1455.

This article should not be taken as definitive legal advice on any of the subjects covered. If you do require legal advice, please contact Rosenblatt as above.  

  • contact

Latest news