A decision of the English High Court handed down yesterday has expressly approved the use of predictive coding (also known as technology assisted review) for a large disclosure exercise: Pyrrho Investments Limited & Anr v MWB Property Limited and Others [2016] EWHC 256 (Ch).

Based on the US's experience, this case will likely mark a turning point for the use of predictive coding in England and Wales. It took the US less than three years from the date of its first judicial approval to reach the point that "it is now black letter law that where the producing party wants to utilize [technology assisted review] for document review, courts will permit it.” Rio Tinto Plc v Vale SA, 1:14-cv-3042 (S.D.N.Y. Mar. 2, 2015).

As noted in Master Matthews's decision, predictive coding has been used relatively infrequently in English litigation. However, Herbert Smith Freehills is among a small number of firms that have already employed the technology in large-scale disclosure exercises in England. While judicial endorsement is not a pre-requisite for a party's use of predictive coding software, Master Matthews's decision is welcome in confirming the benefits of the technology in appropriate cases in England and Wales.

Celina McGregor, a senior associate in our London office, considers the decision below.

What is predictive coding?

Predictive coding uses a combination of technology and manual document review. The document review software applies algorithms to human reviewers’ coding decisions to suggest similar documents for review. In doing so, the system prioritises those documents which it believes are most likely to be relevant, which are then reviewed by human reviewers. The purpose of the process is to identify the documents relevant to the case while reducing the time and cost of the review by reducing the number of irrelevant documents. Predictive coding is seen as potentially beneficial in large-scale disclosure exercises where linear review of documents, using only keyword searches to limit the volume, may not be feasible or proportionate.

A copy of our article providing practical guidance on the use of predictive coding, which is cited at paragraph 25 of the court's decision in Pyrrho, is available here


In his decision, Master Matthews considered the use of predictive coding appropriate because:

  • The experience of other jurisdictions, namely the US and Ireland, has been that predictive coding can be useful in appropriate cases.
  • There is no evidence that predictive coding is a less accurate tool for conducting a review than manual review / keyword searches, and some evidence that it may be more accurate and result in more consistent decisions about relevance.
  • There is nothing in the CPR or Practice Directions to prohibit its use and CPR PD 31B specifically refers to the use of "other automated searches" as potentially appropriate. Master Matthews also noted that the Technology and Construction Court judges support an eDisclosure Protocol that specifically refers to the use of predictive coding.
  • In this instance (where the number of electronic documents to be considered for relevance and possible disclosure was over three million) the use of predictive coding would result in a substantial cost-saving and the costs of using the technology were proportionate to the value of the claim.
  • The parties had already agreed on the use of predictive coding for the purpose of the review.


Overall, the case is a welcome decision for those wishing to use predictive coding software. However, it should not be used as a prescriptive or comprehensive guide to predictive coding. In particular:

  • It is important to understand that Master Matthews's description of predictive coding software (at paragraphs 19 to 24) is only one type of predictive coding workflow. There are a number of other options available to parties when considering the use of this technology, each with their own benefits and drawbacks in terms of time, costs, accuracy and ability to deal with changes to the issues in dispute. As the High Court's decision is based on a particular workflow, a number of the comments made by Master Matthews about best practice and the court's expectation of the technology are particular to the type software used, but not generally applicable.
  • In the present case, it appears that little further human review will be conducted of the documents produced after the predictive coding exercise is checked and stabilised. This may not be appropriate in all cases, in particular, where the resulting documents are likely to include privileged material, irrelevant documents, confidential material, or a non-party's personal/sensitive information. Further, while there may be an immediate cost-saving of not reviewing certain documents, disclosing such material may amplify the other parties' costs and risk an adverse cost order.
  • Particularly in this sort of case, parties may wish to consider the use of a clawback agreement to minimise concerns about the disclosure of privileged or irrelevant documents. In the US, clawback agreements set out the terms on which the parties will return disclosed documents. The status and enforceability of these agreements is uncertain under English law, but there is scope for argument that CPR 31.B(3)(f) envisages this type of arrangement. For further consideration of these issues see our article for PLC Magazine here.
Celina McGregor
Celina McGregor
Senior associate
+44 20 7466 7460