Introduction

When discovery involves large volumes of electronically stored information (“ESI”), efficient and thorough review within budgetary constraints and court-ordered deadlines can be a challenge. New technological advancements such as “predictive coding” (also known as “technology-assisted review”) have become attractive options in litigations involving large numbers of documents. Recent court guidance suggests that such advanced review approaches are appropriate in light of the proportionality framework governing discovery under the federal rules.

Early Decisions Addressing Predictive Coding

Predictive coding is a computerized process whereby complex algorithms analyze document sets using a relatively small “seed set” of documents that have been coded by knowledgeable attorneys. Once properly “trained,” the computer software can label the balance of documents as potentially relevant or otherwise, or rank the documents in order of most likely relevant to least, in order to facilitate various review approaches. Until recently, few courts have spoken on the use of this new technology.

In February of last year, Southern District of New York Magistrate Judge Andrew J. Peck issued the first opinion addressing technology-assisted review in Da Silva Moore v. Publicis Groupe, 287 F.R.D. 182 (S.D.N.Y. 2012), adopted sub nom. Moore v. Publicis Groupe SA, 2012 WL 1446534 (S.D.N.Y. Apr. 26, 2012) (“Da Silva Moore”). We previously reported on this decision in detail in our E-Discovery Update dated April 2012, available at http:// www.kramerlevin.com/Electronic-Discovery-UpdateApril-2012-04-01-2012. Judge Peck wrote in that decision that “computer-assisted review is an available tool and should be seriously considered for use in large-data volume cases where it may save the producing party (or both parties) significant amounts of legal fees in document review.” Da Silva Moore, 287 F.R.D. at 183.

Since Judge Peck’s ruling in Da Silva Moore, several courts have expressed increasing comfort with predictive coding approaches. See, e.g., Gordon v. Kaleida Health, 2013 WL 2250579 (W.D.N.Y. May 21, 2013); Global Aerospace Inc. v. Landow Aviation, L.P., 2012 WL 1431215 (Va. Cir. Ct. Apr. 23, 2012); Hinterberger v. Catholic Health Sys., Inc., 284 F.R.D.94 (W.D.N.Y. 2013). The Sedona Conference Journal, published by the nation’s leading think-tank on e-discovery issues, provided additional encouragement in a Fall 2012 article, noting that “[t]echnology-assisted review procedures have the potential to reduce discovery c o s t s and e xp ed i t e th e p r odu c t i on o f r e l e v an t , n on - privileged ESI.” Hon. Craig B. Shaffer, “DefensibleBy What Standard?, 13 Sedona Conf. J. 217, 234 (Sept. 2012) available at https://thesedonaconference.org/system/files/ LR-Defensible.by.what.standard.pdf; see also Maura R. Grossman & Gordon V. Cormack, Technology-Assisted Review in E-Discovery Can Be More Effective and More Efficient Than Exhaustive Manual Review, 27 Rich. J.L. & Tech., 11, 48 (2011).

In Re Biomet: Keywords versus Predictive Coding

Most recently, the Northern District of Indiana issued an opinion approving the use of predictive coding in a litigation involving tens of millions of documents. In re Biomet M2a Magnum Hip Implant Products Liab. Litig., 2013 WL 1729682 (N.D. Ind. Apr. 18, 2013) (“In re Biomet”).

The defendant Biomet, a medical products manufacturer facing a products liability lawsuit, started with a universe of 19.5 million documents, then used keyword searches to reduce the document count to 3.9 million documents. Id. at *2. After duplicate documents were removed from the set, the document count fell to 2.5 million. Id. at *1. Then, Biomet used technology-assisted review software to further refine the set of documents to be turned over to plaintiff’s Steering Committee. Id. at *2.

Plaintiffs objected to Biomet’s application of keyword searches to the initial pool of 19.5 million documents, contending that predictive coding should have been employed earlier in the process, before the keyword searches had reduced the universe of documents that would be analyzed by the software. Id. at *2-3.

Biomet argued that recommencing the discovery process at square one, and applying technology-assisted review to the initial set of 19.5 million documents would cost Biomet millions of additional dollars. Id. at *4. By April 2013, Biomet’s e-discovery expenses were $1.07 million and were estimated to eventually reach “between $2 million and $3.25 million.” Id. at *1.

In a short opinion written by Hon. Robert L. Miller, the court ruled that Biomet’s discovery disclosure practice had complied with Federal Rules of Civil Procedure 34(b)(2) and 26(b), and that restarting the discovery process, as proposed by plaintiffs, would run contrary to the proportionality provision embodied in Rule 26(b)(2) (C). Id. at *3.

Crucial to the court’s decision were the additional costs Biomet would incur by recommencing the discovery process, and the reasonable discovery efforts already employed by the parties. Id. Judge Miller acknowledged the possibility that “predictive coding , ins t e ad o f a keyword search, at Stage Two of the process would unearth additional relevant documents,” but denied that “the likely benefits of the discovery proposed by the [plaintiff] Steering Committee equals or outweighs its additional burden on, and additional expense to, Biomet.” Id.

Proportionality

Central to the court’s decision was Federal Rule of Civil Procedure 26(b)(2)(C)(iii), which states that:

On motion or on its own, the court must limit the frequency or extent of discovery otherwise allowed by these rules or by the local rule if it determines that:

. . . .

(iii) the burden or expense of the proposed discovery outweighs its likely benefit, considering the needs of the case, the amount in controversy, the parties’ resources, the importance of the issues at stake in the action, and the importance of the discovery in resolving the issues.

Fed. R. Civ. P. 26 (b)(2)(C)(iii).

In addition to suggesting that keywords and advanced software may be used in combination when undertaking document review, perhaps the most noteworthy aspect of Judge Miller’s ruling is the focus on Rule 26’s proportionality standard. In many ways, In re Biomet serves as strong reinforcement to the thematic overtone of Judge Peck’s ruling in Da Silva Moore. Judge Peck advised that “‘[c]ourts and litigants must be cognizant of the aim of Rule 1, to “secure the just, speedy, and inexpensive determination’ of lawsuits,” and that the “proportionality doctrine set forth in Rule 26(b)(2)(C)” furthers these concerns. Da Silva Moore, 287 F.R.D. at 191.

Further, just as Judge Peck in Da Silva Moore noted that while “computer-assisted review is not perfect, the Federal Rules of Civil Procedure do not require perfection,” Judge Miller in In re Biomet acknowledged that some relevant documents out of the 19.5 million may have remained undisclosed, but the Federal Rules of Civil Procedure concerning discovery obligations were nonetheless fulfilled. Da Silva Moore, 287 F.R.D. at 191; In re Biomet 2013 WL 1729682, at *3.

Conclusion

Predictive coding has the ability to save litigants many hours of laborious and costly document review work. Yet the prospect that a number of responsive documents may be omitted from production rests uneasily with counsel and parties who have for decades relied upon the manual review processes. However, as exemplified in Biomet, courts have begun to recognize that no document review and production process is perfect, that perfection is not required, and that emerging technology can therefore play an important role in efficiently fulfilling discovery obligations within the proportionality framework set out in the federal rules.