Last year, Magistrate Judge Andrew Peck of the U.S. District Court for the Southern District of New York issued the first judicial opinion recognizing predictive coding, also known as technology-assisted review, as a legitimate e-discovery tool. Since that opinion, the focus of the e-discovery world has been on predictive coding, seen by many as a more accurate way to identify relevant electronically stored information (ESI) for discovery than the predecessor technology, keyword searching. Other federal and state courts have followed Peck’s lead.
However, the cost effectiveness of an ESI production plan can vary considerably depending on such factors as the importance of the issues, size of dataset and resources of the parties. Specifically, keyword searching can be substantially less costly than predictive coding. And three recent cases demonstrate that keyword searching remains a defensible e-discovery plan.
In one case involving nearly 20 million electronic documents, the court upheld keyword searching as a method for the first phase of identifying relevant documents, with predictive coding then applied to the remainder, over objections from the plaintiffs who argued the keyword search may have eliminated relevant ESI. In a second case, the court reversed its earlier order requiring the parties to use predictive coding and agreed that keyword searching was acceptable due to the relatively small amount of ESI in question. In the third case, the court rejected the defendant’s claim that keyword searching would produce too many documents requiring manual review and therefore be too costly.
“Predictive coding will be appropriate in some situations but won’t be in others,” says Mayer Brown Partner Eric Evans.
In an ongoing class action products liability case in the Northern District of Indiana, In re: Biomet M2a Magnum Hip Implant Prods. Liab. Litig., defendant Biomet used keywords to cull a set of 3.9 million documents from 19.5 million potentially discoverable documents and attachments. The company proceeded even though the plaintiffs’ committee had asked it to wait for a court decision on consolidating multidistrict litigation. K&L Gates Partner Thomas Smith says the defendant was following the courts’ general rule that how to go about ESI production is within the sound discretion of the producing party.
After removing duplicates, the defendant was left with 2.5 million documents, to which it then applied predictive coding. The plaintiffs’ committee objected to Biomet’s use of keyword search, contending that the first phase eliminated a large number of responsive documents. According to the decision, the plaintiffs argued that “Biomet’s initial use of the keyword approach had tainted the process” and sought to require Biomet to redo its discovery, using only predictive coding.
In an April 18 opinion, the court held that the issue was not “whether predictive coding is a better way of doing things than keyword searching prior to predictive coding,” but rather, “whether Biomet’s procedure satisfies its discovery obligations.” The court concluded that Biomet was in compliance and accepted its approach.
The court also stated that the cost of the proposed predictive coding review of all 19.5 million documents was not proportional to the “comparatively modest” increase in the relevant documents that would be found.
In another decision reflecting proportionality in e-discovery decisions, a judge backed away from his earlier requirement that both parties use predictive coding.
In a decision last year, the Dela- ware Chancery Court in EORHB, Inc. v. HOA Holdings, LLC, had required the parties to retain a single discovery vendor to be used by both sides and to “use predictive coding.”
On May 6, upon agreement of the parties, the state trial court modified its prior order. It directed that the parties could now select their own vendors for computer-assisted review and permitted the plaintiffs to use the more traditional keyword searching, instead of predictive coding.
The parties had agreed that, based on the low volume of relevant documents, the cost of using predictive coding would likely be outweighed by any practical benefit of its use and they did not need to use the same discovery review platform or vendor.
No Undue Burden
In a third case, In re Coventry Healthcare, Inc.: ERISA Litigation, Maryland Federal District Court Magistrate Judge Jillyn K. Schulze ruled on March 21 that the keyword searching proposed by the plaintiffs would not create an undue burden.
The defendants claimed they tested the plaintiffs’ proposed search terms on the custodians’ data and hit 200,000 documents. They estimated it would cost $388,000 to process, host and review the data for responsiveness and privilege and argued that their resulting burden outweighed any benefit it offered the plaintiffs.
In response, the plaintiffs argued they had tried to collaborate with the defendants to “develop appropriate searches for ESI by limiting the searches to certain designated custodians” and by shortening the discovery period by three months.
While the court held that “to further reduce any undue burden, plaintiffs may need to refine their proposed search terms to narrow the pool of potentially relevant documents,” it rejected the defendants’ claim of undue burden. The defendants had not suggested “any alternative measures that could reasonably accommodate plaintiffs’ discovery needs other than negotiating more refined search terms,” the court said.
These cases demonstrate that arguing the proportionality standard and providing proof of an undue e-discovery burden when appropriate will continue to be persuasive in many cases. Courts have consistently required only that the e-discovery process be reasonable.
“There is no one-size-fits-all answer,” says Michael Hamilton, a senior e-discovery analyst at Exterro.