In In re Biomet M2a Magnum Hip Implant Products Liability Litig., No. 12-MD-2391 (N.D. Ind. Apr. 18, 2013), the defendant used a two-step process to identify relevant documents for production within a voluminous collection of ESI. First, the defendant used “keyword culling” in a database of 19.5 million documents, locating 2.5 million potentially responsive documents. Second, the defendant employed a predictive coding system to identify relevant documents within the subset of 2.5 million. At the time of the plaintiffs’ challenge, the defendant had spent over $1 million on e-discovery costs. The plaintiffs argued that the use of keyword searches in the first step was inappropriate because, they contended, keyword searching is less accurate than predictive coding and therefore unduly restricted the pool of documents searched in the second step. The court rejected that argument, primarily on proportionality grounds. The court held “I can’t find that the likely benefits of the discovery proposed by the [plaintiffs] equals or outweighs [the] additional burden on, and additional expense to [the defendant]” under Fed. R. Civ. P. 26(b)(2)(C). The court concluded that it was not deciding “whether predictive coding is a better way of doing things,” but only that the defendant’s protocol “complies fully with the requirements of Federal Rules of Civil Procedure 26(b) and 34(b)(2).”