*1 Pursuant to my order of September 12, 2013 (“Sept. 12 Order”), UBS Real Estate Securities Inc. (“UBS”) provided the plaintiff trusts (the trustee of which is U.S. Bank National Association, and which I will call, in conformance with the September 12 Order, U.S. Bank) with 2,719 randomly-sampled documents from two of UBS' so-called “Repurchase Custodians” (the “Sample Documents”), as well as a privilege log containing entries for 411 documents (the “Sample Privilege Log”). The plaintiff has reviewed those documents and submitted its report.

According to the plaintiff, 132 of the Sample Documents “are both clearly responsive to at least one of [U.S. Bank's] document requests, and are not duplicates of any other documents produced by UBS. Moreover, 35 of these 132 documents were hit by the search terms UBS was required to implement pursuant to the parties' agreed search protocol.” (Letter of Sean P. Baldwin dated Sept. 20, 2013 (“9/20/13 Baldwin Ltr.”) at 2). Of the documents included on the Sample Privilege Log, U.S. Bank asserts that “33 ... almost certainly contain at least one search term and are clearly responsive,” because the entries for these documents contain the name “Assured” (Assured Guaranty Municipal Corp. (“Assured”) was the plaintiff in a related case and is the entity providing certain insurance to UBS connected with the repurchase demands at issue in this case); the acronym “AGO,” which is Assured's ticker symbol; or the name of one of the relevant transactions.[1] (Letter of Sean P. Baldwin dated Sept. 24, 2013 (“9/24/13 Baldwin Ltr.”) at 2 & n. 2). U.S. Bank also identifies 73 Sample Documents listed on the Sample Privilege Log that are “likely to be responsive[ ] and ... likely to be hit by search terms.” (9/24/13 Baldwin Ltr. at 2).

UBS contends U.S. Bank's conclusions are “inaccurate.” (9/23/13 Musoff Ltr. at 1). According to UBS, review of the 35 documents identified by U.S. Bank as containing agreed-upon search terms shows that nine are duplicates of documents produced or logged in the original production, one does not contain any search terms, and most of the rest “plainly have no relevance and would not be deemed responsive by a reasonable reviewer.” (Letter of Scott D. Musoff dated Sept. 26, 2013).

Obviously, the parties disagree about the relevance and responsiveness of the documents at issue. I have reviewed the documents and find that they are arguably responsive and relevant, given the “extremely broad” definition of relevance in the discovery context. See Assured Guaranty Municipal Corp. v. UBS Real Estate Securities Inc., No. 12 Civ. 1579, 2012 WL 5927379, at *2 (S.D.N.Y. Nov. 21, 2012) (quoting Condit v. Dunne, 225 F.R.D. 100, 105 (S.D.N.Y.2004)). While these documents may not be “critical” (9/23/13 Musoff Ltr. at 2), that is not the appropriate test.

UBS points out that, out of a universe of hundreds of thousands of documents, U.S. Bank has identified fewer than 200 that are responsive or likely to be responsive; thus the percentage of responsive documents is quite low. (Letter of Scott D. Musoff dated Sept. 20, 2013 (“9/20/13 Musoff Ltr.”) at 2; 9/23/13 Musoff Ltr. at 1). It argues that UBS' “initial review was at a minimum on a par with and likely more accurate than typical document reviews,” citing cases and studies finding that human reviewers are disconcertingly error-prone. (9/20/13 Musoff Ltr. at 2–3); see, e .g., The Sedona Conference Best Practices Commentary on the Use of Search and Information Retrieval Methods in E–Discovery, 8 Sedona Conf. J. 189, 195 (2007) (“A consensus is forming in the legal community that human review of documents in discovery is expensive, time consuming, and error-prone.”).

*2 First, UBS ignores the fact that the error rate of its review is actually quite high. As U.S. Bank points out, if the 298,000 e-mails of the Repurchase Custodians that have not been produced or logged contain the same ratio of responsive documents, the number culled will be significantly larger than the 6,000 documents that have, so far, been produced or logged from the Repurchase Custodians' files. (9/20/13 Baldwin Ltr. at 3; 9/24/13 Baldwin Ltr. at 3). That is, the review missed more responsive documents than it identified. Second, UBS' argument that its results were within the parameters of a typical review conducted by human reviewers sets too low a bar. UBS can, and has, utilized review technologies that can, if used properly, be expected to identify more than a mere half of the potentially responsive documents. UBS must re-review the documents in the Repurchase Custodians' files.[2]

U.S. Bank asks that UBS include additional search terms likely to identify responsive, relevant documents. (9/24/13 Baldwin Ltr. at 3–4; 9/20/13 Baldwin Ltr. at 3–4). The parties have not met to confer on this request, as is required by the Federal Rules of Civil Procedure and the Local Rules. Fed.R.Civ.P. 37(a)(1); Local Rules of the United States District Courts for the Southern and Eastern Districts of New York 37.3(a). Therefore, within one week of the date of this Order, the parties are directed to meet and confer regarding the inclusion of additional search terms, keeping in mind that, if they fail to resolve this issue and the terms agreed upon to date prove inadequate, the documents will have to be reviewed a third time.

Conclusion

UBS is ordered to review, once again, the files of the Repurchase Custodians. Within one week of the date of this Order, the parties shall meet and confer regarding the expansion of the search terms to be used in this review. The request for an extension of time to complete discovery, on which I deferred ruling (Sept. 12 Order at 14), is denied as moot (Memorandum Endorsement dated Sept. 16, 2013).

SO ORDERED.