The volume of electronically stored information (“ESI”) has exploded, often making discovery prohibitively expensive. Supporters of “predictive coding” champion it as a way to reduce significantly the cost of discovery. Opponents, however, label predictive coding as an unreliable and unproven technology that can erroneously exclude responsive documents from production. Recent case law evidences a trend toward court approval of the practice. See, e.g. Moore v. Publicis Groupe, 287 F.R.D. 182 (S.D.N.Y. 2012) adopted sub nom. Moore v. Publicis Groupe SA, 11 CIV. 1279 ALC AJP, 2012 WL 1446534 (S.D.N.Y. Apr. 26, 2012); FDIC v. Bowden, CV413-245, 2014 WL 2548137(S.D. Ga. June 6, 2014). In a matter of first impression for the United States Tax Court, Judge Ronald L. Buch recently followed this trend by overruling the Internal Revenue Service’s objection to the petitioner’s request that it be permitted to use predictive coding to review voluminous documents saved on two backup storage tapes. See Dynamo Holdings Ltd. v. Commissioner of the Internal Revenue Service, 143 T.C. No. 9 (2014). The Dynamo decision is notable not only for its approval of the use of predictive coding, but also for its emphasis on the need for transparency and cooperation when using new technologies in discovery.
Dynamo presents a consolidated tax proceeding in which the IRS determined that certain transfers to the petitioner Dynamo constituted gifts, not loans, for purposes of their tax treatment. During the discovery phase of the dispute, a motion was brought before Judge Buch concerning Dynamo’s request that it be permitted to use predictive coding to review large volumes of ESI stored on two back-up tapes. Id. Dynamo presented expert testimony estimating that the back-up tapes contained well over 3.5 million documents and that manual review of such documents would cost over $500,000. Id. at 6. To ameliorate that burden, which Dynamo characterized as the result of a “fishing expedition” into back-up tapes, Dynamo proposed the use of predictive coding. Dynamo’s expert estimated that the use of predictive coding would reduce the set of documents needing manual review to 200,000-400,000 documents, at an estimated cost of approximately $80,000. Id. at 6. The IRS objected to the use of predictive coding and asked the court to reject it as an “unproven” and unreliable technology. Id. at 5.
The Court’s Analysis: Adopting Predictive Coding as a “Happy Medium”
The provisions of the U.S. Tax Rules of Practice and Procedure governing discovery before the U.S. Tax Court closely correspond to similar provisions in the Federal Rules of Civil Procedure. Id. at 3. While under such rules, the court “is not normally in the business of dictating the process [parties] should use when responding to discovery,” the Dynamo court recognized that because the use of predictive coding was an issue of first impression before the U.S. Tax Court, it would entertain the parties’ dispute on that issue. Id.
The basis for Dynamo’s request to use predictive coding was its position that predictive coding was the only way to “efficiently and economically . . . identify the nonprivileged information [on the back-up tapes] that is responsive to [the IRS’] discovery request.” Id. at 2. In response, the IRS countered that Dynamo could simply produce the entire content of both back-up tapes (including documents containing privileged and/or confidential information), and that the IRS would sign a “clawblack” agreement whereby any such protected documents contained in the production could be withdrawn without waiver. Id. at 3.
In rejecting the IRS’ position, the court held that Dynamo “reasonably resist[s] entering into any [clawback] agreement as part of a plan under which they would voluntarily allow [the IRS] to see all of the privileged or confidential information on the requested tapes.” Id. at 4. The court further held that “given the time and expense involved with [Dynamo’s] review of all the ESI [on the back-up tapes] . . . we likewise do not consider it appropriate to order [Dynamo] to go to that extreme either.” Id. Thus, the court determined that predictive coding served as a “happy medium” between the competing interests of Dynamo and the IRS. Id.at 4. The court described predictive coding as an “expedited and efficient form of computer-assisted review” that can reduce the time and costs spent on the manual review of documents, and rejected the IRS’ position that predictive coding is an “unproven technology.” Id. at 4-5.
Perhaps the most notable aspect of Dynamo was the court’s emphasis on the need to be transparent and cooperative when using new review technology, such as predictive coding. Id. at 5. The court endorsed Dynamo’s “represent[ations] to the Court that they will retain electronic discovery experts to meet with [the IRS’] counsel or his experts to conduct a search acceptable to [the IRS.]” Id. The court thus distinguished Dyanmo’s assurances from the parties’ behavior in Progressive Cas. Ins. Co., a District of Nevada case. Id. (citing Progressive Cas. Ins. Co. v. Delany, 2:11-CV-00678-LRH, 2014 WL 3563467 (D. Nev. July 18, 2014) at *10-*12. In Progressive, the Nevada District Court denied permission to use predictive doing where “the record lacked the necessary transparency and cooperation among counsel.” Id. In contrast, the Dynamo court was satisfied that Dynamo’s transparency would ensure an agreeable result. The court further noted that if, after the use of predictive coding, the IRS found the production incomplete, it could file a motion to compel at that time. Id. at 7.
• The U.S. Tax Court’s decision in Dynamo further evidences the growing trend of judicial approval of predictive coding.
• The court’s decision highlights the importance of cooperation and transparency among parties during the discovery process, especially if requesting the use of new technology, such as predictive coding.
 Predictive coding is a software technology that employs various techniques and algorithms such that after a small sample of documents is reviewed manually, the coding decisions can be automatically populated to a much larger set of documents in order to reduce substantially the total number of documents needing manual review. Predictive coding is also known as Technology Assisted Review (“TAR”) or Computer Assisted Review (“CAR”).