Court Rejects Unilateral Decision to Apply Predictive Coding to Keyword Search Hits, Orders Production of All Hits Pursuant to Clawback Order as Stipulated

Progressive Cas. Ins. Co. v. Delaney, No. 2:11-cv-00678-LRH-PAL, 2014 WL 2112927 (D. Nev. May 20, 2014)

In this case, the parties agreed upon an e-discovery protocol which was memorialized in a court order.  Shortly into its review, Plaintiff determined that the agreed-upon methodology (manual review of search term hits) would be too time consuming and expensive and decided that it would instead apply predictive coding to those documents identified by the agreed-upon search terms—which it began doing without consulting the requesting party or the court.  The requesting party, FDIC-R (FDIC as Receiver), opposed Plaintiff’s unilateral action for several reasons, including the lack of transparency around the predictive coding methodology employed and that the predictive coding protocol did not comport with the recommended “best practices” for the chosen software program.  Ultimately, despite expressing support for the use of predictive coding in discovery, the court ordered Plaintiff to produce all of the documents identified by the agreed-upon keywords, subject to a clawback order, where such a production had been memorialized as an acceptable option in the stipulated order and where Plaintiff had abandoned the option it originally selected (manual review).  The court also noted the FDIC-R’s commitment to devoting the necessary resources to review the documents quickly and thus allow discovery—which had been “stalled for many months”—to move forward.

This case is one of nine declaratory relief actions filed by the plaintiff related to the coverage provided by certain previously issued insurance policies.  The parties submitted a joint Proposed ESI Protocol, which the court approved in a subsequent order.  Accordingly, the plaintiff applied agreed-upon search terms to the approximately 1.8 million documents it had collected, identified approximately 565,000 documents, and began a manual review—one of the two approved methodologies under the agreed-upon protocol.  Eventually, however, Plaintiff deemed the review too “time intensive and expensive” and began exploring its alternatives, including consultation with “a nationally-respected authority on e-discovery.”  Ultimately, Plaintiff determined that predictive coding would be more efficient.  Plaintiff then selected a software program and began to utilize predictive coding to identify relevant documents amongst those identified by the previous keyword search.  Plaintiff did not seek the agreement of the FDIC-R or the court.

The FDIC-R sought to compel Plaintiff to produce the nonprivileged keyword hits without review, subject to a clawback order—the second option contemplated by the stipulated ESI Protocol—or to require Plaintiff to proceed with predictive coding according to the FDIC-R’s suggested protocol, which included applying the methodology to the 1.8 million documents originally collected, and not just the keyword hits.  Among the reasons for the FDIC-R’s opposition to Plaintiff’s use of predictive coding was Plaintiff’s lack of cooperation or transparency around its predictive coding methodology, the likelihood of satellite disputes, and Plaintiff’s failure to adhere to the best practices recommended for the chosen software program.  Plaintiff also acknowledged that “applying predictive coding on top of search terms ‘is not always done.’”

In its analysis, the court plainly acknowledged its potential support for predictive coding and indicated that it would have approved a “transparent mutually agreed upon” predictive coding based protocol, had such a proposal been agreed upon from the outset, but took issue with Plaintiff’s refusal to “engage in the type of cooperation and transparency that its own e-discovery consultant has so comprehensibly and persuasively explained is needed for a predictive coding protocol to be accepted by the court or opposing counsel . . . .” (As the court noted, one of Plaintiff’s consultants was considered something of an e-discovery authority, and had previously written on both the need for cooperation and transparency in predictive coding and litigators’ reticence to accommodate that need.)

Noting its agreement with the FDIC-R that approval of Plaintiff’s proposal would result in more disputes and delay and reasoning that Plaintiff had “elected and then abandoned” the option to manually review the search hits, the court ordered that all of the hits be produced, subject to a clawback agreement.  The court indicated, however, that Plaintiff would be allowed to withhold documents identified by the use of privilege filters and provide a proper privilege log.  The court acknowledged that this would shift the cost of the review to the FDIC-R, but noted that the FDIC-R was “committed to devot[ing] the resources required to review the documents as expeditiously as possible” which would allow the long-stalled discovery to finally move forward.