Exactly one year after the complaint was first filed in a gender discrimination suit, an important, brand new e-discovery opinion is out. Judge Peck has issued the first opinion approving use of computer-assisted review in Da Silva Moore v. Publicis Groupe, No. 11 Civ. 1279 (ALC) (AJP) (S.D.N.Y. Feb 24, 2012). The discovery order has been receiving a great deal of attention, and this week we were pleased to have Seth Row present a summary of the case in the Fios Winter Case Law Update webcast on February 28 (listen here ^[1]).

An American affiliate of a French parent company, Publicis Groupe, is in the advertising and public relations business. Headquartered in New York with offices nationwide, the defendant’s PR division employs over 45,000 people, 70% of whom are female. The plaintiff is a group of females – potentially a very large class – who allege gender discrimination, claiming that there is a glass ceiling for female executives in the company.

Undoubtedly, the case raises issues concerning the scope of preservation and the universe of documents, but the narrow discovery at issue in this opinion is how to search the electronic evidence, especially the emails that were collected.

The parties appear to have agreed on universe of custodians and data – somewhere around three million documents. The parties also appear to have agreed on using predictive coding to search through the documents to find responsive information as an alternative to using a more traditional method such as keyword searching. While parties appear to have agreed on the use of predictive coding (with multiple rounds of quality control testing), the dispute focuses on how to actually do it, and it now seems as if the plaintiffs no longer agree that predictive coding is an acceptable procedure.

So, what happened and what is yet to come?

In a previous hearing, the plaintiffs appear to have agreed to a jointly submitted predictive coding protocol, but before the discovery order was issued on Friday, they filed objections, which were ultimately rejected. Judge Peck addressed the objections directly in the order.

Plaintiff’s objection that predictive coding had not been subjected to Daubert analysis (the establishment of scientific validity through presentation of experts) was rejected by Judge Peck, who noted discovery processes are not subject to Rule 702’s gate-keeping function because they do not in themselves constitute trial evidence.

Plaintiff’s objection that predictive coding is untested and unreliable was also rejected by Judge Peck, who pointed out that perfection is not the standard in discovery. He points out that keyword searching is not perfect either, and an alternative solution is appropriate in the case at hand. Furthermore, in what seems to have been a genuine effort to achieve full cooperation and a high level of transparency, the defendant had agreed to disclose the seed set, including non-relevant coded documents, so that the plaintiffs could see the rationale that informed the predictive coding and assess its accuracy.

Plaintiff has filed objections to the magistrate’s order under Rule 72 with the Article III judge assigned to the case; however the Article III judge can only reject the order if he finds it clearly erroneous or contrary to law, a very high standard. 

Read the opinion (available here ^[2]). The jointly submitted predictive coding protocol is included. Also, monitor the docket for status updates on the case, including the outcome of a foreign privacy law issue and a pending jurisdictional challenge.