Keyword Searching is the Beginning, Not the End of Discovery Productions

Discovery productions are a mix of art and science. I am of the mindset that parties should bring a motion to compel when a production is inadequate, not over what search terms to use. This is one of those cases where the parties fought over how to search for ESI.

In T.D.P. v. City of Oakland, the Plaintiff requested the Court order the Defendant to search specific ESI with keywords. The Plaintiff’s second search request had the implication that the Defendants’ ESI needed to be individually searched for responsive data. T.D.P. v. City of Oakland, No. 16-cv-04132-LB, 2017 U.S. Dist. LEXIS 110731, at *10-11 (N.D. Cal. July 17, 2017).

That…would be a lot of work. The Defendants [rightly] argued keyword searching for responsive ESI would be sufficient, compared to manually reviewing the data. Moreover, the Defendants also claimed manually searching for responsive ESI would be both “impractical” and “oppressive.” T.D.P., at *11.

Anyone who has done document review the “old fashioned way” would agree with the Defendants. However, anyone who has done searches for responsive ESI with keywords alone knows full well the shortcomings of keywords.

The Plaintiffs argued that keyword searching would be “inadequate,” and relied upon the National Day Laborer case:

Simple keyword searching is often not enough: “Even in the simplest case requiring a search of on-line e-mail, there is no guarantee that using keywords will always prove sufficient.” There is increasingly strong evidence that “[k]eyword search[ing] is not nearly as effective at identifying relevant information as many lawyers would like to believe.”

T.D.P., at *11-12, citing Nat’l Day Laborer Org. Network v. United States Immigration & Customs Enf’t Agency, 877 F. Supp. 2d 87, 108-09 (S.D.N.Y. 2012).

The Court explained that the Nat’l Day Laborer addressed search strategies beyond keywords including predictive coding, not “falling back on attorneys physically poring over every piece of ESI created in the relevant period.” T.D.P., at *12.

Magistrate Judge Laurel Beeler held that there was nothing that demonstrated keyword searching would be inadequate in finding relevant ESI in the current case. T.D.P., at *13. The Court gave Defendants permission to use their search strategy. Judge Beeler did eliminate two overly broad search terms and offered the following advice:

Toward the end of discovery, when parties’ roles and communications are more obvious, it can be helpful to do targeted ESI searches among identified parties and identified times. (For example, all communications among these three people in this two-week period.) This level of specificity is usually possible later in discovery. The court thinks that the initial ESI production may illuminate subsequent targeted searches.

T.D.P. at *14.

Bow Tie Thoughts

Keywords are the most basic way to start document review. The only thing more basic would be launching a review application and starting relevancy review with DOCID000001. That being said, keywords should not be applied bluntly to a database.

If a request for production is for all emails within a specific date range, between specific individuals, regarding specific subjects, a “strategic search” can be created within those specific parameters. Those “hits” can then be reviewed for relevancy and privilege. This strategy can be repeated for each request for production.

That is the beginning to doing a production. If using a review application that has predictive coding, prediction models can be created based on individual requests for production or a model including all data identified as responsive.

The prediction model can be applied to un-reviewed records that were not hits to the search terms. The model can identify ESI for review that is likely responsive to the discovery requests.

It is important to treat responding to discovery requests as a process, not just running one search for ESI. Search terms can have both false positive and false negative results. An iterative process allows attorneys to narrow review based on search terms, and then leverage predictive coding to identify other potentially responsive ESI.