The search term Tug-of-War: but what about machine learning.

Posted by Servient Team -


The lead counsel in any litigation is in a unique position as she or he stands before the court representing the client, engaging with opposing counsel, discovering evidence, and managing the entire process as a case moves towards resolution. The stakes are often very high, and litigators are no doubt entitled to a little leeway. The frustrating thing in matters related to ESI (electronically stored information) is the ease with which excellent counsel can step into the quagmire that results from merely using search terms to surface relevant evidence.


So it’s always worth noting examples of the widespread acceptance by the courts of advanced search technology, and the availability of industry experts to testify in support of it. It’s also useful to be aware of recent court decisions showing that merely using keyword search is difficult and dated.


A prime case in point emerges from the recent Webastro Thermo & Comfort v. BesTop, Inc., 2018 WL 3198544, No.16-13456 (E.D. Mich. June 29, 2018). This was a patent case involving a convertible automobile’s roof and roof-opening mechanism. Of particular interest to us, though, is a ruling handed down in the course of this litigation that supported a motion for a protective order for email discovery. The motion was granted by Magistrate Judge R. Steven Whalen after finding that the Plaintiff had met its burden of good cause by “sufficiently articulat[ing] specific facts showing clearly defined and serious injury resulting from the discovery sought.”


The backdrop for this clash was a stipulated court order governing the production of ESI, intended to promote early resolution of ESI disputes without court intervention. This ESI order provided that “each requesting party shall limit its email production requests to eight (8) key custodians and a total of ten search terms per custodian per side.” It also provided guidance on search term construction, with examples of prohibitively broad searches and methods for reasonably narrowing search terms.


In support of its motion (filed after the parties had reached an impasse regarding the ESI discovery), the Plaintiff filed two declarations. The first included attachments of correspondence in which defense counsel proposed search terms and plaintiff’s counsel objected to their breadth. Plaintiff’s counsel specified that a pre-production search using the defendant’s proposed terms yielded an astonishing 53 gigabytes of email for just three of the eight custodians.


The Plaintiff’s second declaration stated that “just a subset of the email discovery requests propounded by [Defendant] have returned more than 614,000 documents, comprising millions of individual pages for production.” Counsel further declared that the first 100 sequentially numbered documents were not relevant to the “claims or defenses of the parties”, but rather involved other business of the Plaintiff. Finally, counsel noted that the search term “sales” had yielded documents only because of job title signatures in individual emails.


In response, the Court agreed that the Plaintiff had “provided hard numbers as to the staggering amount of ESI returned based on Defendant’s search requests” and distinguished a case cited by Defendant’s in which the party had offered no facts but merely made conclusory claims to that “producing such emails would increase the email universe exponentially.”


The Court reasoned that many of the Defendant’s search terms were “overly general on their face”, citing specific examples of the Defendant propounding requests with Plaintiff’s product name, which had been expressly prohibited by the ESI order. The Court therefore ordered the parties to meet and confer to focus and narrow the Defendant’s search terms. The Plaintiff’s request for cost-shifting was denied, the court ruling that this would only be considered if the Defendants failed to reasonably narrow their requests.


The lesson in this case? Perhaps that engaging an information retrieval expert in advance would have been helpful in formulating the ESI order. Machine learning platforms are out there waiting to be exploited with defensible, cost effective results.

Recent Posts

Linklaters: Using Technology To Change The Delivery Of Legal Services

read more

New York Courts Embracing Machine Learning

read more

The search term Tug-of-War: but what about machine learning

read more