The E-Discovery Consulting & Services practice within FTI Consulting’s Technology segment recently completed e-discovery for a Federal Trade Commission quick look as part of the second request for a merger between two healthcare industry companies. Using Brainspace’s Technology Assisted Review (TAR) model, the team developed novel workflows to enable efficient e-discovery in the face of a number of unique circumstances. We recently chatted with the team involved on the case, which was led by Senior Managing Director Wendy King and included Managing Director Jon Murdock and Senior Director Myron Williams, to understand how the matter was different from other second request inquiries. They discussed the challenges involved and shared details around the new workflows FTI developed as part of the engagement.

FTI Technology Insights Webinar Series - Second Request Case Study Roundtable
Recorded Webcast | FTI Technology Insights Webinar Series - Second Request Case Study Roundtable
This case involved a number of challenges. Can you explain those?

Wendy King: Our team has a lot of experience in dealing with second requests, which typically involve high-stakes scenarios, large volumes of data and tight timelines. All of these factors were present in this case—we were dealing with a collection set of nearly 4 million documents, some of which had been self-collected by the client. Other documents had been collected from mobile devices using a third party vendor, and dozens of boxes of paper documents were identified at the last minute. So right out of the gate, we were delayed in getting the dataset into our environment to begin discovery. That was when we started seriously considering Brainspace’s Continuous Multimodal Learning (CMML), a flexible interactive TAR tool, as a key way to ensure we could get through the documents on time. The workflows we created to meet the regulator’s parameters, quickly stabilize the Supervised Learning model and get through the documents at a fast pace were really the most unique element of this matter.

Myron Williams: What Wendy hasn’t mentioned yet is that as we were getting underway, the COVID-19 pandemic was just hitting the U.S. States were beginning to issue stay-at-home orders, and we didn’t know if our reviewers would be allowed to continue working in the review center. We didn’t know whether or how the new restrictions would affect the production.

King: Right. So while it’s not unique to have time constraints, or large volumes of data, or the need for a courier to help deliver a production in these types of matters, it is unique to be forced to pivot unexpectedly under that kind of pressure. And to face so many unknowns about whether reviewers will be forced to begin working remotely (causing extreme delays), whether the FTC will be available to accept a production in-person as planned or what the cost implications might be for our client.

So, all of these factors, and only three weeks to complete e-discovery. What steps did the team take at the outset to hit the ground running?

Williams: While the pandemic added a lot of unknowns, we were already in a good position because of the tech and innovative workflows our team set up from the start. These had all been applied before COVID-19 started threatening to impact the project. Had we been using a different set of analytics and workflows, it would have been much harder to hit the deadline once things escalated due to the pandemic.

John Murdock: The workflow allowed us to move quickly and efficiently through the documents so the reviewers were able to get through everything over the weekend, before lockdowns went into effect. Were it not for having an effective predictive model in place, we would not have been as far along in the project, and the reviewers may not have been able to get through the documents as efficiently as needed.

King: It’s also worth mentioning that our team’s experience in these types of matters (COVID-19 factors aside) gives us a strong running start. We are deeply familiar with the tools available, which types of workflows lend themselves to which types of matters and the nuances of working with regulators. Because of this experience, we start out every matter with sound project management and strategy, which makes a huge difference in hitting deadlines and getting ahead of unexpected challenges.

You’ve all mentioned that the workflows were the most unique aspect of this matter. What were the specific innovations?

King: With Brainspace’s CMML, we began training the analytics model before all the data was received. One team of five reviewers (subject matter experts) was given sample sets of documents for which they determined, based on predetermined criteria, the relevancy of each document.

Murdock: While that work was ongoing, another team of reviewers was provided with documents that could not go through the prediction process, including images, multimedia files and scanned paper documents from which text could not be easily pulled.

Williams: A third team, comprised of specialized reviewers, worked on a highly focused review of the predicted responsive documents.

Murdock: Overall, using these three teams working in tandem, and analyzing metadata in Brainspace, we cut the dataset sent for review from roughly 40 percent considered responsive to only approximately 20 percent that was ultimately deemed responsive and sent for review.

FTI’s experts are familiar with building workflows across all of the leading e-discovery and analytics platforms. Why was Brainspace the right fit for this case?

Williams: Brainspace’s CMML offers a lot of ways to train the model. One that’s really beneficial is that it allows us to find the documents that it thinks will serve for training the model. When it shows us documents for review that are better suited for training—both responsive and not responsive—it makes for a tighter model and faster path to the end result. It also allows simultaneous review and places less emphasis on judgement calls from specialized reviewers.

King: This is a second generation of technology assisted review (TAR), and really makes a difference in taking many variables into account and leveling out human decision making in training the predictive model. This and the fact that the tool supports simultaneous review, allowed us to move very quickly to meet the deadlines and get the review done before COVID-19 lockdowns hindered our progress.

Will the workflows created for this case be applied to future matters? If so, what are some of the potential benefits for clients?

Williams: We’re already using the domain analysis approach for other second requests to help reduce documents sent for privilege review.

Murdock: Our teams are always looking for ways to use smart workflows to reduce the population in a matter without compromising the work product. In this matter, we helped the client save more than $100,000 just by applying analytics in a smart way. We defensibly reviewed nearly 4 million documents in a matter of weeks, and delivered productions on-time to the FTC, under the exacerbated circumstances of the mounting global crisis. The lessons learned, workflows used and benefits that were realized are all repeatable for future client matters.