ALM Properties, Inc.
Page printed from: Corporate Counsel
Select 'Print' in your browser menu to print this document.
McDermott and DOJ Embrace Predictive Coding
Law Technology News
The legal community has been increasingly eager to identify opportunities to leverage review technology to better manage the scope and cost of discovery. McDermott Will & Emery recently took advantage of such an opportunity when members of its antitrust and discovery groups negotiated with the U.S. Department of Justice to develop a win-win predictive coding protocol that met the department's and our client's needs.
McDermott is an international firm, with more than 1,100 lawyers and 20 litigation support personnel. We have eight offices in the United States, 10 overseas and an alliance with MWE China Law Offices.
In the summer of 2012, our client, Constellation Brands, agreed to purchase the U.S. beer businesses of Corona and related brands from Anheuser-Busch InBev. That purchase was contingent upon the results of a DOJ investigation into Anheuser-Busch Inbev's acquisition of Mexican brewer Grupo Modelo. Time was of the essence for all parties. The deal, if approved, would transform Constellation Brands into the third largest brewer in the U.S. market.
The mutual decision between the DOJ and Constellation Brands to use predictive coding (aka technology-assisted review) proved to speed the review process and the DOJ's investigation of the merger.
Constellation Brands benefited because it was able to meet a tight deadline while saving significant costs. Using kCura's Relativity electronic data discovery software, our McDermott team reviewed millions of documents within two weeks to substantially comply with the DOJ's request for information.
The DOJ, on the other hand, benefited by receiving a targeted production that included a high percentage of information that was helpful to its analysis of the proposed merger. The DOJ ultimately approved that merger in April 2013.
The collaborative use of predictive coding technology taught all parties important lessons about cost and time savings, but there were other additional insights along the way that will help all of us in future interactions. Here are the top five lessons we learned.
1. Take a deep breath
Attorneys involved in the merger review process are under the gun from the outset. Clients want their proposed merger approved as soon as possible and want their lawyers to respond to the DOJ's requests without delay. This often leads to a flawed approach, where a legal team hurriedly begins a review project before it has the information necessary to properly plan it, leading to spinning wheels and wasted resources.
When parties intend to use advanced technology, such as predictive coding, their legal and technical consultants should take a deep breath and spend time laying out the plan before jumping in. Thoughtful planning at the front end will save time and money on the back end.
2. Details matter
When working with the DOJ, make sure that you begin the communication process early and remain open and honest throughout the process, so everyone knows the rules of the game. The DOJ has detailed protocols for all aspects of the discovery spectrum, from what to collect through how to cull it to the form of production. Without waiting for a clear, comprehensive plan to be developed through an unimpeded communication channel, clients will very likely end up redoing or correcting something that they did, spending unneeded time and incurring unnecessary costs.
For example, in this particular matter, the DOJ did something somewhat unusual by prohibiting the use of search terms when using predictive coding technology. Had the client used its internal resources to apply search terms to identify the information that its law firm would eventually collect (as is often done), those efforts would have been wasted, and a second collection would have had to be performed, thus offsetting the savings realized through the use of the technology
For example, the millions of Constellation Brands documents reviewed in just over two weeks using predictive coding would have taken five attorneys more than two years using more traditional methods — an amount of time no merger party has the luxury of taking.
3. Take it from the top
Predictive coding represents a different approach to document review. In traditional "linear" review projects, the task of wading document-by-document through hundreds of thousands of documents gets pushed down to the least-expensive qualified reviewer available — because each individual decision made by those reviewers is of relatively low value.
With predictive coding, however, each decision is crucial because it will help "teach" a computer algorithm how to find the information you're looking for. Thus, each decision will result in thousands of other computerized decisions to be made across the data set.
One wrong decision could have a substantial ripple effect that will have to be corrected in future iterations of decision-making "seed sets," creating additional effort, frustration, costs, and delay. Consequently, senior members of the team — who are most likely to get each decision right because they are most familiar with the case and its issues — must be involved in reviewing the sample (seed) sets of documents.
4. Have confidence in the technology
Successful matters showcasing the use of predictive coding are helping bring this methodology forward into more common use, but it is still in its infancy. While there continue to be legal and practical challenges to technology-assisted review, it is important not to lose sight of the significant potential benefit to clients. In this case, McDermott's suggestion to the DOJ that predictive coding be used resulted in three important benefits:
5. The rewards are obvious, but the risks might not be
With all the benefits, potential risks cannot be ignored. DOJ's protocols may require greater transparency than most clients are accustomed to providing. For example, the DOJ may require a validation process in which a DOJ "clean team" reviews a sample set of documents drawn from both the responsive and non-responsive data sets.
This process creates the potential, however remote, for the disclosure of troubling or damaging documents that would otherwise have been withheld from the government. As a practical matter, the likelihood of turning over a truly problematic document in a small, random sample of documents is very low (and some would argue much lower than if the review was performed entirely by humans in a traditional, linear review). But it remains a real risk that should be disclosed to the client and weighed against the benefits.
It is the obligation of counsel to understand the possibilities along with the risks, and advise clients accordingly. More often than not, the predictive coding rewards will substantially outweigh the minimal risks, particularly in the growing number of projects when both sides of a matter can sensibly agree on the value of using predictive coding and other state-of-the-art technology.
Geoffrey Vance (email@example.com) is a litigation partner and head of McDermott Will & Emery's discovery practice group, based in Chicago. Alison Silverstein (firstname.lastname@example.org) is the group's managing director, based in Washington, D.C. Antitrust partner Warren Rosborough contributed to the article.
This article originally appeared in Law Technology News.