Technology has fundamentally changed litigation discovery obligations, processes and, most importantly, expenses. The days of clients sending file folders or bankers boxes of paper files to lawyers to review and produce are long gone. And, correspondingly, the expense of electronic discovery is the most rapidly increasing item in the average litigation budget.

In fact, a 2010 Duke University survey of major companies revealed that from 2006 to 2008 the average discovery costs ranged from $621,000 to just less than $3 million for average litigation matters. For larger litigation, the costs spanned from $2.3 million to $9.7 million.

This rapid and continuing growth in e-discovery expenses is even more alarming when one considers that there is no evidence that the increasing cost of discovery has resulted in a corresponding increase in the volume of relevant or important material being produced in litigation. Rather, in the major cases that went to trial in 2008, it is estimated that the ratio of pages discovered to pages ultimately introduced as exhibits at trial or arbitration was a mere 1,000-to-one. On average, almost 5 million pages of documents were produced; however, only 5,000 of them became exhibits.

E-discovery practitioners, alongside trusted vendors, continue to develop new search methodologies that use technology to reign in the growing costs and time associated with electronic document review. One of the newer strategies being employed is known as “computer assisted review,” also known as “predictive coding” or “predictive analytics.”

Predictive analytics is the nonspecific term for a computer program that uses complicated algorithms to sample and predict relevancy across large collections of electronically stored information (ESI). Essentially, after attorneys and document reviewers identify representative responsive documents from a large document pool, those documents are analyzed by the software program, which is then run against the entire data collection in order to find documents that are conceptually related to the representative documents identified by the document reviewers (i.e., “more like this”).

Other aspects of technology-assisted review include being able to:

  • Organize and prioritize entire document collections for more efficient review
  • Categorize documents by key phrases, concepts and names
  • Determine the relevance, responsiveness and privilege status of any document
  • Locate all similar or related documents irrespective of keywords
  • Algorithmically reduce the number of documents requiring eyes-on review, therefore reducing the related billable hours

Predictive coding appeals to lawyers, particularly in-house counsel, because it aims to leverage technology to reduce the need—and expense—for human attorneys reviewing every document. However, there are several formidable arguments against it:

  • It is not yet a standardized methodology
  • There is still a general lack of understanding among the industry as to how this technology accomplishes what it does and how effective it really is
  • There are concerns about the need for required non-attorney personnel and lack of transparency in the behind-the-scenes technical process
  • There have been instances of inconsistent interpretations of the same data, unnecessary production of nonrelevant data and inadvertent production of confidential or potentially privileged data

In order to maximize the benefits and cost reductions with predictive analytics, the litigation e-discovery team must be a proactive participant in every stage of the process. This team should consist of at least one senior ESI counsel or e-discovery information lawyer who is experienced with the new technologies and how to avoid the pitfalls associated with the variety of methodologies.

Most importantly, it is critical that your discovery team remember that human judgment is still required. For predictive coding to be effective, experienced attorneys intimately familiar with the case must initially identify a proper “learning” data set before the computer can begin its processing. Successful use of predictive coding depends upon the accuracy of the original data sample, which ideally will originate from the reasoned analysis of experienced document reviewers and supervising attorneys.

Predictive coding alone is not a solution for companies facing complex litigation with huge amounts of email correspondence and ESI. Nor can it replace attorneys in conducting the review. It can, when used and controlled properly, be an efficient and cost-effective tool. Companies and counsel must make sure that lawyers and others employing the applications understand the limits as well as the benefits.

The time and expense of predictive coding processes can be a fraction of a full manual review, but potential users should proceed with caution. Now that the first legal decisions regarding the acceptability of predictive coding are surfacing, its application will also likely evolve to one day require less scrutiny. But until then, the effective and appropriate use of predictive coding and analytics will require careful and precise management by the attorneys, parties and the courts altogether.