More On

E-Discovery Burden, the Judges' Guide, and an Alternative to AFAs

For years judges have relied heavily on the counsel appearing before them to learn how electronically stored information (ESI) had to be processed for review and production. Of course, at times the lawyers for the producing parties had more interest in arguing how burdensome and oppressive the production requests were than in enlightening the court about cost-effectiveness; at other times counsel may have had only limited experience with new technology.

The eDiscovery Institute, a 501(c)(3) nonprofit research organization, has just released a publication for judges that provides a detailed, vendor-neutral look at technologies and processes that can greatly reduce the cost of handling ESI, the "Judges' Guide to Cost-Effective E-Discovery," by Anne Kershaw and myself, with a foreword by the Hon. James C. Francis IV, Magistrate Judge for the Southern District of New York.

Used in combination, those technologies can cut ESI costs by 90 percent or more, a fact that seems particularly relevant given that ESI review costs can be the largest single budget item for litigation.

Among the technologies and processes covered in the Guide are:

  • deNISTing (i.e. not processing files known to be associated with commercial software, such as help files, user documentation, templates, etc., as compiled by the National Institute of Standards and Technology)
  • Duplicate consolidation across custodians as opposed to just within custodians (average savings: 38 percent for across-custodian vs. 21 percent for within custodian)
  • Email threading (36 percent more than just deduping)
  • Predictive coding (45 percent more than just deduping and email threading)

Parties who will be arguing burden before judges who are familiar with the technologies and processes outlined in the Guide ought to be prepared to show how their ESI processes take advantage of these known technologies or risk losing credibility with the court.

There has been much discussion about the use of alternative fee arrangements to lower legal bills. The hope of course is to reverse the somewhat perverse disincentive for law firms to be efficient under the hourly billing model. One option in lieu of hoping that competitive pressures introduced by AFAs will force firms to be more cost-effective would be to keep the hourly billing approach for litigation when it makes sense but explicitly require firms to use of processes and technologies as outlined in the Guide. If firms aren't using these technologies and processes it is virtually impossible for them to be cost-effective, and in win-win AFAs, corporate legal departments want their firms to be making a reasonable profit.

Oftentimes cost-cutting involves a trade-off between quality and cost, but in the case of these technologies there is a quality trade-up, not trade-off. For example, with proper duplicate consolidation, all the information about the various copies of emails and other files is preserved, but with only one copy being reviewed the risk of different decisions on different copies (e.g. privileged, responsive, nonresponsive) is practically eliminated. Grouping emails by email conversations or threads assists reviewer comprehension of the discussion and once again helps ensure consistent treatment of all the emails in the thread so that, e.g. the contents of an email that was listed as privileged don't get produced when they are included in the body of a later reply or forward that is produced.

As discussed in an earlier Inside Counsel column, with predictive review, at least the results are replicable. By contrast with traditional linear review the odds of a document that was selected as responsive being selected by a second reviewer are about comparable to a coin toss - see the studies cited in the Guide which show the odds to be between 48 and 62 percent.

The Judges Guide has been distributed in hard copy form to all U.S. Magistrate Judges and is available for download free of charge at

Join the Conversation

Advertisement. Closing in 15 seconds.