Predictive Coding: Not Just for E-Discovery

    Feb 25, 2014

    Relying on employees to appropriately label and manage records is a flawed approach, according to a panel of experts who spoke at the recent LegalTech New York. Law Technology News reported that one of the experts, Warwick Sharp, vice president of marketing and business development at Equivio, equated this approach to a game of telephone in which the retention officer “is responsible for sending a company’s record policy all the way down the chain to a company’s last user employee (who is ostensibly responsible for labeling files in his own email with up to 300 categories).”

    The impracticality of this approach is obvious, especially when a company has thousands of employees and hundreds of policies, Sharp said – and this is assuming that every employee is trying to follow the policy. In his opinion, predictive coding could be a much better solution considering that a records retention expert can train software to label documents with categories. The goal: an automated process that is consistent, systematic, and defensible. 

    Panelist Laura Kibbe, managing director of expert and professional services at Epiq, related a case study that demonstrated the usefulness of predictive coding to one of Epiq’s clients. Kibbe said the client was able to clean up the company’s e-mail repository by categorizing 40% percent of 1.4 million stored documents as junk e-mails, with 80% accuracy – a rate that has been accepted as legally defensible in court, according to the panelists. This approach yielded substantial savings in storage and e-discovery costs for that material alone, persuading the company to further analyze and categorize the not-junk category. 

    The use of predictive coding is not a new technology. It has been used in e-discovery and is expected to become more commonplace. Its role in overall information governance, however, is fairly new and one that reportedly came up throughout the convention. The experts were quick to remind their audience that the goal of predictive coding isn’t perfection, but that perfection wasn’t possible in the days of paper either. It is, however, a more defensible process than relying on 20,000 employees to follow a detailed schedule.

    For more information about the use of predictive coding for information governance and to reduce e-discovery costs, see these free ARMA International resources:

    © 2016, ARMA International