Article / By Caitlin Kasmar

Predictive Coding and Threading.

While predictive coding addresses many of the shortcomings of the traditional linear-review model, if not approached thoughtfully, it may amplify some of the very shortcomings it purports to address.

View All

Many attorneys have now embraced predictive coding as a viable cure for discovery ills—a notable and significant change in attitude six years after Judge Peck’s ringing endorsement of the technology in Da Silva Moore. But a panacea it is not.

While predictive coding addresses many of the shortcomings of the traditional linear-review model, if not approached thoughtfully, it may fall short of addressing some of the challenges it purports to address.  To address some of those pitfalls and achieve optimal efficiency, attorneys should consider combining predictive coding with another Technology Assisted Review (TAR) tool: email threading.

Predictive Coding: A Brief History

Predictive coding was born in direct response to the explosive discovery costs as email became increasingly commonplace twenty years ago.  It has been a belated technological solution to a problem created by technology itself, promptly followed by risk-averse lawyers mulling its merits, still longing for the days when one could analogize a gigabyte of email to a mountain of boxes in exchange for nods of judicial sympathy.

Today the e-discovery marketplace may seem deluged with a growing list of competing technology platforms that relegate the linear review model back to the dawn of the century where it belongs.  At a high level, the underpinnings of the predictive coding workflow are virtually the same: (i) attorneys review document samples to identify relevant documents; (ii) the predictive coding system uses the attorneys’ coding to identify additional potentially relevant documents; and (iii) human reviewers fine tune the system by confirming or overturning its categorization of relevant and not relevant documents.

Key Takeaway

Email threading dramatically reduces the number of emails that reviewers have to analyze.

The Importance of Consistent Coding

Yet the efficiency of the predictive coding systems’ workflow may suffer due to a familiar culprit of the time-tested linear review workflow: the propensity for different reviewers to code conceptually similar documents differently.  Such inconsistencies are difficult to control and remediate in any project.  But when left unchecked, they may prolong the amount of time that predictive coding process needs because conflicting coding may “confuse” the system in its own categorization.  In many cases this can be remedied by keeping the review team to a small size and by reconfirming the coding that the system has flagged as inconsistent—steps that may require close and timely monitoring by more senior and expensive attorneys.

"When left unchecked, inconsistent coding may prolong the amount of time that the predictive coding process needs because conflicting coding may ‘confuse’ the system in its own categorization."

Adding Email Threading into the Mix

The problem with inconsistent coding may be significantly mitigated by leveraging another TAR tool—email threading—which limits the review to only the most complete and unique (and usually the latest) threads of an email chain.  For example, if employee A emails employee B with an attachment, and A and B go on to exchange five emails in the same thread, email threading will present only (i) the initial email from A to B with the attachment and (ii) the latest email between A and B in the same thread.

When used in a traditional linear review model, email threading dramatically reduces the number of emails that reviewers have to analyze.  But when used in conjunction with predictive coding, the suppression of largely repetitive content has an added benefit: it may reduce the likelihood that different parts of the chain are coded inconsistently, thereby prolonging the amount of time that may be required to stabilize the predictive coding categorization.

But that’s not all.  The cost savings of the review process may be extended further if the parties agree upfront to produce only the most inclusive threads.  A smaller email production in turn would speed up the preparation of a privilege log, if applicable.

To be sure, there’s a lot to be excited about when it comes to innovation in e-discovery.  But it’s often the utilization of multiple tools that yields the optimal outcome.

About the Author

Caitlin
Kasmar

Partner

Caitlin is a partner at Buckley LLP and the attorney lead for FORTÈ. She has spent most of her career in the e-discovery trenches, advising clients on responding to subpoenas and discovery requests in litigation and enforcement matters.

View Bio

Each day, the average office worker receives 121 emails and sends out 40. Source: Lifewire, The Number of Emails Sent Per Day (and 20 Crazy Email Statistics)

Share

Are you ready to love e-discovery?

Introducing FORTÉ, the better way to do e‑discovery.

Let's Talk