On September 17, 2014, the U.S. Tax Court issued its first opinion regarding the discovery of electronically stored information (ESI).  In Dynamo Holdings, Ltd. vs. Commissioner, 143 T.C. No. 9 (Sept. 17, 2014), the Tax Court permitted a taxpayer to use “predictive coding” as a first-level review of a large pool of documents.  The court’s opinion is important for taxpayers faced with requests for a substantial amount of ESI, and has the potential to reduce the burden and cost of complying with such requests.

Background

Most documents today originate in electronic form, and taxpayers may have a duty to preserve ESI.  ESI has traditionally been subject to discovery requests, but it was only recently that the judicial system amended its rules to specifically address ESI.  The case law and procedural laws emphasize that e-discovery should be a cooperative and reciprocal process, and that parties should not be subject to undue burden or cost.  Failure to comply with e-discovery obligations can result in sanctions, which vary depending on the type and severity of the violation.

To respond to requests for ESI, taxpayers and their advisors have several tools to assist in extracting and analyzing responsive information.  One such tool is predictive coding.

What Is Predictive Coding?

Predictive coding is a type of computer-assisted review software technology that leverages human expertise and sophisticated computer algorithms to iteratively train a computer to apply coding decisions on a sample set of documents across a universe of documents.  Typically, in a predictive coding workflow, senior team members with a strong understanding of the substantive legal issues in the case review a small set of “seed” documents in order to identify documents that are responsive to a particular request or issue.  These initial decisions are then captured, analyzed and used to teach the computer to apply similar logic-based decisions to conceptually similar documents within the database.  This process is then repeated several times with the human decisions in each seed set informing and refining the computer’s identification of responsive documents.

Dynamo Establishes that Predictive Coding Is an Acceptable Document Review Technique

In Dynamo, the Internal Revenue Service (IRS) determined that certain cash transfers to a related party were actually disguised gifts to the taxpayer’s owners.  The taxpayer argued that the transfers were loans.  During litigation, the IRS sought the production of ESI that was contained on two backup storage tapes.  The taxpayer asked the court to deny the IRS’ request because it was merely a “fishing expedition.”  Alternatively, the taxpayer asked the court to permit it to use predictive coding to reduce the number of documents to be reviewed to determine the relevant and not privileged documents that would be produced to the IRS.  Allegedly, there were up to seven million documents to review to determine the relevant and not privileged documents.

The IRS argued that the taxpayer should be required to produce the entire universe of potentially relevant information (including any privileged documents) because it wanted to review not only the documents themselves but also all of the metadata associated with each document.  The IRS also argued that the court should deny the taxpayer’s request to use predictive coding because it is an “unproven technology.”  However, the IRS assured the court that, to the extent that the set of documents produced contains privileged documents, the IRS would agree that the disclosure of those privileged documents would not constitute waiver of any privilege claims that the taxpayer may assert.

First, the court reviewed its discovery rules and determined that ESI is discoverable in Tax Court.  Second, the court weighed the parties’ competing interests.  One the one hand, the IRS is entitled to a complete response to its discovery request.  On the other hand, the taxpayer has the right to protect its privileged information and avoid the substantial costs associated with a so-called “linear review,” a review of each putatively responsive document by a human reviewer.

The court determined that predictive coding is a “potential happy medium” that addresses each side’s concerns.  In making its decision, the court recited articles and studies that indicated that, in general, predictive coding is more accurate than a traditional, linear review because humans tend to make more mistakes than computers.  The court also found support for the use of computer-assisted review by other courts facing similar issues.

The court’s opinion is a welcome development for taxpayers in the current electronic age, particularly given the limited number of tax cases addressing e-discovery.  The IRS has broad authority to request documents from taxpayers, and the time and cost associated with such requests can skyrocket when large amounts of ESI are involved.  Tools such as predictive coding can make the discovery process more efficient for all involved.

Predictive Coding Makes Large Document Review Projects More Efficient

The advent of email and electronic media has made discovery during litigation an exponentially complicated and expensive ordeal for litigants.  Predictive coding may answer the question of how to be efficient and reasonable when you have large amounts of electronic data.  Predictive coding offers several benefits.  First, it offers a proven, viable method to assist courts with their duty to limit discovery within reasonable bounds.  The Federal Rules of Civil Procedure state that a “court must limit the frequency or extent of discovery” if the discovery proposed is “unreasonably cumulative or duplicative” or if the information can be obtained from “a more convenient, less burdensome, or less expensive” alternative source.  Fed. R. Civ. P. 26(b)(2)(C).  Tax Court Rule 70(c)(2) contains a similar “undue burden or cost” exception.  Also, predictive coding can substantially improve rates of precision and accuracy.  The computer’s concept identification, multi-dimensional “thinking” and continuous fine-tuning can be better than simply using key words to find similar documents and identify true responsiveness.  Finally, leveraging an advanced review technology, such as predictive coding, is a smart way to proactively manage the often massive scope and cost of discovery, resulting in substantial savings to clients.

Computer-assisted Document Review Is Here to Stay

Being facile with predictive coding and other ESI review tools will become increasingly important as more courts sanction their use.  Cases in which one side has significant ESI and the other does not will continue to present hurdles for the use of sophisticated ESI review tools.  But in complex cases, where both sides are sophisticated litigants and have large amounts of ESI, we expect more and more cases to involve party-negotiated and court-approved ESI-review terms.  Indeed, ESI discovery in complex cases will become increasingly impractical in the absence of the advantages created by recently developed ESI review tools.  In that sense, the trend in favor of using those tools is inevitable.

How Best to Prepare for ESI Discovery

There are several things that clients can do to take advantage of the trend.  First, clients should take stock of their ESI.  The more that a client knows about its data (e.g., the quantity of data, how the data are stored and backed up, retention policies), the easier it is to identify the tools that can efficiently analyze the data.  Part of this process may require updating document retention policies—given that individuals are more likely to retain electronic data than paper documents—to ensure that unnecessary ESI can be disposed of prior to time any duty to preserve may apply.  Second, clients should delegate internal controls for ESI to a single, responsible individual who can interface with outside counsel.  In the authors’ experience, clients that have a designated ESI officer understand the ESI issues and options more deeply than those that do not.  Finally, clients should consider investing in a system that would make iterative ESI collection easier, faster and more efficient.  This may be especially important with clients that are routinely involved in litigation.  ESI collections can be very expensive and time consuming, but developing a system in which subsequent collections can build off earlier ones may reduce per-collection costs and improve efficiencies.