Litigation Support

 View Only

TAR Protocols: A Practical Guide - Second Blog

By Joel Sandler posted 01-27-2022 15:22

Please enjoy this blog post co-authored by Joel Sandler, eData Practice Support Manager, Morgan Lewis & Bockius LLP, Ethan Ackerman, Senior Attorney, Morgan, Lewis & Bockius LLP and Leonard Impagliazzo, Associate, Morgan Lewis & Bockius LLP.


As we explored in the recent “Crash Course in TAR”, Technology Assisted Review (“TAR”) refers primarily to the use of machine learning to predict the relevancy of documents in a discovery project.  The technology continues to evolve; we have progressed from the static models of TAR 1.0 to the active learning models of TAR 2.0 and are witnessing further innovation with workflows described as TAR 3.0.

As the technology evolves, one question remains consistent:  does the traditional use of TAR require some form of agreement between adverse parties in discovery?  Often marred in disagreement over technical specifications, negotiated use of TAR can fall short of its promise to deliver cost-effective results. However, TAR does not have to be limited to only those instances where parties agree to its use. Legal support technologists can work with attorneys to identifies opportunities to leverage TAR without formal discovery agreements.

Traditional ‘Negotiated’ Application of TAR

In a traditional negotiated application of TAR, parties in discovery come to an agreement on whether the technology may be used at all to identify documents for production. Given the impact that TAR will have on the outcome of the document production process, the parties typically agree on various inputs, outputs and settings of the TAR workflow. These agreements may require that exemplar documents used to train the model are identified (aka “seed sets”), that samples of predicted non-relevant documents are validated, or that specifics of the underlying technology and selected workflow are divulged.

Once the parties have come to an agreement, attorneys for the producing party work in tandem with their legal support colleagues to prepare the TAR models.  In most scenarios, the attorneys manage the review side of the project, while the support team administers the technical aspects of the platform.  Active collaboration between the legal and technical roles is critical throughout this process, as there are many decisions to be made and milestones to document.

Once the model is established, the TAR software will assign a “score” to each document in the review set and eventually the non-privileged documents above a certain threshold score will become eligible for production.  The TAR agreement may require the producing party to divulge the process by which that “cutoff score” is determined.

In asymmetric discovery, such as a regulatory demand for documents or a consumer-facing class action, the producing party may have little leverage to negotiate a TAR protocol that is favorable. The attorney may end up using whatever bargaining power is at their disposal dealing with TAR logistics rather than the merits of their case. From a strategic perspective, they may simply not want to fight that battle.

Even in a more evenly-balanced discovery context, such as a corporate IP litigation, the parties may have reason to avoid a detailed and invasive TAR agreement. For example, the parties may not want to divulge their exemplar seed sets, as doing so may shed light on what they perceive to be their key documents. For all of these reasons, parties may not ultimately come to agreement on TAR protocols even though they have access to the technology and its myriad benefits.


Alternative Application of TAR

Absent agreement between parties on a TAR protocol for production, there are still many ways to leverage the technology. A defining feature of the traditional approach described above is that the technology is being used to assist with a threshold determination of what documents to produce and which to withhold.    Key features of the alternative workflows discussed below is that they can be used for initial investigatory purposes, to prioritize internal review tracks and to aid in quality control processes. 

Attorneys working side-by-side with legal support technologists can establish innovative workflows, especially leveraging TAR 2.0 platforms. Free from constraints of a fixed agreement, these are scenarios where one can leverage TAR 2.0’s flexibility, in particular that scoring models can be spun up with much less overhead compared to TAR 1.0.  It is critical that the attorney clearly articulate the goals of the use of TAR, while the legal support technologist asks the appropriate questions to accomplish the identified goals.

One useful application is leveraging TAR to assist with the arduous task of identifying privileged content within an email review.  Once the legal team has identified a validated set of confirmed privilege designations, they can work with their support team to seed a TAR model. With minimal additional inputs, the TAR software can be used to assign predictive scores to the remaining documents. That score can be used as a proxy for likelihood of privilege content in those documents. The high-scoring documents that have not yet been designated as privilege can be identified for further review.

Taking this a step further, the entire thread of any high-scoring documents can be evaluated to ensure privilege consistency throughout the email thread. This workflow can help to avoid inadvertent privilege disclosures, as well as embarrassing over-assertions of privilege. While TAR scoring should not be used as the only quality-control measure for identifying privilege documents, it can be a great addition to the overall privilege review protocol.


TAR can often be leveraged to significant impact on any matter, even when not utilized in the ‘traditional’ workflow of selecting documents for production. No matter the desired usage of TAR, one factor remains true; continuous collaboration and clear communication between attorneys and legal support technologists is key to maximizing the benefits of any form of TAR. Working together, the teams can use new technology in innovative and exciting ways.