Access Official Technical PDF

Share Summary:LinkedInEmail

Technology-assisted review can help teams focus attention on a large body of information, but the process depends on more than software. A 2018 court declaration in a securities class action describes how predictive coding and attorney-led analysis were used to manage discovery involving hundreds of thousands of pages. Although the setting was legal rather than industrial, the account provides a concrete example of a technology-supported workflow in which tools helped narrow a search while people remained responsible for assessing the material.

Technology-assisted review in a large document workflow

The case concerned allegations about Liquidity Services Inc. and statements regarding the company’s business performance. During discovery, the class representatives reviewed approximately 500,000 pages produced by defendants and third parties. The declaration also reports that the defendants produced about 120,000 pages in response to discovery requests from the other side, while non-parties produced approximately 223,000 pages.

Handling material at this scale required an organized process. The class representatives’ documents were placed in an electronic database, managed on secure servers, and organized using Relativity. Attorneys reviewed and analyzed the production over nearly a year. Their work supported deposition preparation, responses to interrogatories, mediation preparation and the development of case strategy.

The important distinction is that the technology supported the review; it did not replace the reviewers’ role. Attorneys examined documents, considered their relevance and used what they found to inform subsequent legal work. In this account, the workflow combined a digital repository, analytics software and experienced human review.

How predictive coding fit into discovery

The parties agreed to a protocol under which defendants would use predictive coding to identify material responsive to discovery requests. The stated alternative was relying on other collection approaches, such as searching electronically stored information through search terms. Predictive coding was therefore part of a defined discovery process, not a standalone application operating without procedural limits.

In the class representatives’ review, technology-assisted review (TAR) helped narrow the universe of documents requiring attorney attention. After that initial reduction, attorneys used targeted searches and document attributes to locate material for further consideration. These included:

  • Text within documents
  • Authors and recipients
  • Document type, such as email, memorandum or securities filing
  • Dates and Bates numbers

The declaration describes the resulting workflow as a combination of software-assisted focusing and human analysis. The tools helped direct attention toward potentially significant material; attorneys then reviewed documents in context and identified relevant, irrelevant and particularly important items for additional examination.

Why human oversight remained central

Document classification and relevance depend on the questions being asked. In this case, the review team’s task was connected to specific allegations, discovery requests and litigation needs. Attorneys needed to understand not only whether a document matched a search or appeared important, but how it related to the claims, witnesses and evidence.

Review attorneys also met regularly with more senior lawyers, generally on a weekly basis, to discuss significant documents, deposition preparation and case strategy. This provided a channel for escalating findings and aligning review activity with the broader objectives of the case. The declaration thus presents TAR as one component of a managed workflow, supported by professional judgment and ongoing coordination.

For technical decision-makers evaluating any data-intensive process, the account underscores a practical consideration: reducing the volume of material to inspect is not the same as completing the analysis. A workflow still needs clear review objectives, appropriate access to the information, a way to examine selected records and a process for communicating findings to the people making decisions. These are observations about the documented legal process, not claims about performance in industrial applications.

Process oversight and the limits of automation

The use of predictive coding also became a subject of discussion between the parties. They negotiated over the scope and manner of document production, including questions about the predictive-coding process and the timing of productions. At one point, the court instructed the class representatives to submit questions to obtain additional transparency about that process. Further correspondence and discussions followed, and the parties resolved many of the disputed areas.

This history shows that a technology-supported workflow may require oversight beyond the tool itself. In the case described, the parties had to address how the process worked, what information was being produced and whether production was timely. The declaration does not provide details about the predictive-coding model’s configuration or its measured accuracy, so it cannot support conclusions about technical performance or comparative efficiency.

It does, however, make clear that the review involved several connected elements: agreed procedures, a digital document environment, analytics, targeted searching, attorney review and communication among legal teams. Treating any one element as the entire solution would overlook the coordination required to make the workflow useful.

Practical takeaways for data-intensive work

The case offers a grounded example of using analytics to help manage a large information set. Its lessons should be applied within the limits of the source: the work involved litigation discovery, not factory operations or industrial control systems. Still, the workflow highlights general questions that organizations can ask when adopting tools to help process large volumes of data:

  • What specific information needs to be found, and for what decision?
  • How will software-assisted prioritization connect to human review?
  • What procedures will make the workflow understandable to the people overseeing it?
  • How will findings be communicated and used in later stages of the work?

Technology-assisted review in this case was not simply an automated search. It was a structured process that combined predictive coding, document analytics, targeted examination and attorney judgment. The practical implication is that tools can help focus attention, while reliable use still depends on defined objectives, human oversight and transparency about how the process operates. For teams responsible for complex data workflows, that distinction is essential: automation can support analysis, but the surrounding process determines how its outputs are evaluated and applied.

Access Official Technical PDF

About The Author