Leveraging eDiscovery for Fact Development: Practice Pointers

In this article for New York Law Journal, lawyer Alisa Melekhina discusses best practice advice for making the most of a litigation data set and provides ideas to assist in leveraging litigation review workflows and positioning review teams for efficient, multi-purpose responsive and fact development reviews.

While responsive and privilege reviews are at the core of eDiscovery, these types of reviews also offer the opportunity to simultaneously pave the way towards establishing key facts relevant to the claims and defenses at issue. Federal courts often opine that the relevancy requirements of Federal Rule of Civil Procedure 26 should be read broadly. See Fed. R. Civ. P. 26(b)(1) (permitting discovery “regarding any nonprivileged matter that is relevant to any party’s claim or defense and proportional to the needs of the case”); Gilead Scis., Inc. v. Khaim, No. 24-CV-4259, 2024 WL 5192075, at *3 (E.D.N.Y. Dec. 22, 2024) citing New Falls Corp. v. Soni, No. 16-CV-6805 (ADS) (AKT), 2020 WL 2836787, at *2 (E.D.N.Y. May 29, 2020) (“Relevance is a matter of degree, and the standard is applied more liberally in discovery than it is at trial.”).

Accordingly, in light of these rulings, it is not uncommon for litigants to experience voluminous outgoing and incoming productions in complex cases. Below are ideas to assist in leveraging litigation review workflows and positioning review teams for efficient, multi-purpose responsive and fact development reviews.

Defensive Review
Defensive review protocols often contain complex issue and key document criteria. This can be a benefit to the fact development process as it offers a head start in flagging important themes and key document escalation instructions for reviewers. Highlighting the substantive guidance already built into review protocols can create opportunities for re-using coding and key document work product.

Streamlining Issue and Key/Hot Tags. Consolidating issue coding selection “tags” to correspond to Requests for Production (RFPs) can be the first inflection point at which to establish the themes of the case. If responding to numerous RFPs, creating one issue tag per RFP may become unworkable and clutter the review pane, and likely would slow down the speed of the review team.

• Instead, this can be used as an opportunity to think through: which tags can be consolidated? For example, do multiple RFPs touch on the same underlying conduct, meeting, or financial or economic event? Unless there is a situation where it is necessary to break down productions by RFP response, these types of overlapping themes can be used to establish a broader, single categorical issue tag.
• Note that issue tags can be used in conjunction with Key or Hot document “escalation” tags. In other words, if there is a concern that consolidation may miss particular documents, key documents could still be queried within each issue category by using tagging combinations.

Examples, Examples, Examples. When training reviewers, a strongly written review protocol can be further bolstered by providing document examples pre-selected by the subject matter experts. These examples may demonstrate what would fall under various key, issue, PII, or privilege tags, or even what is, in fact, non-responsive. Early concrete document exemplars may prove helpful in aligning early on complex responsiveness and privilege determinations.

• These core or example documents can often be identified with targeted searches at the outset, if not already available from pleadings or prior fact development.

Tracking and Escalation. Establishing a workflow early on that makes clear how key documents will be escalated, and to whom, is likely to aid in early fact development efforts. One way to accomplish that is as part of daily tracking metrics compiled by the eDiscovery vendor or your litigation technology support team.

Examples of metrics to include are the day’s document determinations as well as links to saved searches for important issue and/or Key tags. Another option is to export a daily excel of key documents that includes (i) the basic metadata, such as date, to, from, cc, bcc, subject, and filename; and (ii) any comments included by the reviewing attorney noting why they determined such documents warranted key treatment.

• For this suggestion to work, instructions must be included in the review protocol for how key documents and issue tags should be escalated, and any instructions for how and where key documents should be summarized (such as a Comments field). Coding is merely the first step—and one that has the potential to be fruitless without further having a plan for leveraging the tagged key documents.

Offensive Review
“Offensive” fact development reviews draw on the same set of processes and skills as traditional eDiscovery reviews. Working with incoming or already produced documents may present opportunities to incorporate more creativity with searches and technology tools. In contrast to defensive reviews, offensive reviews do not carry the same risks of inadvertently producing documents.

Planning Ahead. As with deadlines for outgoing productions, incoming productions require planning ahead. Fortunately, the rolling and substantial completion deadlines for all parties are often dictated by the common case schedule.

• When formulating a document review staffing plan, it may help to have prompt and early discussions with the team regarding the offensive review plan. Who will perform the review—contractors, associates, Technology-Assisted Review, and/or Generative AI?
• Each case has different staffing and budgeting considerations. What makes sense for defensive privilege review may not make sense for building out highly substantive fact modules. Having early discussions will help to tackle incoming productions in real-time.

Keyword Searches and/or CAL. Depending on the volume of incoming productions, linear review may not always be feasible. While keyword searches may serve as a helpful starting point, consider how to go beyond searches previously tailored to RFP responses. Crafting targeted searches (with limiters) can be achieved through tracking key terms and phrases from early identified key documents. These initial rounds of searches will often lead to other buzz words and related documents.

• As a backstop, or even as a stand-alone, consider enabling Computer-Active Learning (CAL) training models in the background. This offers the benefit of escalating high-scoring (likely to be relevant or important) documents that may fall outside the search terms.
• While Generative AI is outside the scope of this article, a use-case could exist to explore—cost-permitting—already produced data such as incoming productions. Consider whether AI-identification and generated summaries of key documents over a voluminous data set could be a value-add.

Identifying High-Performing Reviewers. The defensive review could serve as an opportunity to identify knowledgeable reviewers who could quickly pivot to offensive fact development review. Other than relying on overturn metrics to assess performance, consider incorporating qualitative assessments such as check-in calls or rolling key document summaries. Another criterion is how the reviewers incorporate feedback on key document determinations going forward. Reviewers can be an integral part of the team and may have amassed institutional knowledge that can be leveraged as fact development progresses.

Setting Work Product Expectations. While the work product for responsive and privilege reviews typically focuses on coding determinations, fact development reviews emphasize a more substantive output. Consider how to best align fact work product with the case team’s needs. For example, would key documents be best communicated via saved searches with comments, ongoing chronology updates, or document summaries by email?

• Consider assigning a dedicated team member, such as a subject-matter expert associate, to aid in this important work stream. As with all eDiscovery workflows, managing fact development comes down to open communication, feedback loops, and setting expectations upfront.