In a September 2021 blog post, Holly Vedova, Director of the Federal Trade Commission’s (FTC) Bureau of Competition, outlined several “new process reforms” for Second Requests, designed to result in both a “more streamlined and more rigorous” approach.1 The FTC appeared to be especially focused on revamping its eDiscovery procedures for Second Requests with the reforms addressing topics such as the identification of custodians, companies’ maintenance of IT systems and potentially responsive data, the approval and use of eDiscovery technologies, and privilege log formats.
Perhaps the most significant change was the development of stricter requirements and new processes for the approval and use of eDiscovery tools and methodologies:
…the FTC’s second requests will now require each company under investigation to provide information about how it intends to use e-discovery tools before it applies those tools to identify responsive materials. Complete and accurate information is critical in any investigation and there are substantial benefits to ensuring up front that e-discovery processes will identify required information. In addition, this change will more closely align the FTC’s Model Second Request with that of the Department of Justice.2
In practice, the FTC reforms have and will impact three areas of eDiscovery practice on Second Requests issued by the FTC: (i) general disclosure requirements, (ii) search terms, and (iii) Technology Assisted Review (TAR). These reforms are of particular note because they rectify the prior divergence in approach of the FTC and Department of Justice (DOJ). Such reforms are noticeable, in part, in the FTC’s updated Model Second Request.
General disclosure requirements
Specification 29 of the prior FTC Model Second Request governed parties’ disclosure obligations relating to eDiscovery, requiring them to “[i]dentify any electronic production tools or software packages utilized by the Company in responding to this Request for: keyword searching, Technology Assisted Review [(TAR)], email threading, de-duplication, and global de-duplication or near-de-duplication…”3 While there were additional requirements as it relates to the use of TAR, this Specification simply required the identification of the technology used, not a description of how the technology was employed. Further, parties were able to submit their response to Specification 29 after completing productions , allowing them the flexibility to get started quickly without having their use of TAR or other eDiscovery technologies pre-approved by the FTC.
New language in Instruction I(5) of the updated FTC Model Second Request, which is identical to Instruction 4 of the DOJ Model Second Request, includes new mandates, requiring that “before using software or technology (including search terms, predictive coding, deduplication, or similar technologies)” parties must “submit a written description of the method(s) used to conduct any part of its search.”4 Notably, parties before the DOJ must provide this information prior to use of the eDiscovery technology.
In practice, this reform is a departure in form, i.e., timing, rather than substance, i.e., what must be disclosed, from past FTC practice. Appropriate planning and preparation will permit parties to minimize disruption or delays created by this upfront requirement. The disclosure of similar information is often required in other matters, and most eDiscovery practitioners will have standardized workflows for the applicable technologies with written descriptions of such procedures they use. Further, as described more fully below, parties utilizing TAR were already required under Specification 29 of the previous FTC Model Second Request to provide similar descriptive information.
The requirement to provide this information at the outset of the process, however, may prove to be more burdensome if it indicates the FTC’s intention to heavily scrutinize the technologies and methodologies used by parties prior to granting approval or allowing the waiting period to expire, “to ensur[e] up front that e-discovery processes will identify required information.”5 If so, then parties may consider building in additional time for these discussions and consider proactively engaging with the FTC to avoid any delays. Where parties anticipate a Second Request, it may be advantageous to conduct eDiscovery planning sessions and begin drafting the required written description during the initial waiting period, allowing them to move quickly once the Second Request is issued and avoid any bottlenecks caused by the eDiscovery process.
Search terms
The DOJ and FTC Model Second Requests both specifically address the use of search terms as a method of reducing data sets and identifying potentially responsive information. The FTC’s previous requirements were more general than those of the DOJ, requiring parties using search terms to simply “provide a list of the search terms used for each custodian.”6
Instructions in the updated FTC Model Second Request (again, identical to those in the DOJ’s version), call for parties to provide more in-depth information regarding their use of search terms, including “(a) a list of proposed terms; (b) a tally of all the terms that appear in the collection and the frequency of each term; (c) a list of stop words and operators for the platform being used; and (d) a glossary of industry and company terminology.”7
While most of these additional items are fairly straightforward, responding to item (b) (the tally of all terms in the collected data and frequency of each term) can take more time to complete than other requirements and may require custom solutions depending on the eDiscovery processing engine and search index used. If employing search terms on FTC Second Requests, outside counsel and eDiscovery providers should take steps in advance to determine how best to provide this information and whether doing so may require extended lead time or bespoke solutions.
As discussed in more detail below, the most significant difference to date in the use of search terms before the DOJ and the FTC has been the ability to apply search terms prior to the application of TAR, which the DOJ does not permit. It is still unclear whether the FTC will formally adopt a similar approach. Parties in civil litigation and past matters before the FTC have used the combination of well-crafted search terms and TAR to further reduce cost and save time with no reduction in recall metrics (i.e., the percentage of responsive documents from a document collection that are successfully identified and produced). The FTC could very well adopt a middle-ground position, continuing to allow search terms to be used in conjunction with TAR but requiring that the data population excluded through the use of search terms be subject to a sampling and validation process similar to that used for documents excluded by TAR. To date, there is no official FTC position on the topic, which may indicate that a party’s approach will be evaluated on a case-by-case basis.
Technology Assisted Review
The updated FTC Model Second Request addresses guidelines for TAR workflows in two separate areas. First, Specification 30, maintained verbatim from previous versions, requires parties to:
Describe the collection methodology, including (a) how the software was utilized to identify responsive documents; (b) the process the company utilized to identify and validate the seed set documents subject to manual review; (c) the total number of documents reviewed manually; (d) the total number of documents determined nonresponsive without manual review; (e) the process the company used to determine and validate the accuracy of the automatic determinations of responsiveness and nonresponsiveness; (f) how the company handled exceptions (“uncategorized documents”); and (g) if the company’s documents include foreign language documents, whether reviewed manually or by some technology-assisted method; and
Provide all statistical analyses utilized or generated by the company or its agents related to the precision, recall, accuracy, validation, or quality of its document production in response to this request.8
Parties often maintain this information as a standard practice and such information is typically accessible with little disruption or delay to day-to-day eDiscovery processes. This is especially true without an explicit requirement to provide the information before implementation of TAR. Parties have typically been able to meet these requirements by preparing a memorandum outlining the requested data, often completed close to the conclusion of the eDiscovery process.
Second, new language in Instruction I(5) (again mirroring language in the DOJ Model Second Request) adds the following regarding the use of TAR:
For any process that relies on a form of Technology Assisted Review to identify or eliminate documents, the Company must submit (a) confirmation that subject-matter experts will be reviewing the seed set and training rounds; (b) recall, precision, and confidence-level statistics (or an equivalent); and (c) a validation process that allows Commission representatives to review statistically-significant samples of documents categorized as non-responsive documents by the algorithm.9
These new instructions are functionally the same as those in Specification 30, with the exception that parties now also need to (i) provide confirmation that “experts” will complete the training rounds and (ii) incorporate a validation process that includes agency review of samples of predicted non-responsive documents.
Notably, the DOJ maintains a Predictive Coding Model Agreement separate and apart from its Model Second Request, which includes a wide range of additional standards and guidelines to which parties must agree when using TAR.10 While the FTC has updated the written TAR guidelines in its Model Second Request to match those in the DOJ’s Model Second Request, the FTC has not released a Predictive Coding Model Agreement. As such, it is unclear if the FTC intends to also adopt any of the more substantial requirements in the DOJ’s Predictive Coding Model Agreement.
The FTC’s decision in this area will likely have the greatest impact on the various changes discussed here, given the increasing utilization of TAR on large-scale, time-sensitive matters and the fact that the requirements in the DOJ’s Predictive Coding Model Agreement are considered burdensome by some.11 According to the DOJ’s Predictive Coding Model Agreement, parties must adhere to the following, at least when employing traditional TAR workflows:
- No pre-culling of data sets: parties cannot use search terms or other analytical tools (such as email threading) to reduce data sets prior to the application of TAR
- No supplemental human review for responsiveness: after documents are identified as responsive by the TAR tool, they cannot be excluded from production via either manual review as nonresponsive, or search methods, without written approval from the DOJ.12
These additional requirements have, in our experience, created hesitancy among some parties to use TAR before the DOJ, with parties citing the following concerns most commonly:
- Disallowing any documents classified by the algorithm as “Responsive” from being manually re-reviewed for responsiveness is unreasonable. While human review is certainly not perfect, neither is the algorithm. Documents incorrectly predicted by the system as “Responsive” may contain sensitive information, which parties neither want to produce nor would be obligated to produce absent use of TAR as, in-fact, non-responsive. In addition, without the flexibility to update responsiveness decisions, parties may need to invest additional time in correctly identifying privileged information in otherwise non-responsive documents (as opposed to simply excluding from production). Accurately identifying this privileged content, completing necessary redactions, and compiling privilege log entries are all tedious and time-consuming steps that further increase cost and delay compliance.
- Restricting the use of search terms and email threading before, or after, utilizing TAR negates some of the potential efficiency gains of these tools. eDiscovery practitioners frequently use search terms or email threading alongside TAR to reduce data sets in large-scale matters, and believe this combination to be more accurate and efficient than either approach on its own.13 Even when using TAR to make predictive decisions regarding responsiveness, most parties will still opt to review a portion of these documents for privilege, confidentiality, or key issue identification prior to production. The larger the universe of documents, the larger the scope of this manual review—and the cost and time required to complete it before production. That may reduce the benefits of TAR, especially given the other concerns noted above.
Further, negotiating and reaching agreement on TAR protocols in advance of starting the eDiscovery process will often delay compliance with the Second Request. Parties are often eager to begin as quickly as possible with the understanding that slight modifications to their eDiscovery protocols may be agreed to after starting. Having to settle on these protocols and standards in advance, even if this process is typically streamlined, can lead to lost time and add to parties’ reluctance to leverage TAR. The FTC’s statement that it intends to require parties to provide additional information before they apply certain eDiscovery tools may indicate the potential for similar delays in situations where there is disagreement or questions regarding a party’s proposed approach.
Notably, both agencies’ Model Second Requests and the DOJ’s Predictive Coding Model Agreement appear to contemplate what is known as a TAR 1.0 model, in which human review of an initial “seed set” and training rounds of documents are used to inform the system, after which it then makes predictive determinations for the overall data set. More recently, many practitioners have shifted to technologies that incorporate an active learning framework (TAR 2.0), in which the TAR algorithm continually updates its relevance scoring based on ongoing human review of documents prioritized for responsiveness by the tool. The DOJ has not publicly addressed the use of these updated workflows. Anecdotally, prior to revisions to their Model Second Request, the FTC had been receptive to both TAR 1.0 and 2.0 workflows, despite written guidance from both agencies appearing to contemplate only 1.0 models. Neither agency has formally updated their protocols to specifically accommodate these approaches.
Impact on future Second Requests
The FTC’s changes to its specifications regarding the approval and use of eDiscovery tools and/or processes should have relatively little impact on general disclosure requirements (i.e., the approval of eDiscovery tools or proposed processes) or the application of search terms (in non-TAR contexts). These updates do not represent significant changes to past FTC guidelines and much of the newly required information should be easily accessible or routinely maintained by eDiscovery providers. Similarly, the recent changes made to the TAR protocols in the updated FTC Model Second Request, while imposing some additional requirements on parties, should not prove overly burdensome as written.
However, if the FTC also imposes requirements similar to those in the current DOJ Predictive Coding Model Agreement, this may significantly impact the use of TAR on Second Requests. As it stands now, it appears somewhat more common for parties before the DOJ to forego the use of TAR, finding that the potential delays and additional burdens associated with the agency’s written guidelines present new risks and reduce the perceived benefits of employing otherwise useful technologies. On the other hand, parties had generally found the FTC’s guidelines (as written and in practice) to be flexible and user-friendly. If the FTC expects parties to follow some of the DOJ’s more rigid requirements for TAR, parties may implement new eDiscovery strategies on future Second Requests in response. In any event, parties before the FTC should expect their processes to come under heightened scrutiny regardless of the technology or process used.
