Category: Predictive Coding
Proportionality Considerations and Computer-Assisted Review
The concept of proportionality is echoed throughout the Federal Rules of Civil Procedure. Specifically, Rule 1 provides that the Rules should be construed and administered to secure the just, speedy and inexpensive determination of every action and proceeding. Rule 26(b)(2)(B) further provides that a party need not provide ESI that is “not reasonably accessible because of undue burden or cost.” Perhaps the most direct guidance regarding proportionality is Rule 26(b)(2)(C)(iii), which provides that a court “must limit the frequency or extent of discovery otherwise allowed if it determines that the burden or expense outweighs its likely benefit.”
Generally, courts examine several “cost/benefit” factors in making a proportionality determination, including the following:
As specifically applied to computer-assisted review (CAR), some additional proportionality considerations to examine in this cost/benefit analysis include:
In determining whether computer-assisted review is right for a particular case, consider CAR as compared to more traditional alternatives such as linear review or perhaps targeted manual review based on search terms, custodians or date range. To gauge the likely effectiveness of CAR, determine whether your particular data set lends itself to CAR – what is the likely level of richness of the data, what amount of data is likely to be privileged and what is the quality/prevalence of textual content? If there are a large number of non-text-based documents or the documents have poor quality text, CAR is not likely to be an efficient or effective review tool. Finally, ask your vendor whether or how the results of the CAR tool can be statistically validated. This will allow you to defend your results if pressed to by the court or by opposing counsel.
You should also examine whether using (or not using) computer-assisted review would cause an undue burden in terms of time or cost – what is your deadline and how many reviewers would be needed to meet it? A properly trained CAR tool can make it through a large data set relatively quickly, limited primarily by how much time your subject matter experts can devote to reviewing training sets. If you have a short timeframe for completing a review, the larger the data set, the more expensive a linear review of that data will be. Conversely, the per-document cost of running a data set through computer-assisted review typically decreases as the size of the data set increases. Finally, you should consider how likely CAR is to yield unique, probative evidence. If your data set is both large and largely text-based, CAR will likely be a more efficient way to review for responsiveness. However, if your data set is relatively small, or is large but has a substantial amount of non-text content such as numbers or images, CAR is not as likely to yield unique, responsive evidence with probative value for your case.
If you examine all of these factors at the outset of your case and work with your e-discovery vendor to ensure that CAR is the right tool for the job, you will be well-prepared to present your proportionality argument and get your CAR discovery plan approved by the court.