The Little-Known World of TAR
Sierra Palmer, Law Clerk
Schwartz Semerdjian Cauley & Moot LLP
Discovery document review is one of the most time-consuming and sometimes discouraging tasks in the litigation process. It costs time and money, and if done haphazardly, can end up resulting in sanctions. But what if document review were as simple as finding a good show to watch on your Netflix queue? Several predictive coding experts say it very well could be.
When you hit the couch with a bowl of popcorn to watch your favorite Netflix series at the end of the day, you scroll through the movie options, pick your favorite Rom-Com, and enjoy. Each time you do this, Netflix learns more about what types of movies you are looking for and is better able to give you recommendations each time you log on. Technology-Assisted Document Review (TAR) uses the same predictive coding technology to operate document review in a similar fashion, only it helps organize legal documents instead of Rom-Coms.
The Way TAR Works
The TAR process begins when an attorney reviews a “sample” document and tags it as either relevant or non-relevant—essentially training the system on what it is that he is looking for. The TAR system then takes that content information and uses it to draw inferences about the other documents up for review, subsequently organizing each document by relevance to help guide the review process. The TAR system bases its relevance on substantive concepts of the case instead of strictly chronological details such as times, dates, and names. For example, if a litigator is dealing with document review involving a wrongful termination case, the TAR system can hone in on tangible concepts such as hiring details or discrimination language. The review system can also detect information indicating privilege and confidentiality in each document.
When copious amounts of discovery documents are given in a litigation case, TAR allows the attorney to remove up to 95% of unresponsive documents from review, enabling a more narrowed and focused evaluation of the documents that really matter, doing so in less time for less money. (See TAR for Smart People: How Technology Assisted Review Works and Why it Matters for Legal Professionals, Catalyst Secure (2015), https://catalystsecure.com/pdfs/book/ Catalyst_TAR_for_Smart_People.pdf (last visited Aug 9, 2017)).
Is it Judicially Acceptable?
The fear of discarding up to 95% of submitted discovery documents per the suggestion of a computer, in addition to the seemingly-complex technology components of this program, have discouraged numerous firms from taking the leap into predictive coding technologies. However, the TAR system got the court’s official stamp of approval despite Plaintiff’s objections in Da Silva Moore v. Publicis Groupe, 287 F.R.D. 182 (S.D.N.Y. 2012) when it held that computer-assisted review can now be considered judicially-permitted for use in appropriate cases. Id. at 193.
In Da Silva Moore, a gender discrimination case, the parties agreed on certain ESI sources to condense three million electronic documents from the agreed-upon custodians. Id. at 184. After several conferences with opposing counsel and the judge, Plaintiffs’ vendor noted that they did not think the predictive coding agreement would work because it is “new technology and it has to be proven out.” Id. at 187. Plaintiffs subsequently objected to the judge’s ruling asserting his acceptance of defendant’s predictive coding approach on the grounds that it violated Fed. R. Civ. P. 26(g)(1)(A). This rule states that by signing a disclosure, a party certifies that it is complete and correct at the time it is made to the best of the person’s knowledge. Id. The court ruled that Fed. R. Civ. P. 26(g)(1)(A) is not in contention with TAR because that rule has nothing to do with a defendant’s obligation to respond to a plaintiff’s discovery requests. Id. Fed. R. Civ. P. 26(g)(1)(B) was written separately to address discovery responses, and it does not require the same certifications required in Fed. R. Civ. P. 26(g)(1)(A). Instead, it requires that discovery requests be consistent with the rules and warranted by existing law or by a non-frivolous argument. The court noted that while computer-assisted review is not perfect, the Federal Rules of Civil Procedure do not require perfection, and those rules are in place partly to secure the speedy and inexpensive determination of lawsuits. Id. at 191. The court concluded that although computer-assisted review may not be appropriate for all cases, it is a useful tool that should be considered for use in cases with a large data volume where it may save parties significant legal fees in discovery document review. Id. at 194.
The court in Rio Tinto PLC v. Vale S.A., 306 F.R.D. 125 (S.D.N.Y. 2015) confirmed that, in the three years since Da Silva Moore was decided, “case law has developed to the point that it is now black letter law that where the producing party wants to utilize TAR for document review, courts will permit it.” Id. at 127. In Rio Tinto PLC, the parties agreed to abide by protocol that discloses all non-privileged documents in their control sets. Id. at 129. The court opinion noted that, because of this, the court need not rule on how transparent and cooperative the parties must be with respect to their training sets, which is an eDiscovery issue that remains open. Id. at 128. Rather, the court wrote the opinion because of the eDiscovery community’s interest in TAR protocol. Id. The opinion stressed that it is inappropriate to hold TAR to a higher standard than a manual or keyword review because doing so would discourage parties from using TAR. Id.
When to Use TAR
Although there is no magic number of documents that permits or discourages the use of a TAR system, it is recommended for cases involving voluminous discovery. Many TAR vendors implement an equation to estimate how much money a litigator can save or lose using TAR by analyzing several factors: how many documents are up for review, the hourly billing rate of the document reviewer, the estimated target recall, the documents per hour reviewing rate, and other pricing dynamics. (See Predictive Ranking ROI Calculator, ROI Calculator, https://catalystsecure.com/products/insight-predict/roi-calculator (last visited Aug 9, 2017)). Some vendors charge by the amount of data in the case while others charge for the number of users, but pricing varies with each review system and vendor.
To give an example of a document review estimate, one vendor website projects that if an attorney has 100,000 documents up for review, an hourly billing rate of $50 for the first pass reviewer and $250 for the second pass reviewer, an estimated target recall of 80%, and a 50 documents per hour review rate, the final costs for traditional review would be $100,000 while the vendor’s final charge for the TAR system would be $16,847—saving the client $83,153 in discovery costs. The projection also notes that document review hours required for traditional review would amount to approximately 1,680 hours, while TAR would take about 246 hours. (See Predictive Ranking ROI Calculator, ROI Calculator, https://catalystsecure.com/ products/insight-predict/roi-calculator (last visited Aug 9, 2017)). However, the site estimates that with the same billing rates and target recall, the attorney would lose money if there were 2,000 or less documents to be reviewed. The document review hours for 2,000 documents under traditional review are still far greater at 34 hours compared to TAR’s 13-hour estimation. Regardless of cost, TAR is consistently less time-intensive than traditional review, so attorneys who are in a rush may choose to spend the extra money under particular time restrains.
What the Future Holds
As of 2016, predictive coding technologies were used in only 1% of cases, yet document review consists of approximately 73% of the discovery process. (See Highlights from the Northeast eDiscovery &IG Retreat 2016: Predictive Coding and Other Document Review Technologies–Where Are We Now, Clustify Blog, https://blog.cluster-text.com/ (last visited Aug 9, 2017)). Because the TAR process is still fairly foreign to the legal community, traditional document review methods remain the industry standard. However, predictive coding technologies are becoming more familiar to the public at large by way of everyday use, including Netflix queues, Pandora playlists, and Amazon product recommendations. As a new generation of attorneys familiarize themselves with this increasingly ubiquitous technology, TAR will likely transform into a tool that is more than an impressive but irresolute concept.
Judge Peck, who was the sitting judge for both Da Silva Moore and Rio Tinto PLC, predicts that there may come a time when it may be unreasonable for a party to decline to utilize TAR because it will be so widely used. (See Should You be Using TAR? Judge Peck Recommends You Do, D4Discovery (2017), http://d4discovery.com/discover-more/judge-peck-recommends-using-tar#sthash.vxjOPvTs.dpbs (last visited Aug 9, 2017). Although he admits that the legal world is not at that point yet, it seems safe to say that TAR is here to stay and is likely to be used more commonly in the future, so it may be time to grab your popcorn and let TAR queue up your document review.