Originally posted on August 27, 2012
As I read some of the commentary I have seen posted on Kleen Products, I felt compelled to weigh in as an industry participant closely involved with predictive coding. Kleen Products was a useful case for a number of reasons.
First the case showed us that a party which educated themselves on information retrieval was willing to go into court and argue that parties should have to use technology to improve the search and retrieval of responsive ESI. Having witnessed some of the hearings first hand, I believe the plaintiffs were genuine in their desire to have the defendants move beyond keyword searching because they were and probably still are deeply concerned that they will have a hard time finding what they need from key word and Boolean searches supplemented with some concept searches. The reality is it was a tremendous burden faced by the Kleen Products plaintiffs to argue that the defendants in Kleen Products should have done something different from what lawyers have traditionally done with key words, especially when these defendants had already reviewed millions of documents after key word searching.
This task of proposing a better approach was made even more difficult by the plaintiffs arguing the approach in the abstract. If predictive coding was a synonym, they might have had more success. But just like a key word usually has lots of meanings, predictive coding means different things to different people. For a good example of this, listen to the ESIBytes Predictive Coding Power User podcast I moderated at the Carmel Valley E-Discovery Retreat with Maura Grossman, Tom Gricks, and Bennett Borden as attorneys who each use technology assisted review with different tools and approaches with Herb Roitblat and David Lewis as technology experts.
Maura Grossman vigorously advocated for seeding predictive coding with seed sets of responsive documents while Tom Gricks advocated his experiences accomplishing the same task with random sampling of a collection. Bennett Borden chimed in that he used analytics and agile review teams to accomplish similar tasks. As a result, predictive coding is a GPS system and there are different ways to attack your data and utilize forms of technology assisted review and they mean different things to different people.
To contrast the Kleen Products situation, in Global Aerospace, Inc v. Landow Aviation, L.P. there was a sense of the workflow being proposed to the other side with the ability to describe the predictive coding solution as well as the advantage that no work had previously been done with an opposing strategy. Personally, even with years working with predictive coding tools, I could never get a good feel from Kleen Products transcripts and testimony what the plaintiffs were exactly asking for except they wanted to use more predictive coding technology and less key word driven solutions.
Secondly, Kleen Products most important legacy might be it provides some good training for future litigants who are on the receiving end of productions to identify better approaches to challenge key word searching. To those who believe this case shows that key word searching is a position that is easy to defend, I would recommend putting down that celebratory beer and sober up. My prediction is you should be prepared for additional challenges post Kleen Products. This is due in no small part by the effort of the Kleen Products plaintiffs and their experts who invested time and resources into starting this debate and raising the public awareness of their misgivings on key words searches. Their arguments were based on scientific principles and can be evolved into better arguments which do not face the same tremendous burdens I outlined above.
The next challenge raised by a receiving party will be far more sophisticated and harder to defend. Especially if the opposing party says go ahead and use key words but due to our reluctance in accepting them, we are insisting on sampling the null set and we reserve the right to request additional review if the sampling shows the key word searches were not effective.
For those who want some precedence on the use of predictive coding, Da Silva Moore and Global Aerospace provide examples of cases where predictive coding is being used and Judge Scheindlin’s recent dicta in Nat’l Day Laborer Org. Network v. United States Immigration & Customs Enforcement Agency, 2012 U.S. Dist. LEXIS 97863 (S.D.N.Y. July 13, 2012) highlights judicial recognition that predictive coding can offer improved results than key word searching alone. Kleen Products is an important piece of the predictive coding puzzle because it can only result in better arguments against key word searching being used by future litigants. Any entrepreneur will tell you, the best business lessons come out of making mistakes which turn into better future ideas. My prediction is Kleen Products biggest legacy will be in planting the seeds for the next argument raised by a party receiving ESI from key word searching to challenge key word searching.
Karl Schieneman, Esq./MBA is newest contributor to the eDiscoveryJournal. He is the President and Founder of Review Less, a predictive coding consulting company which has served as experts in the Global Aerospace case as well as created dozens of examples of educational content in articles, blog posts and ESIBytes podcasts describing predictive coding, workflow approaches, and how to optimally staff predictive coding or analytical review projects. Review Less also has also evaluated numerous predictive coding software packages.