Predictive Coding And The Real World

Predictive Coding, a.k.a. technology assisted review (“TAR”), is all the rage. It seems, everywhere you turn, someone is expounding this technology as the cure for all e-discovery woes. Well, thanks to a ruling by Magistrate Judge Nan Nolan in the Kleen Products v. Packaging Corp. of America matter, some reality has been injected back into the hype.

Kleen argued that the defendants have to use a particular flavour of Predictive Coding to ensure the accuracy of their production.  Packaging Corp countered with the assertion that they have a tried and true methodology involving iterative keyword searches that has served them well in the past.

After many months of back and forth, Kleen finally gave up the fight and agreed that Packaging Corp can use their standard method for the initial production request and any production request made before Oct 1, 2013. After that date, all bets are off.

While the “Stipulation and Order Relating to ESI Search” does not detail why Kleen gave up the goat, it was likely because they were unable to come up with an acceptable way of evaluating the value of the different search methodologies. Studies such as TREC clearly show that the different incarnations of TAR (keyword searching included) all produce different results when applied against the same set of data. They all have differing costs associated with them as well. Which one is best for each circumstance is something that is hotly contested between vendors, but up to now, no one has come up with a definitive way to quantitatively measure it.

While the application of some type of technology to e-discovery tasks will certainly be less expensive that reviewing every scrap of digital paper, the technologies that best suits a particular situation may not necessarily include Predictive Coding.

​For more information on apply technology in the e-discovery process, contact Wortzman Nickle.