Meet the new eDiscovery, same as the old eDiscovery?

What’s most striking about the Gibson Dunn 2015 Mid-Year E-Discovery Update is that there’s not much at all striking about it. In fact, the only truly new topic of interest that has emerged this year — the so-called “Internet of Things” — is dismissed by the firm as requiring “a rather fertile imagination” to identify how the IoT can seriously factor into a business dispute.

Gibson has eloquently and thoroughly assessed the state of eDiscovery affairs since 2009. And since then, some has changed, but the predominant issues remain the same. Compare 2010’s recap with the latest review. Confusion over predictive coding? Check. Percolating rules amendments? Check. Courts indecisive on social media? Check.

The stasis from 2012 to 2015 is even more apparent.

2012 quote: “Has Predictive Coding’s Time Finally Come?”

2015 quote: “Predictive coding continues to be the big underachiever of e-discovery.”


2012 quote: “Reform of Federal Rules Makes Significant Process”

2015 quote: “It has taken an interplanetary probe nine and one half years to travel to Pluto, so the five years that the latest e-discovery amendments… have taken to come to fruition seems pretty good in comparison.”


2012 quote: “Several years ago, the cross-border transfer and disclosure of information…was far less frequent than it is today.”

2015 quote: “The cross-border transfer and disclosure of information remains a challenge.”


2012 quote: “Marking a departure from previous updates, the subject of sanctions is not our lead story for 2012.”

2015 quote: “Sanctions are an ever-present danger in the e-discovery world… [c]ontinuing a trend that we have seen in recent years.”

You get the point.

To be sure, there have been some significant developments within each of those broad categories, but those “advancements,” if we can call them that, have generally occurred on the margins. Judicial guidance on cost recovery, for instance, is in the same state of disarray as it was prior to the Third Circuit’s 2012 would-be seminal ruling in Race Tires AmericaInformation governance hasn’t taken hold in the all-consuming fashion in which it was touted, and some have since questioned whether it’s even worth pursuing. The duty to preserve, yes, still applies to text messages. And, no, we still haven’t figured out how to capably collect those suckers.

What’s also disheartening is the way in which eDiscovery continues to be framed: not as a way to gain a strategic advantage or insights into facts and arguments, but as a threat — a costly, risky elephant in the ointment that must be dealt with. Indeed, the very first line of the 2015 review is: Progress on Some Fronts, But Significant Dangers Remain, and New Dangers Emerge.

Gibson is right on the money. eDiscovery is dangerous. Don’t shoot the messenger.

Predictive coding’s holding pattern

The firm’s comments on predictive coding deserve particular attention. Gibson mentions that:

One commentator recently opined that [the reason predictive coding has not caught on] is not because of software makers, vendors, judges, governmental investigators or clients, many of whom favor and promote its use, but rather it’s due to litigation counsel “not understanding the value of TAR” in appropriate cases…. There may be some truth to this, but we think the story is more complex.

The story is absolutely more complex. Keep in mind…

a) The use cases for predictive coding are limited to begin with because, in general, you need a lot of data to justify the up-front costs, technical complexity, and hand-holding to use it. In other words, “mass adoption” of predictive coding is unrealistic in the short-term because it’s a workflow/technology that, in its current state, isn’t economically feasible or useful for the vast majority of organizations. How many 50-person law firms can afford to plunk down the set-up fees for Relativity?

b) Assessments of predictive coding’s adoption are skewed by the unrealistically large potential market that observers, practitioners and providers have identified for it. Due to its complexity, cost and limited use cases, it is pretty much impossible for at least 95% of litigants to actually use predictive coding. So in that light — focusing on the 5% of the market that actually exists — adoption is probably much more robust. Whether predictive coding’s application has been successful is another matter. Anecdotally, the prevailing response among non-experts tends to be “It’s great… when it works.”

c) The reason adoption of predictive coding has stalled, if it has stalled at all, is not due to litigation counsel “not understanding” its “value” — there’s nothing hard to understand about a value proposition of faster, cheaper, more accurate — but because that value prop has been tremendously overstated by predictive coding providers and proponents who have long struggled how to convey in plain language how their offerings actually work. To be sure, user error limits the tools’ efficacy as well — but such is the nature of complex technology.

d) Here’s the caveat: predictive coding is almost definitely being used in some significant number of cases where it isn’t being disclosed to the other side — and thus not being disclosed, period.

Which leads to another point. Gibson also states that “probably the biggest impediment” to adoption is concern that parties seeking to use it must share coding decisions, including potentially disclosing work product.

While that may be true, what continues to be unclear is why parties disclose the use of predictive coding in the first place. Certainly there is no obligation to share specific search methodologies, only to defensibly fulfill discovery obligations. A responding party need only make “reasonable efforts” — if those efforts includes a room full of attorneys, search terms, or predictive coding, so be it.

It could be the case that disputes are only likely to arise — and court intervention is only necessary — should the opposing side challenge an allegedly deficient production. It might also be true that not disclosing the use of predictive coding is also a way to avoid the so-called “TAR tax,” which Gibson describes as an attempt by the requesting party to “effectively preclude the responding party’s ability to use predictive coding by demanding conditions that the responding party will likely find unacceptable.”

Surely the show-your-hand approach has yielded mixed results.

Recent events in the Rio Tinto v. Vale case, in which Judge Peck declared predictive coding to be “black letter law,” are particularly worthy of mention. In the wake of the court’s proclamation that predictive coding limits “undue burden,” multiple burdensome disputes over how to limit burden have broken out. First the parties battled over whether search terms could be used to whittle down the document universe prior to application of predictive coding (The court’s answer: Yes).

Then, perhaps more significantly, the parties battled over whether a special master should be appointed to assist in executing — and resolving disputes over — the predictive coding protocol the parties negotiated. Potential battles loom over the adequacy with which Vale trained its workflow, and the mathematical rigor Rio Tinto applied to the recall and precision estimates of its own performance. Maura Grossman of Wachtell Lipton will referee.

That appointment, of one of the figures behind the studies that assert TAR’s superiority, and on which Judge Peck’s own assessment’s of predictive coding’s acceptability rely, will do nothing to dispel the notion that working with the other side to craft search methodologies is a huge gamble.

Indeed, a casual onlooker might say: You have two of the most prominent authorities on the subject mediating the dispute (Judge Peck, Grossman); two leading law firms (Quinn Emanuel, Cleary Gottlieb); and leading vendors (Deloitte, Precision Discovery)… And even that group can’t figure it out.

Writes Philip Favro, Senior Discovery Counsel at Recommind: “That a special master would be required to help the parties – where they entered into a predictive coding use protocol – seems counterintuitive and underscores the need for counsel to proceed with caution before agreeing to such protocols.”

Also, the special master will be paid $850 per hour, plus the rate of whoever she taps for help. So add that to the tab.

‘Consumerized’ eDiscovery?

In the interest of full disclosure, Logikcull does not offer predictive coding in the context it has been described here. Logikcull could be (mis)construed, however, to be among the companies attempting to “consumerize” eDiscovery —  which Gibson describes as attempts “to sell e-discovery software as a service (SaaS) directly to end users (e.g., individual lawyers) much like an app, with little or no professional services component involved.”

The post goes on:

Vendors offering these cloud-based e-solutions typically target law firms and companies with professional litigation support staff with the skills to manage complex projects and execute difficult tasks. The consumerization approach, by contrast, directly targets individual lawyers and rests on a sales pitch that implies e-discovery is easy. The problem is that all too often, even with the best technology, it isn’t, and the risks and consequences of failure, as always, remain great (see our sanctions discussion above).

This is probably an accurate assessment of some companies, but also consider:

  1. This view presents a distinctly “Biglaw” perspective. In many instances, especially outside of federal courts and/or complex disputes, it may not be within the bounds of proportionality to hire a traditional vendor to assist in discovery, considering the amount of information at issue, the value of the case and its importance. If there are companies suggesting that shorthanded lawyers use off-the-shelf products for a 500 GB tort case valued at $100 million, they are wrong and should be called out.
  2. The “consumerized” approach to directly target lawyers likely has less to do with their roles as “end users” than their roles as decision makers and holders of the pursestrings. Any lawyer that is performing discovery without any help at all — from, for instance, a technically adept paralegal — is likely in a position that he or she must do so (e.g. a solo practitioner).
  3. So called “do-it-yourself” discovery might be necessary in some instances (see point 2) — and in those instances, the level of expertise of the judge and opposing counsel is potentially such that more leeway is given as to the adequacy of discovery efforts.
  4. This “consumerization” trend is even more reason for lawyers to obtain basic discovery skills — and to achieve competency.
  5. Standalone discovery products may be of more use to corporate consumers than to law firm consumers, who are likely to pass the cost of discovery back to the client, irrespective what methods are used (and thus have less incentive to limit cost). “Consumerized” products are of particular value to the extent they can empower capable corporate legal departments to pare down data before outside counsel and vendors are brought into play.

A word on cost recovery

Finally, on the topic of cost recovery, a correction: Gibson cites to Comprehensive Addiction as evidence that several courts “continue” to rely on the Third Circuit’s March 2012 decision in Race Tires America to narrowly award costs. While the Colorado district court in Comprehensive Addiction did indeed cite to the Race Tires decision, its own decision to broadly award costs to the prevailing defendant certainly does not reflect the Race Tires court’s approach or rationale. In fact, the Comprehensive Addiction case might be the best evidence yet that courts are either misinterpreting, confronting in disagreement, or flat-out ignoring the Third Circuit’s (not to mention the Supreme Court’s) rationale. A more detailed discussion of Comprehensive Addiction and its consequences is available here. We have also covered the troubling unintended side effects of taxing electronic discovery costs.

Robert Hilson is a director at Logikcull. He can be reached at This post reflects his opinions and not necessarily (but probably) the opinions of his employer.