January 15, 2021

Recommind’s “Predictive Coding” Patent: More PR Than IP

Since Recommind officially announced the decision by the US Patent and Trademark Office to award the company a patent for what its press release termed a “Workflow and Technology for Computer-Assisted Document Analysis and Review”, the blogosphere has been flooded with discussions regarding its implications for competitors and the e-discovery landscape at large.

In the days since the announcement many of our competitors and colleagues have observed that the methods and technology specified in the patent are both too broadly defined and not sufficiently unique to stand up to scrutiny or challenge. Some contend the patent should never have been awarded at all. While relatively new to the legal world, as Catalyst CEO John Tredennick noted in yesterday’s Law Technology News (LTN) article, Recommind Intends to Flex Predictive Coding Muscles, “the technology, known as relevance feedback among computer scientists, began in the 1970s.” Tredennick went on to say, “We’re puzzled that you can get a patent on what seems to be 40 years in the making in the academic community.”

At the heart of the matter is Recommind’s sense of ownership over the term “predictive coding” and the lengths it to which it appears they are willing to go to enforce its use. Most interesting is Recommind’s continued use of the trademark symbol TM, even after the United States Patent and Trademark office rejected their registration of “predictive coding” citing the term was too broad. The term is confusing for many reasons beyond the ones stated above. The technique and its associated technologies are really an evolution of machine learning, a region of inquiry, as the name suggests, that has been going on in retrieval systems for over 50 years – when computers were called “machines” because they actually resembled a die press and operated using punch cards.

Many other blog posts, tweets and articles point towards a general belief that the announcement is little more than window dressing. Am I engaging in a bit of wishful thinking? Are criticisms of Recommind’s actions merely sour grapes? Are others simply jealous they didn’t think of it first? In short: no. As Amir Mile, CEO of Equivio pointed out in his response to the news:

…at the time of Recommind’s filing, in May 2010, there were many vendors actively offering predictive coding applications in the ediscovery market. This was clear to anyone attending last year’s LegalTech New York conference in February 2010. In their 2010 survey report on predictive coding vendors, the eDiscovery Institute lists 11 predictive coding providers. In addition to Equivio and Recommind, the companies surveyed included Capital Legal Solutions, Catalyst, FTI Technology, InterLegis, Kroll Ontrack, Valora and Xerox.

Milo goes on to say that numerous industry vendors have filed patents that pre-date Recommind’s application and their claim of exclusivity for predictive coding is “disingenuous”. I agree wholeheartedly. But it goes beyond that. It further muddies the water of an already murky pool. They have made it clear that they do not intend to prosper by winning in the war of ideas, but by declaring a propaganda war of hype and intimidation. Such a gambit (and it is a gambit, in the true sense of the word) only stifles creativity, hampers legitimate discourse and competition and engenders a race to the bottom. My guess is the e-discovery community is too savvy to take the bait.

What is most striking about the announcement is the aggressive stance of the press release and subsequent comments by company officers. It would be difficult to characterize the tone of Recommind General Counsel and Vice President of Marketing Craig Carpenter as anything but intimidation in an article posted on Law Technology News (LTN). “‘I’m cautiously optimistic that it will self-correct the market,’ Carpenter said. ‘Let’s revisit this in a quarter or two and see where things stand. Hopefully we’re not talking about a lawsuit.'” One can almost see him mournfully shaking his head before gazing meaningfully at the portrait of Don Corleone hanging above his desk.

At the very least, the announcement, ongoing discussions, and potential cascade of challenges is a well-timed PR boon for Recommind. Why well timed? In the LTN article, Carpenter “added that [an] IPO plan is in the ‘not-too-distant future’ without elaborating, and emphasized that Recommind is not for sale.” In case anyone was wondering.

The reality is that we actually may have someone else to blame for this move. Adiscov, holder of a patent issued in 2004 (filed 11 years ago) entitled, “Method and System for Providing Electronic Discovery on Computer Databases and Archives Using Artificial Intelligence to Recover Legally Relevant Data”, sued Recommind in Virginia for patent infringement in February, 2011. As there is little information available on Adiscov beyond its appearance in such suits, it is up to readers to determine whether the plaintiff is a patent troll or a legitimate competitor. Whatever the case, it raises the question as to whether it inspired Recommind to play the same game.

In his blog post, Dawn of the Predictive Coding Wars?, commentator Barry Murphy noted, “Clearly, Recommind feels it has pioneered the concept of predictive coding and doesn’t want any competitors riding on coattails.” As noted earlier, though Recommind’s trademark application was denied, as Murphy notes, this hasn’t stopped them from buying the URL and using it as a marketing vehicle.

The e-discovery market is intensely competitive—even contentious—at times. It’s a huge and growing market. Recommind’s press release cited a recent Gartner study that predicts the e-discovery software market alone will be worth $1.5 billion by 2013. What they did not include was that the same study predicted that 1 in 4 vendors will be eliminated by market consolidation by 2014. While I sympathize with Recommind’s apparent frustration and its attempts to curb the overuse, and sometimes flat out misappropriation of the term “predictive coding”, I can’t condone its decision to assume the role of enforcer. Is it ultimately more constructive to collaboratively work towards more defined industry standards or to cow competitors into paying tribute in the form of licensing fees under threat of litigation? (Hint: the correct answer is the one that doesn’t hurt customers by stifling innovation.) One thing is certain: the patent grab is nothing more than hyperbole. The patent contains nearly no technical advancements and is little more than workflow. Prediction: it doesn’t sustain a legitimate challenge.

Devin Krugly

Devin Krugly is the VP of Marketing and Business Development at AccessData. He joined AccessData from ExxonMobil to guide the growth of the company’s marketing initiatives, lead generation and to modernize the company’s global VAR and partnership network. Prior to his current role, Devin led several large multi-million dollar solution design and implementation projects for the world’s largest publically traded company, ExxonMobil. His most recent experience was a three year effort to grow an in-house e-discovery team with proper tools to successfully execute data collection and processing related to litigation. The scope of that project included a year-long process to evaluate potential vendors which led to 24 months of assessing fit and purpose of an e-discovery team and design of an IT infrastructure to support the team’s activities. Prior to his role with ExxonMobil, Devin held a position with Halliburton in their Global IT Security department working on NISPOM compliance and developing best practices related to government classified information. Devin also served in the US Army and was deployed to Bosnia-Herzegovina in support of operation Joint Endeavor/Joint Guard during peace-keeping operations in that region.

More Posts


  1. Thanks, Devin, for your excellent analysis. I have additional comments about the history of predictive coding and the prior art relevant to Recommind’s claims here. http://orcatec.blogspot.com/2011/06/competitors-press-release-about.html.
    I don’t think that the problem with the patent is that it is too broad. Rather, I think that it applies specifically to a very narrow implementation using Relevance Feedback (prior art), Probabilistic Latent Semantic Analysis (for which they have a patent), and Support Vector Machines (prior art). This is acknowledged in their patent application.

    OrcaTec uses different technology to provide predictive coding and will continue to do so.


  1. […] “Predictive Coding” Patent: More PR Than IP Source: original article Author: Devin Krugly Categories: Electronic discovery, e-discovery, […]

Speak Your Mind