Revolutionizing eDiscovery With Predictive Coding

Friday, July 1, 2011 - 01:00

The Editor interviews Robert Tennant, CEO, Recommind, Inc.

Editor: Bob, tell us about Recommind and your involvement.

Tennant: Recommind provides intelligent software applications to solve business problems. Our general area of emphasis is information governance and knowledge management for the legal market, particularly in the field of eDiscovery and related information management. As for myself, I first got involved with the company when I provided it with seed funding at the end of 2000. The company was launched in 2001, and we just had our tenth anniversary.

Editor: How has business been for Recommind the past 12-18 months?

Tennant: It's been great; our revenue has been growing at a rate slightly over 100 percent this year. This has been driven largely by our leadership in providing innovative eDiscovery applications. After a number of years of our evangelizing, the demand for Predictive Coding and the application of our other machine learning techniques is growing. There is also strong interest in areas related to corporate information governance, including data remediation, data classification, automated filing of documents and the application of retention schedules.

Editor: What types of customers does Recommind work with?

Tennant: The short answer is all types, but our main client base is composed largely of corporations. We also serve many law firm and government clients, including various regulatory agencies. Our software provides soup-to-nuts solutions for the entire range of eDiscovery challenges within larger corporations and is also available as specific applications for smaller organizations with more limited needs, as well as in a hosted software-as-a-service setting so that organizations can pick and choose from applications in a rapidly deployed, on-demand model.

Editor: Recommind operates in some pretty competitive markets. What makes Recommind different? Why do customers choose Recommind?

Tennant: Recommind is differentiated largely through our technical leadership and innovation. We are unique in having invested so much in developing the science of machine learning and developing solutions based on this expertise. Most organizations in the legal space don't have the background that we have in working with the more sophisticated algorithms. Our customers appreciate this expertise and our investment in their success.

Editor: Recommind products run the gamut, from information management through eDiscovery. Is there a connection between information management and eDiscovery?

Tennant: By engaging early in good information management practices, our customers can protect themselves from headaches and additional costs down the road when they face litigation or other discovery needs. Because better information management is the key to success, we developed our CORE, which stands for Context Optimized Relevancy Engine. This common platform provides a basis for e-mail management, knowledge management, eDiscovery and other applications. By organizing information ahead of time, when the time comes for eDiscovery, Recommind provides an easy way to transition from information management to eDiscovery, reducing risk and saving a lot of time and money.

Editor: Recommind made a big splash recently by announcing it has been issued a patent for Predictive Coding. Can you tell our readers what Predictive Coding is and why it's such a big deal?

Tennant: Predictive Coding is the application of machine learning techniques in an iterative process to code documents in an automated or semi-automated manner. This process engages attorneys early on to leverage their knowledge and understanding of a case. With a few examples the attorneys train the system to code documents more accurately and consistently than with manual review. The system can be utilized either with supervision, where reviewers are still coding each document, but at a much faster rate, or without supervision, where the computer codes most documents but where quality checks are iteratively performed by the review leaders. In both cases the process is much, much faster and more accurate, which translates directly into cost savings.

By engaging the legal team's more senior attorneys earlier in the process, better case decisions can be made with greater certainty that the team has the information necessary to make the right decisions. By utilizing the computer to assist the review team, the reviewers get the benefit of (1) better review organization, and (2) guidance, both of which help speed up the review dramatically, even when reviewers look at every document, as well as (3) a quality check.

People and intelligent computer applications have different error patterns. When people concentrate and put something in a bucket, they're usually right - but they lose concentration and get tired. Computers, unlike people, don't get tired, so by utilizing the attorneys to guide the computer, you get the best of both worlds. In many ways the Predictive Coding process is very much like the review process today, where the more senior attorneys tell reviewers what they want them to look for, but instead of a Word document description, examples are provided to the computer.

The Predictive Coding process that we've patented is iterative. This enables the system to be retrained to recognize documents that become relevant or may be privileged because of knowledge gained or changes in issues as a case proceeds, which helps the reviewers a lot. This iterative process provides greater certainty that you're not missing anything. The computer can give you an actual statistical confidence estimate of how accurate you are, which you can then take to the bank in terms of defensibility, even where reviewers are not looking at every document. As I mentioned earlier, though, in many cases the reviewers still look at every document - just a lot more quickly and accurately. In that case, the defensibility argument is the same as it is today - someone looked at each document.

Predictive Coding is a big deal because our clients tell us that 70 percent of the cost of document review is lawyer review time. Predictive Coding cuts that dramatically, and with a much higher degree of accuracy. So the end customer, the company, saves a lot of money and can be more confident as they make decisions throughout a case.

Patent protection for Predictive Coding is significant, first, because it's a recognition of Recommind's innovation and significant investment of effort and money over many years to make document review more accurate - and far less costly and time consuming - and, second, because it affords us some protection from others trying to copy our innovations, which will let us concentrate on continuing to innovate.

Editor: For which types of clients and cases is Predictive Coding appropriate?

Tennant: It's appropriate for everybody, but the benefits increase with more data. If you need to review just a few thousand emails, you will start to get real savings and accuracy improvements immediately.

Editor: What about the role of Predictive Coding in early case assessment (ECA)?

Tennant: Recommind has a powerful set of analytical tools, including sophisticated concept search, the underlying technologies that form the basis of Predictive Coding, and a number of other machine learning and other analytic and data visualization tools that we employ in our ECA application. Predictive Coding can be utilized to assist in ECA, but it's about methodically reviewing material quickly. Many of the other tools we have are more effective when you want to just quickly find out what is contained in a data set, which is why we have them. When customers have our ECA and Collection application, these tools can be brought to bear before data is even collected, making for smaller and less painful collections. We provide the "early" analytics in ECA.

Editor: How does this all relate to information governance?

Tennant: The information governance challenge is that organizations just don't know what data they have. What we call information governance is helping people understand what information they have, so that data can be extracted, eliminated, filed appropriately, or policies can be applied. Most companies that we speak with today say they are keeping everything because they are scared of possible spoliation claims or simply of eliminating something they need to keep for regulatory or business reasons.

With our information governance applications, we help customers file what they do need and eliminate data they don't need, when it's appropriate for them to do so. Being able to deal with data effectively can produce significant cost savings as a result of reducing the IT resources required to maintain extraneous data. It also enables organizations to actually implement their retention policies in practice, not just theory.

Editor: Do customers use Recommind products on their premises, in the cloud or both? Why?

Tennant: Customers use our products in both settings. We have applications that cover the entire EDRM lifecycle, with components that help during each phase. They are all built on the same platform so that you don't have the risk of dropped handoffs between applications or even just the pain of managing multiple applications. Some clients don't need a full suite of applications, in which case they can buy individual modules within the suite for use on-premises or they can just utilize our applications in a hosted environment on a case-by-case basis.

The hosted solution provides an online review platform to which you can send data. We will put your data in our systems and make it available for you or your counsel to review online. There are no IT infrastructure requirements or anything of that nature.

Editor: What do the next 12-24 months hold for Recommind?

Tennant: There is a lot more that we can do, and we're continuing to invest in areas where we bring unique value. We plan to continue our innovation and leadership in eDiscovery and in Predictive Coding. You can expect some new product offerings around information governance based on our CORE platform and continued evolution of e-mail management for both productivity and governance. Of course, we will also continue to emphasize the great customer service that our clients receive from us. We're only as successful as our clients are, and we know it.

Please email the interviewee at with questions about this interview.