A Mathematic and Scientific Approach to 21st Century In-house Litigation

• Citation: Vol. 1 No. 6
• Publication year: 2018

Christopher J. Groh^*

This article discusses the procedures to build an empirical, quantitative model for purposes of producing data-driven settlement figures. The model is intended to substantially decrease litigation risk and reduce litigation cost.

Data speaks, we listen, interpret, contemplate, and then react; data speaks but does not have to control. Data is, at the very least, our baseline from which to act. Data says here is what happens (based on what has happened throughout history). With our data on what normally happens, we can (1) predict the outcome of any future occurrence and (2) test our data's accuracy based on what actually happens. Then, we have even more data on the likelihood of new data not fitting our standard curve of baseline data. So, data is used to predict the outcome of future occurrences, and actual future data is measured against predictions to create variation probabilities (statistical term?).

Setting the Stage

Now, a new case comes in, regression pins a ___ percent chance of ___ outcome. The case actually results in ___. The results are added to the model. But how does the model know why the results did not come out as predicted? It might not—it depends on the sophistication and breakdown of variables (moving parts); however, this is likely a determination made by experts, which can then be used to train our model (machine learning?). Eventually, with enough cases and expert opinion, the model can detect relationships and learn how to act.

The model is essentially quantifying every possible event and outcome. It is generating pathways (this can be visually depicted with a decision tree). It is saying if X, Y, and Z are present, A is the

[Page 403]

most likely outcome; but if X, Y, and F are present, B is the most likely outcome; or ¾ standard value of X makes the likelihood of a Z event increase by 50 percent, and so on. It is analyzing the data in a way that humans never could. Not only does it compare events to outcomes, but events to other events. For example, it could potentially find, on average, that every $1.00 spent on discovery beyond $xx.xx yields diminishing returns.

How do we create such a work of art? With human experts. It starts by understanding the entire process, identifying and defining every moving part; classifying and/or creating a system for classifying standard variables, and mapping standard relationships. Then, we can collect our data, analyze the data, and build a standard model to serve as our "baseline" prediction of future occurrences. Next, we make predictions (run regression, etc.), interpret and quantify actual results, test (compare) our predictions against actual results, and assist the model in determining the cause of deviation. Human experts are still the warriors; the model is the sword.

Incentives to Using Data

In-house legal departments have a tremendous opportunity and incentive to utilize data to make better decisions. The opportunity is now available due to the current data practices of most companies—most companies have some sort of data strategy and collect routine data either for defensive or offensive reasons.¹ Nonetheless, data is available or easily obtainable on almost anything.

The incentive to use data is twofold: (1) increased accuracy in decision-making and (2) corporate standards. The effect of increased accuracy in decision-making substantially reduces risk and cost and increases foreseeability. Risk is reduced because the very nature of risk is the possibility of an unknown (i.e., risk is uncertain negative consequences). Data hedges risk because it allows clear, reliable foresight based on patterns and trends in history. Data reduces cost because it allows for informed planning, budgeting, and cost decisions. For example, if a company uses billing data to understand the cost of outside counsel, per attorney, and sees that each additional attorney costs on average $30,000 extra, depending on the severity of the case, it might be wise to limit the number of attorneys involved—a data-driven decision used in planning and budgeting for litigation. Second, a

[Page 404]

majority of corporations now hold legal departments to the same business standards—i.e., focused on performance standards. These standards focus on efficiency, auditing, cost-reduction strategies, etc., all of which are simply different ways of leveraging data to make decisions or measure performance. Moreover, we all know that the quickest way to impress an executive is by increasing the bottom line, but since legal is inherently a cost center, counsel must get creative to show processes and efficiencies conducive to maximum cost savings. Cutting legal cost absent downside risk is as effective as proportionately increasing revenue—it does not matter where the dollars come from. Therefore, let's use data in the legal department to reduce costs while simultaneously decreasing risk by making mathematical, data-driven decisions.

Pleading the Use Case: Litigation

A major unfettered and inefficient area of in-house legal operations is the handling of litigation disputes. Litigation is inherently risky due to the potential for damages, setting a bad precedent, or harming reputation; however stated, litigation is a major cost and impacts business goals.

Therefore, absent data-driven predictability of the consequences of a claim, the only risk mitigation strategy is to settle. But who then determines when to settle and what amount to settle for? At what point do we experience diminishing returns? The Lawyers for Civil Justice, Civil Justice Reform Group, U.S. Chamber Institute for Legal Reform, sought to provide empirical data regarding the issue of litigation costs affecting the ability of companies to compete in a global economy. In a statement at the 2010 Conference on Civil Litigation at Duke Law School, the group noted that "[t]he reality is that the high transaction costs of litigation, and in particular the costs of discovery, threaten to exceed the amount at issue in all but the largest cases."² The group conducted an empirical study, sent to all Fortune 200 companies, seeking to obtain data regarding long-term U.S. litigation trends. The group attributed a lack of current data to the notion that companies are hesitant to "provide litigation data to researchers because of significant concerns about confidentiality coupled with the difficulty and costs of retrieving and providing data in the formats and for the time periods sought. The resulting lack of empirical data leaves the important question

[Page 405]

of litigation transaction costs to be addressed primarily through anecdotes—which can be compelling but also easily dismissed."³ The survey results indicated that "[c]ompanies are spending billions of dollars yearly on litigation"⁴ and "[l]itigation costs continue to rise and are consuming an increasing percentage of corporate revenue."⁵

As noted, settling a dispute can be an extremely effective cost-reduction strategy. In addition, foresight into anticipated demand allows for efficient planning, budgeting, resource allocation, and informed cost-saving decision-making. Unsurprisingly, the root of this tree is data; the right litigation data allows for the experts to make these cost-reducing decisions. As stated by Daniel Katz, "[a]ided by growing access to large bodies of semi-structured legal information, the most disruptive of all possible displacing technologies—quantitative legal prediction (QLP)—now stands on the horizon. Although different variants of QLP exist, the march toward quantitative legal prediction will define much of the coming innovation in the legal services industry. And it will occur whether you like it or not."⁶

Approach

This article is intended to discuss the procedures to build an empirical, quantitative model (decision model) for purposes of producing data-driven settlement figures. The model is intended to substantially decrease litigation risk and reduce litigation cost. Creating such a model delivers ancillary benefits, such as creating the framework for standard processes, performance reviews, and other major cost-saving, performance-enhancing practices.⁷

Once the model is developed, you do not have to see it as your new boss; even employing it in the background for a set period of time will allow the collection of systematic litigation data until enough has accumulated to begin predictive analytics. As Brenton and Flaherty stated perfectly, "[o]nce you have determined what, why and how to measure, you still have to measure long enough to amass a reasonably sized data set. At that point, you are likely to find that some of the choices you originally made were less than perfect. So you tweak and cleanse the data based on lessons learned and establish a baseline. At this point, you can begin to project the return on investment of your proposed innovation."⁸ The goal is to

[Page 406]

determine the what, why, and how of litigation data. Then, from there, turn the data-knowledge into value by producing quantitative settlement figures and making litigation predictions.

Methodology

The process for indulging in empirical analysis, as opposed to reporting the findings of a traditional case study, will be explained. Nonetheless, the analysis is framed to encompass components of a case study approach, as well as descriptive, exploratory, interpretive, and persuasive analysis.⁹ In addition, this article will explain the implications and unique characteristics of a problem (ad hoc litigation management) while focusing on a single unit of analysis (litigation settlement). A single line of analysis (quantitative settlement model) will be employed while using knowledge gained through research, course work, and practical experience to lay forth the groundwork for creation, implementation, and further analysis.

Research Design

Research design is the process of choosing a topic of study, generating a research question, and choosing the method to...