Today, managing legal defense costs is often relegated to legal e-billing software that was developed some 20 years ago. State-of-the-art at the time, these applications, which were designed to automate billing and manage legal spend, have now been asked to do more than they were intended—uncover data that can be used to inform business decisions of insurers and captives. The problem: Most legal e-billing software has limited access to critical data that is needed by sophisticated analytic technologies, which are used to develop new intelligence for managing legal claims defense costs.
Legal e-billing software typically forms the core of current enterprise legal management platforms, which include matter management, workflow management, content management, and reporting and analysis among other tools. The e-billing component of an enterprise legal management platform was originally designed to automate the billing process and to provide a repository for checks and balances on defense cost spending that could help with claims litigation management. But with more recent and often unsustainable year-over-year increases in insurers’ and captives’ legal defense costs, many in the industry have turned to legal e-billing software as a way to develop information that can be used to manage defense costs.
Legal e-billing software is based on a set of codes known as Uniform Task-Based Management System (UTBMS) codes, which at first glance appear to be a goldmine of information regarding the defense costs of law firms. In fact, they do provide information about pretrial pleadings and motions, case assessment and administration, discovery, appeals, and expenses. But a deeper dive into these codes has revealed several fundamental flaws.
One of the most important shortcomings is that the codes are prone to wide interpretations by paralegals and administrative support teams that can result in inconsistencies and inaccuracies in the core data used in any analysis. For example, a common activity like creating an email can be coded as drafting by one law firm and as communication by another. Some firms that tend to take shortcuts often combine the coding of two or more activities, such as traveling to a deposition and attending it, into one code. Alternatively, a firm may ignore important distinctions within coding categories, which often happens in failing to separately code expert depositions from all other depositions.
Taken individually, the coding inaccuracies may not seem significant but, for insurers or captives that manage several or even hundreds of law firms, these types of inaccuracies and inconsistencies become magnified by the varying interpretations of dozens or hundreds of paralegals and other support staff, making the output from this data highly suspect.
The data pulled from code-based legal e-billing software, if it can be trusted, is also limited to summary types of output. For example, an insurer typically can only see how the total defense spend breaks down by activity, but cannot determine how much work went into any of the particular activities. For example, assuming the manually coded UTBMS data was coded correctly and consistently across law firms (a significant assumption that in practice does not happen), an insurer or captive may be able to see how much was spent on depositions in total, but it cannot determine how many depositions were taken. Without knowing how many depositions were taken, it cannot be determined whether the average cost of depositions is changing or whether Law Firm A has higher or lower deposition costs than its peers.
A new worldview
These shortcomings led actuaries at Milliman to look in another direction. Instead of using UTBMS codes, they decided to look into the possibility of using invoice line descriptions as a more transparent and complete source of data. They used natural language processing (NLP) techniques, a branch of artificial intelligence (AI) tailored to defense costs invoices. They were able to extract the intelligence and data from the invoice line item descriptions, thus allowing insurers and captives the ability to develop data-driven strategies for managing defense costs.
By deploying a singular data mining algorithm, Milliman’s legal claims analytics eliminate code-based inconsistencies and inaccuracies because every single line description from every law firm from every client from every state is read in the same way. This approach results in a fully consistent and reliable data set with much more granularity than is possible from any UTBMS-based data set.
The use of machine learning technology in Milliman’s legal claims analytics translates into a platform whose understanding of invoice line item descriptions improves over time. Differing abbreviations in invoice descriptions and contextual idiosyncrasies become part of the dictionary or language database, which cannot be replicated with the use of code-based e-billing applications.
With this data, Milliman’s claims litigation analytics, known as Milliman Datalytics-Defense®, can develop new intelligence for insurers that differentiates among types of depositions taken; and determines the number of depositions or motions that each lawyer within a law firm has taken or filed, the success rate of the motions filed by lawyer, the types of depositions by lawyers and law firm, and the number of trial days involved in a case; and it identifies the type of witnesses who were deposed, among other capabilities.
This level of legal claims analytics allows insurers and captives to better understand the cost drivers of their defense spending, identify law firms with outlying cost structures on both ends of the spectrum, and make more informed decisions regarding the allocation of their legal defense spending. In this way, insurers can direct more cases to the most effective law firms and phase out less effective performers.
The cart before the horse
What has often evolved from legal e-billing software in use today is a patchwork of fixes aimed at incorporating analytics into UTBMS-based data sets. Frustrated with the output from their legal e-billing software, some insurers and captives have tried to reverse-engineer the development of analytics by cleaning up the UTBMS data that results from their e-billing applications. While the results are somewhat of an improvement over the bare output from their legal e-billing software, this backed-in approach to analytics falls far short of what text-mining technology can deliver because it is still based on the fundamentally incomplete UTBMS-based data set.
Milliman’s Datalytics-Defense includes a legal e-billing software, which automates the invoicing process and provides checks and balances to defense costs that can assist with claims litigation management, but its e-billing application is the data source for the development of the analytics. In this sense, Milliman’s e-billing application is the means to accessing core invoice line descriptions, which powers its text-mining technologies.
In developing new intelligence around defense costs, Milliman’s legal claims analytics allow insurers and captives to drive down these costs, using risk-adjusted benchmarks. Insurers and captives can compare the expected defense costs of a case against its actual costs. The goal is to maximize the allocation of legal defense dollars to the most effective law firms—an approach that has saved some insurers up to 15% of their defense costs—rather than micromanage these resources through sometimes tedious audits, which under the best of conditions can reduce costs by only 2% to 3%, typically.
When based on uncompromised, clean data, legal claims analytics can be a powerful tool in managing legal defense spending. But taking the next step in managing defense costs requires a seismic shift in accessing the information content needed to analyze spending. This shift requires moving away from code-based e-billing applications and adopting technologies that can add visibility to the allocation of legal defense resources, increase transparency, and inform business decisions about one of the industry’s largest expenses.