Lessons from Libor: How to Apply Machine Learning for Document Digitization
The imminent demise of the benchmark Libor interest rates is one of the most important developments in the history of financial markets and one that will pose a challenge for all financial institutions. Fortunately, recent technology innovations are available to assist in the monumental task of defining issues, managing data and creating new processes and protocols for progress. In particular, the challenges posed by Libor can be well served by the application of machine learning.
In our white paper, Libor as a Template for Digital Document Transformation in Financial Services, we lead an exploration of the ways in which machine learning can be utilized to answer the challenges posed by Libor, including pitfalls and important lessons to be learned. In this blog post, we address some of the key ways that the project should be structured to ensure maximum success.
Key Elements of a Libor Change Initiative
Libor has been used for most contracts and agreements in derivatives, bonds, mortgages, commercial and retail loans. In documentation, the term is used trillions of times in hundreds of millions of contracts and agreements. Successfully managing the transition to a post-Libor financial world is a large and sensitive task and therefore requires a comprehensive approach that marries data science with change management. Some of the most important functions include:
Setting the table: Just as a house rests on its foundation, a successful change effort is based upon the work put in on problem definition and setting success metrics. The more that the current state is analyzed and comprehensive and realistic goals are set upfront, the more that success flows down the road.
Data is fundamental to success: Data is also foundational to the process. In order for the effort to pay off, care needs to be taken from the beginning or outcomes will be compromised. The issues are complex and important steps include access to appropriate and necessary data, take time for data labeling, and to perform rigorous exploratory analysis.
Utilize appropriate tools and processes: You’ll need to use the right tools for each step and the number of different skills needed for such a comprehensive undertaking is broad. Two examples are optical character recognition (OCR) and multiple, additional ML models to separate documents into categories. From there, other steps include business model grooming, feature engineering, and error analysis.
Make the results work for you: Too often, there is a failure to fully follow through and map results to success metrics and follow up to strategize and create a roadmap forward. Candor and objective analysis are necessary to achieve optimal results and commitment from stakeholders is an essential ingredient as well.
Successfully Applying Machine Learning to the Libor Data Challenge
The end of the Libor era is one of the most important and far-reaching events in the history of financial markets and making the most of the transition to a new regime can be viewed as an opportunity to accomplish a broad transformation for a large number of important areas within the enterprise. The task isn’t simple but the potential rewards are large. Making the most of them will help determine which firms do well and which fall behind. Tackling the challenges with the application of new technologies such as machine learning as detailed in Libor as a Template for Digital Document Transformation in Financial Services is critically important.
Maven Wave helps drive the future of financial services with innovative business outcomes, fueled by cloud, with risk top of mind. To help organizations maximize economic outcomes and advancements, Maven Wave brings a rich blend of industry-specific technological expertise, agile-integrated design, and best practices for transformation.