Transforming Data into Actionable Insights - An Evolution of Data Analytics in the Asset Management Industry
Smart analytics have the potential to significantly alter the way asset managers do business — and that’s important since the industry touches about one-third of all Americans. Members of the Investment Company Institute (ICI), an association representing regulated funds globally, manage $20.7 trillion in the U.S. and serve more than 100 million U.S. shareholders.
In the decade since the global financial crisis, fintech companies have debuted new tools that have drastically enhanced the analytical capabilities of financial services companies to keep pace with regulators. In-house compliance departments— which are responsible for ensuring that firms are operating within all rules, regulations and industry best practices — have new tools that enable them to better monitor the activities impacting their clients.
Global asset manager Legg Mason has embraced recent trends in digital transformation to enhance the way it efficiently dissects large data sets to extract meaningful insights. Legg Mason has $727 billion in assets under management as of Dec. 31, 2018. Founded in 1899, it has about 3,300 employees spread across its Baltimore, MD headquarters as well as 38 other offices around the world.
In recent years, there has been a push for increased automation and advanced analytics across all industries. To keep pace, Legg Mason has made enhancements to its infrastructure, adding a cloud deployment to its technical strategy and allowing it to leverage new technologies.
In an effort to encourage these changes, the firm created a collaborative cross-departmental artificial intelligence (AI) task force in 2017. It was charged with enhancing enterprise awareness of AI capabilities and identifying a range of use cases ripe for disruption. The group focused on opportunities to deploy robotic process automation, machine learning, and deep learning, separating the hype from reality. The deliverables of the task force included over 100 possible use cases. Additional work ensued; the task force weighed the difficulty of implementing each case against the impact on the organization. Ultimately, it developed a use case prioritization framework and a list of recommended projects to proceed with.
One of the projects the task force chose to showcase was the impact of fintech on electronic communication surveillance. Legg Mason currently uses a third-party vendor that utilizes natural language processing (NLP) and machine learning to sift through thousands of emails a day. The system only alerts the compliance department to the ones with the highest behavioral risk score, which could indicate unethical or harmful actions by an employee. Combining elements of NLP and behavioral science means compliance department employees receive a focused list of electronic communications to review. The benefit: A significant reduction in the human hours that were spent reviewing emails.
Another highly analytics-driven function in compliance is analyzing the trade records and accompanying data sets related to employees’ investing activities to identify regulatory and portfolio compliance risks. Historically, these reviews were conducted by looking for known fact patterns with the assistance of a business intelligence system. These systems were very good at looking for patterns that were already well defined but were not as effective in proactively identifying new behaviors that could be risky. Today, though, Legg Mason is harnessing the abilities of open source tools along with a new visual analytics system to drive an advanced cognitive analytics approach.
Legg Mason’s digital toolbox has been rapidly expanding in recent years with the addition of a data visualization tool, R (a statistical computing language), Python (a data manipulation and data analysis language) and a cloud ecosystem. The team is leveraging the power of open source data science languages to create more sophisticated tests focused on exceptions and anomaly detection. The open-sourced data science journey began with R and then transitioned to python due to the wide array of packages available and the language’s dominance in machine learning algorithms. The diversity of modeling tools and a rich data science community in Baltimore offers abundant resources across various disciplines that all build analytics solutions in python.
Another benefit of coding in python is that it will scale with the company in terms of complexity of tests and with the types of data that are brought in. As Legg Mason moves toward greater analysis of unstructured data and deep learning models, the programming language will stay the same.
As more intricate tests are developed, it becomes increasingly important for Legg Mason to provide clear and concise information to end users. This is where the visualization side of business intelligence systems shine; they are able to tell a story in a way that the human brain can easily consume. In many cases, the data being displayed is the same as the data found in a traditional dry business intelligence report but the results are conveyed to the business users in a quick and meaning full way so they are able to quickly absorb all of the information and act on it.
For example, the compliance department receives reports that highlight employee securities trading activities. The reports identify trading around high prices and accumulation. The reviewer is quickly able to see the price movement of security as well as an employee’s holdings before and after the price surge. This could prompt a conversation where the employee is asked to provide additional details on the trade.
Building smart analytics involves human talent, the right tools and infrastructure, and a collaborative environment. Legg Mason has made great strides in all three of these areas by encouraging experimentation and innovative thought.