Fighting financial crime with advanced analytics
Financial services organizations increasingly are using advanced analytics by supplementing the underlying financial crime systems to improve the quality of alerts and to increase efficiency in identifying suspicious activities. Among various analytics, machine learning (ML) has been particularly effective, as it can identify patterns of behavior from large data sets. Machine learning’s capability to identify financial crime risks more accurately than rules-based techniques means that it can improve the efficiency and effectiveness of financial crime controls. ML is often used to automate complex processes and ultimately save time. Machine learning uses two types of techniques: supervised learning and unsupervised learning.
With supervised ML, a model is trained with historical data to produce an output that predicts future activity. Supervised ML can efficiently absorb large volumes of data and produce output data so that it predicts future behavior. While it can reduce manual work and cost, supervised ML also has the potential to introduce bias. As the process is conducted with limited human decision-making, it also can be hard to understand and to explain.
In contrast, unsupervised ML analyzes input data and then finds hidden patterns or intrinsic structures in historical data. The process is more driven by algorithms such as k-means clustering and anomaly detection, and it requires human interpretation to make sure it reveals insights of the underlying problem. With unsupervised ML, financial services organizations can more easily explain the process and outputs.
The ability to explain an ML model is critical for financial services organizations’ model risk management governance and for regulators. Regulators generally have been more comfortable with the use of advanced technology when there is some human decision-making involved in the process and when there are controls around the approach that allow them to validate a model.
Adopting advanced technology: How to get started
While some financial services organizations still view ML as nice to have but not an essential component of financial crime prevention, those with the appropriate resources are beginning to incorporate ML and other forms of advanced analytics in their financial crime detection efforts.
For those interested in getting started, the first step is to identify the technology that best meets the business’s needs and its financial crime program’s objectives. Financial services organizations should conduct a financial crime technology assessment to understand what data is available, understand their risk profile and business needs, and determine what future-state needs could be met with the use of technology.
The next step is to conduct a data quality analysis. The availability of data in financial services organizations – like in many sectors – has grown dramatically, opening the door for advanced analytics that can require large volumes of data. However, it’s important to note that the outputs will only be as good as the data. Before diving into any ML endeavor, financial services organizations need to have a good sense of the quality of their existing data.
As financial services organizations begin to explore the possibilities with advanced technology, they need to make sure that the outcome is tied to a specific objective within their financial crime program. Rather than blindly interrogating data in search of insights, financial services organizations should be purposeful about what they’re trying to accomplish. When looking for data insights, organizations often identify inefficient data analysis processes as they either conduct certain processes manually or repetitively. Although ML might be an option given the resources and timeline, another technology to consider is targeted process automation or robotic process automation (RPA).
Many organizations are turning to RPA, which is a form of automation that can enhance existing defined, structured, and repeatable processes by increasing efficiency and reducing the likelihood of human error. A bot can mimic the actions of a human worker on repetitive or mundane tasks, freeing up the human workforce to spend time on more meaningful and value-added activities. In comparison to ML, RPA can be less time intensive and can provide an immediate efficiency to the organization’s operations.
Like with ML, the process of integrating RPA must begin with a clearly defined and desired business outcome that is tied to a specific objective as well, such as reduction of time to prepare a case for investigator review. Before a task can be automated, financial services organizations should consider whether process improvement is required, as automating an existing inefficient process is a common RPA pitfall. By applying process optimization techniques first, financial services organizations can be more confident the deployed automation will generate a positive return on investment.
Placing technology side by side with compliance
Financial services organizations today generally have a dedicated Bank Secrecy Act and anti-money laundering team made up of management, investigators, and analysts to respond to alerts. The team might include individuals focused on transaction monitoring, sanction screening, and customer risk rating with different levels of expertise, all reporting up through the chief risk officer. Larger organizations might also have data analytics functions to support the compliance function that could be made up of just one or two people or an entire department.
The compliance team should not be siloed from the frontline staff that opens accounts and leads the initial customer onboarding process. Suspicious activity detection begins at onboarding, and technology changes can affect frontline systems and processes. Therefore, cross collaboration and communication between compliance and the frontline is important.
Additionally, ownership of any automation should be shared between compliance and IT, as compliance is just one end user that can benefit from automation. The underlying technology and governance over the controls could also be considered for placement within a center of excellence so that management, enhancement, and reporting can be centralized.
Doing more with advanced technology
The goal of implementing financial crime technology is to effectively detect suspicious activity while enhancing user experience and minimizing pain points and manual processes. With advanced analytics and automation, financial services organizations can reduce manual error rates, create efficiencies, do more in less time, and create more consistent, scalable, and sustainable processes.
