How advanced analytics can enhance and transform Transaction Monitoring
Financial Institutions (FIs) play a pivotal role in the prevention, detection, and reporting of suspected financial crimes. Acceleration of new and disruptive technologies are making it increasingly evident that FIs cannot tackle today’s challenges with yesterday’s strategies. The use of advanced analytics in areas such as transaction monitoring (TM) is one way FIs can become more resilient.
The threat and impact of financial crime on society is at an all-time high. Financial Institutions (FIs) play a pivotal role in the prevention, detection, and reporting of suspected financial crimes. Acceleration of new and disruptive technologies are making it increasingly evident that FIs cannot tackle today’s challenges with yesterday’s strategies. The use of advanced analytics in areas such as transaction monitoring (TM) is one way FIs can become more resilient.
FIs have been operating increasingly sophisticated systems to monitor, investigate, and report suspicious transactions over the last few years. These systems sift through vast amounts of data on a daily basis, and as a result flag many transactions that require investigation of possible suspicious activities. However, FIs are facing a range of challenges in operating these systems.
Current challenges in TM
FIs have to navigate a global web of legal and regulatory frameworks on a daily basis. These expectations are constantly shifting, and the increasing focus by regulators on TM constantly reveals significant gaps that require immediate attention.
Due to the added business demands and cost pressures arising from transforming business models, FIs are finding it increasingly difficult to drive and manage the necessary change required in their TM compliance strategies.
Even with adequate time and money injected into TM strategies, finding and retaining the right talent with a specific skills set that combines deep TM knowledge with technical capabilities and regulatory savviness is a challenge. Organizational silos and uncoordinated responsibilities also contribute to the high complexity of executing TM effectively.
As FIs expand and transform their businesses constantly, it is getting harder for institutions to access consistent, high-quality, and standardised data. Poor technology assets, including outdated and legacy tooling coupled with compliance staff using subsets of data that suffer from poor data quality in an ad-hoc manner lead to immature use of data for TM purposes.
Poor data quality unfortunately also has a ripple-effect. For many FIs, their current TM implementations have low alert precision. These institutions have reported that it is mostly due to poor data quality and that processing these false alerts is one of the largest unnecessary manual overheads in AML compliance.
At the same time, monitoring techniques do not utilize the full potential of data and technology. TM is traditionally based on logical business rules, that often lack a sound risk focus and result in many false alerts as well as missed risks.
Current investigation workflows rely on static alert prioritization methods which result in non-focused handling and reviewer fatigue. Poor alert prioritization and alert tagging has an impact on the quality of alert handling, leading to potential overlooking of suspicious activities.
Unpacking the Advanced Analytics toolbox
Responses to these challenges need to balance multiple goals, such as reducing costs, boosting innovation, enhancement of risk frameworks, and alignment with the supervisor. The application of data-driven methods through advanced analytics as an addition to a risk-based approach and responsive design can help to find the optimal balance and make TM more risk focused, more compliant, and more efficient.
The widespread application of business rules is based on familiar patterns in financial crime. When such a rule’s threshold exceeds predefined limits, an alert is being generated. However, this type of monitoring typically returns many false positives, because its predefined approach and often imprecise monitoring criteria necessarily include a large number of regular transactions. This effect is even stronger when the business rule logic is insufficiently based on a well defined risk scenario.
Although the use of business rules cannot be totally abandoned, it needs to be complemented with more advanced analytics techniques. The most widely explored method is to apply predictive modeling (also referred to as supervised machine learning) to generate alerts. Instead of expert knowledge, the identification of potentially suspicious alerts is then based on computer models learning to predict AML risks from historic data. Application of this technique is known to be extremely powerful in reducing false positive alerts, but also requires safeguards to ensure completeness, validity and the right focus of the model (suppressing false positives while keeping false negatives limited).
Profiling, clustering and network analysis
Besides predictive models, many FIs nowadays apply a form of monitoring against client or transaction profiles. At its basis, customer segmentation algorithms can be applied to finetune monitoring scenario’s for specific client segments. For instance, peer grouping can be applied in which a customer’s transactional pattern is compared for specific variables with the average transaction profile of its peers. Or multivariate anomalies can be detected by combining both client attributes and statistics on transaction behavior.
In a more advanced application, data clustering methods (also called unsupervised machine learning) may help to identify clusters of transactions that are unusual and may carry specific AML risks. This way, also unknown risk scenario’s that could not be identified based on known risks or historic data can be traced.
When sophisticated criminals are able to avoid direct monitoring of behavior, Social Network Analysis (SNA) can help. SNA looks beyond isolated customers by trying to uncover networks of collaborating criminals. This is especially relevant as many money laundering schemes involve groups and networks rather than individuals. Visualization tools play an important part in these analyses. Amended with statistical network analysis, this can create a valuable detective approach to unravel the full scope of a financial crime network to which the FI is exposed.
Cognitive analytics leverages artificial intelligence and may help operations professionals to handle TM alerts, as well as to propose new AML countermeasures. Armed with the ability to process volumes of unstructured data (images, sounds, contracts, e-mails and manual handling) in seconds, it can make human handling efforts much more efficient. Ultimately, it derives meaning from data and generates hypotheses on the actual risk that only needs confirmation from a human expert. Although currently in their early days of evolution, these methods may very well transform the way FI’s handle their TM alerts to an unprecedented level.
Advanced Analytics for TM Multiple techniques with multiple purposes
The traditionally linear TM process will change to become an orchestrated combination of monitoring techniques. Most FIs are only just departing on their journey to enhance TM with advanced methods, with investments in proof of concepts (PoC) and isolated technology. However, it will become essential that FIs will develop a targeted strategy to reach the full potential of these advanced techniques and to make investments on the right priorities. Because it is only under such a strategy that TM effectiveness can be increased and maintained compliant, while long term costs are kept under control.
More information on Transaction Monitoring?
Do you want to know more on Transaction Monitoring? Please contact below.