Advanced analytics in transaction monitoring: how to make it happen


Advanced analytics in transaction monitoring: how to make it happen?

In this blog, we address the question of where Financial Institutions (FIs) can apply advanced analytics to existing Transaction Monitoring (TM) processes. Additionally, we identify key success factors to be taken into account when implementing advanced analytics in TM.

In our previous blogs, we wrote about the potential of advanced analytics for transaction monitoring (TM) and the technical conditions under which it can be applied successfully. In this last blog in a series of three, we make things practical and provide guidance on how to make it happen.

Embedding advanced analytics in the TM process

Advanced analytics can be applied in various ways in TM processes, but on a higher level we distinguish five key strategies:

  1. False positive reduction: traditional methods of TM, based on business rules commonly produce large numbers of false positive outcomes when trying to find the needle in the haystack. Predictive models, or machine learning, may be applied to calculate risk scores for each alert generated by such methods, in order to prioritize or even close business rule output without manual intervention. These models may initially be applied on top of traditional alert generation. In a more mature phase the models might replace the existing rule based alert generation for certain risk scenarios.
  2. Identification of new and specific risks: advanced analytics methods can be applied alongside more traditional methods to detect and monitor specific risks and behavior. Risks that are hard to capture with traditional methods may be tackled by applying more advanced monitoring. For example, monitoring of terrorist financing and other network related forms of financial crime can be executed in a more effective way by making use of pattern and network analytics rather than business rules only.
  3. Client segmentation: advanced analytics allow for sophisticated definition of segments of clients and their behavior. By using data clustering techniques (also referred to as unsupervised machine learning), subgroups (segments) of clients be identified based on a combination of client attributes and characteristics of their regular transaction behavior. This provides insight in the usual or common behavior. Subsequently, for each of those client segments, specific risk factors are analyzed based on historic data. This allows for a more tailored approach to the various client segments, which results in much more specific monitoring measures and therefore more relevant output.
  4. Offline performance evaluations: in evaluating the effectiveness of TM, advanced analytics methods such as feature selection processes and data visualization can help analyze performance and identify areas of improvement. In this case advanced analytics is not applied directly to monitoring processes (‘online’) but as a tool to analyze alert generation and handling processes (‘offline’) for weak spots or potential efficiency gains.
  5. Optimizing alert handling processes: lastly, advanced analytics can also be used to improve the handling of alerts. Similar to optimizing existing alert handling processes with Artificial Intelligence (AI) and Robotics, advanced analytics might enable alert handlers to perform their task in a more effective and efficient way. For example, enabling alert handling staff with a single holistic report presenting all relevant background information and risk factors of the client (based on extensive analysis of internal and external sources), could enable much faster and more accurate judgment of risks associated with TM alerts.

The strategies above are commonly applied within or from the existing TM frameworks, in order to make specific enhancements to this framework. Ultimately however, application of more advanced monitoring techniques could imply a paradigm shift and lead to further reduction of existing boundaries between ‘know your customer’ (client due diligence) and TM processes within FIs. In that scenario, a fully integrated and continuously updated view on the client and its (real time) behavior could enable fully integrated AML risk monitoring of the client.

Key success factors for implementation

While the execution of Proof of Concepts (PoC’s) to test the potential of advanced analytics can be organized relatively easy, implementing such methods within TM production processes provides specific challenges. Aside from the technical and regulatory preconditions discussed in our previous blog, we see the following operational challenges:

  1. Software: the well-known TM systems are being challenged by niche players advocating new and more advanced monitoring methods. Integrating new capabilities with existing systems is a common challenge FIs have to solve. Besides reconsidering existing vendor relationships, FIs might also explore creating bespoke solutions and processes to bring their integrated capabilities to the next level.
  2. Data: a classic obstruction in the application of advanced analytics is the availability of high quality data. Because the value of advanced analytics is strongly dependent on the availability of the right data, FIs need to consider making strategic and targeted investments in their data quality and data management structure.
  3. Process complexity: as monitoring techniques become increasingly complex, so do the associated business processes. This is true both from a risk management and operational perspective. Introducing advanced analytics methods may make answering trivial questions like ‘Do we cover all relevant risks?’ or ‘Why has this TM alert actually been generated for this client’ suddenly complex. It is therefore key for FIs to apply and maintain effective governance over all monitoring methods in place to ensure all risks are managed as intended. For more details on this topic, refer to our first blog in this series in which we state that FIs should make considerate decisions to apply techniques for a specific purpose and as such form an ‘orchestrated combination’ of monitoring measures.
  4. Knowledge: to fully utilize the potential of advanced analytics, it is not sufficient to just hire capable data scientists. Staff involved with TM, both from a risk management, compliance and operational perspective, will need to be educated on the working and interpretation of outcomes of the advanced analytics models. Only then will FIs be successful in integrating advanced analytics capabilities in the existing TM processes. This not only required to obtain the necessary support for these techniques within the organization, but also allows the organization to fully leverage and integrate the outcomes these models provide.
  5. Maintenance: bringing an advanced monitoring model into production mode can be challenging in itself, but keeping it updated and performing well is another one. Upon a first implementation cycle, FIs will need to form teams, procedures and processes to keep their models up to standards. As noted in our second blog of this series, multiple technical factors are critical for maintaining advanced monitoring models successfully.


The scenario in which application of advanced analytics within TM leads to a new paradigm in which client risks are continuously and holistically monitored by an advanced set of monitoring measures is very appealing to many professionals in the field. This is for clear reasons: such a design could greatly enhance both effectiveness and efficiency in the role that FIs play in combating financial crime and bring the field of Anti Money Laundering (AML) to an exciting next level. But before that effect materializes, multiple strategic, technical, regulatory and technical challenges have to be tackled. We hope our blogs may contribute to the journey that FIs are experiencing in making that happen.

More information on Transaction Monitoring?

Do you want to know more on Transaction Monitoring? Please contact Hilko Van Rooijen at +31882887771 & Martin Van Tienhoven at +31882887750

Did you find this useful?