You are here
The use cases reviewed as part of this research illustrate the value of data analytics for anti-money-laundering (AML) supervision and enforcement. For the purposes of this research, the term ‘data analytics’ refers to methods allowing users to turn data into knowledge that would not be revealed through a human review of the data in question. This includes traditional statistical methods and more recent developments relating to ‘big data’ or machine learning.
If used strategically and at scale, data analytics can support a dynamic and more precise assessment of money-laundering risks, and thereby enhance the efficiency and effectiveness of authorities’ AML efforts. This includes opportunities to monitor macroeconomic trends in real time, such as the evolution of financial flows in response to geopolitical or regulatory developments, or to more systematically analyse the financial activity of institutions or geographic areas to determine whether it is consistent with a range of contextual factors.
Notwithstanding the growing use of data analytics by AML authorities, the international dialogue on existing methods has been limited to date. This paper provides the first comparative study of the experience of national authorities in using data analytics for AML purposes and offers recommendations on how to maximise associated benefits.
The use cases reviewed for this research present a number of differences. First, they have been developed to serve different purposes. Some are primarily designed to detect or investigate criminal conduct, whereas others seek to enhance a sector’s risk-mapping as a basis for better allocation of supervisory resources. Second, the datasets underpinning the analysis vary in nature (data relating to transactions versus data on contextual factors), scope (data on suspicious transactions only versus data on larger sets of transactions) and level of detail (aggregate versus granular data), with the most effective tools generally relying on a consolidated review of several datasets. Third, the umbrella term ‘data analytics’ includes a range of different methods that have been successfully used or are currently being explored by authorities, including trend analysis, network analytics and, more recently, machine learning.
Considering the diversity of potential models, the choice of an analytical model should be informed by several contextual and operational factors. Contextual factors include each jurisdiction’s financial crime threat landscape, economy, data protection standards and non-AML policy objectives (such as facilitating business). Operational factors relate to the specific objectives and priorities of agencies, the availability of human and technological resources needed for the use of data analytics, and an agency’s pre-existing access to data.
Taking into account both contextual and operational factors, the use of data analytics should begin with a determination of the priority questions to be answered through the analysis and a clear understanding of how the results are relevant to AML efforts. For example, a method of analysis can be designed to determine if individuals use the services of several money service businesses to remit funds to the same destination, on the understanding that this behaviour indicates a higher risk of money laundering or terrorist financing.
As a next step, authorities should determine what data is needed to answer this question, whether such data is already available, and how additional data (if any) should be accessed, while keeping privacy considerations in mind. The selection of a dataset should be informed by consultations with other authorities, reporting entities and international partners. Irrespective of the specific method used, authorities need to develop a system to verify and ensure the quality of the data, including through routine verifications of consistency.
From the outset, each method should include an ethical framework clarifying how and for which purposes data may be used (for example, only for supervisory rather than investigative purposes, or only to support an ongoing investigation instead of revealing new leads), as well as a strategy for disseminating results to the intended end users. Finally, use cases should be subject to a periodic review of effectiveness, assessing if they meet the agreed objectives and how they contribute to overall AML efforts. This will inform any adjustments to the method and help policymakers assess whether resource and privacy implications are proportionate to the ultimate benefits.
In light of this, the paper concludes with the following recommendations:
- Supervisors, financial intelligence units and relevant law enforcement agencies should adopt a strategy to identify and harness opportunities for the use of data analytics in an AML context, taking into account available use cases and the factors discussed in this paper and in consultation with relevant private sector actors. This will provide a stronger basis for a transparent discussion with policymakers on necessary long-term investments and the balance with privacy considerations.
- Regional organisations such as the EU should foster a dialogue on the different approaches developed in this area by member states, to encourage a harmonised approach. Such coordination will reduce compliance costs for businesses operating in several countries and is likely to benefit the quality and usefulness of the data submitted under each of the national regimes.
- Assessment bodies such as the Financial Action Task Force and the IMF should more systematically review to what extent member countries have considered and harnessed the potential of data analytics for AML intelligence, supervision and enforcement purposes, which will allow for peer learning.
As the pace and complexity of the financial system continues to increase, the effectiveness of global AML efforts will be contingent on the readiness of governments to innovate in a strategic and coordinated manner.