The second installment of our sanctions compliance blog series looks at ways organisations can achieve greater efficiency based on our whitepaper, Excellence in Sanctions Compliance: The Role of Effectiveness, Efficiency and Explainability.
The first blog in our “Excellence in Sanctions Compliance” series detailed the many regulatory requirements that must be met to achieve effective compliance. If not managed optimally, those requirements can translate into heavy costs for financial institutions.
Efficiency, on the other hand, is all about allocating the right resources, while optimising results and maintaining effectiveness.
According to recent research we conducted with analyst firm Chartis Research, a third of respondents identified operational and resource constraints as the number one factor influencing their approach to sanctions screening. Efficiency can be improved by streamlining operational processes to reduce the headcount required, tuning models to minimise false positives, and automating tasks like list updates.
Efficiency is critical in supporting effectiveness, reducing costs, and mitigating the risk of fines and reputational damage. For example, in late 2018, a Gulf region bank was placed under a consent order by New York State Department of Financial Services (DFS), in part because it could not keep up with alerts and Suspicious Activity Reports (SARs) that needed to be filed. The bank was receiving over 1,500 alerts per month and, due to lack of efficient software and processes, it was filing SARs very late.
The staggering cost of fines – $1.3 billion as of October 2019 from OFAC alone – are a strong reminder that sanctions compliance lapses pose a significant risk. Efficiency improvements can play a major role in reducing operational risk by limiting the need for manual processes that are inherently error prone.
Improving the efficiency of an organisation’s screening filter can be accomplished by leveraging artificial intelligence (AI) techniques to reduce false positives, while ensuring no true positives go under the radar. This is a big efficiency win if the hit rate can be reduced to single digits, dramatically reducing the number of alerts that need to be assessed.
There are many different types of AI, but the type most commonly used in sanctions screening is machine learning. With machine learning, large amounts of data are made available to an algorithm tagged to designate true positives, false positives, true negatives and false negatives.
The algorithm uses this data to connect attributes to outcomes. When new data is fed into the system, the algorithm assesses the data based on the training data (with known outcomes) and produces an output, which could be no alert, or an alert with a score and some descriptive elements.
If the inputs to the model are not good, the outputs will not be good either. It is important that the watchlists being referenced are frequently updated and parsed so they can be matched effectively without creating low-quality alerts.
The goal when applying AI to the filtering process is to reduce the number of false positives requiring manual review. This has several operational merits:
Once the filter has been applied, the output is a number of potential matches. These “hits” need to be reviewed and decisions made, generally by a human analyst. By augmenting human and artificial intelligence in this process, there are significant opportunities to achieve far greater efficiency.
Sanctions data changes constantly and organisations need to ensure they keep up.
Efficient operational processes support the effectiveness of the sanctions screening programme by improving the data available to analysts, increasing the speed at which alerts can be reviewed, and increasing how quickly models can be maintained and updated.
With better data, analysts can spend less time gathering information from different systems and there is a smaller margin for error when documenting decisions on alerts. Better data could include transaction information, identity elements (such as date of birth and passport information), and detailed reason codes and match descriptions. Better data also provides for full-featured audit trails.
Pulling together all the relevant information allows analysts to review, document, and make decisions on alerts faster. This increases the overall throughput of analysts and means fewer are needed to manage the volume. This does not necessarily equate to a reduction in headcount, but allows for resources to be reallocated to optimise effectiveness.
If models aren’t assessed and kept up to date, they may drift over time and begin to produce false positives. This makes them less effective and can increase the alert rate, requiring more time from the compliance team to review them. To the extent possible, model maintenance should be automated and provide actionable intelligence to fine tune the system.
Regular testing of the model and any changes before and after implementation can help to ensure models stay on track. Regular testing also supports compliance with New York State’s Department of Financial Services Part 504 regulation. Being able to quickly provide model test and validation documentation to auditors and regulators increases the efficiency of the examination process. Testing can help identify new opportunities to refine the screening process, by finding greater efficiencies in the system and reducing false positives.
Having processes and procedures to perform regular model testing and validation means performing these tasks can be targeted, less time consuming and a regular part of maintaining model health.
There are many ways to improve efficiency in a sanctions compliance programme. Some, like reducing false positives are fairly clear. Others, such as finding efficiencies in the model testing process, are somewhat less obvious.
Join our webinar or download our white paper, Excellence in Sanctions Screening: The Role of Effectiveness, Efficiency and Explainability to learn more.