How to reduce false positives in AML

As part of their regulatory compliance obligations, banks and other financial institutions (FIs) must develop and implement risk-based AML programs to counter the money laundering and terrorism financing threats they face. After performing risk assessments on their customers, firms should ensure suitable monitoring and screening measures are in place to alert them when high-risk customers engage in activity that could indicate money laundering.

However, an AML alert does not necessarily mean money laundering is taking place. The sensitivity of transaction monitoring and screening measures often means firms have to deal with false positive alerts when detecting genuine criminal activity. False positives constitute a significant proportion of the AML alerts generated by monitoring and screening measures, which means they can cause significant problems for firms by diminishing the speed and accuracy of their compliance processes. 

What is a false positive in AML?

A false positive occurs when AML software issues an alert for a transaction that seems suspicious but turns out to be legitimate. For example, a customer who makes several cash withdrawals from different bank branches on the same day may trigger an AML alert because these transactions are considered unusual – even if the customer has legitimate reasons for making them. Likewise, variations in global naming conventions can result in customers being incorrectly flagged during sanctions screening processes. 

An FI’s AML compliance team would need to investigate both of these cases due to the alert – even if they are ultimately revealed to be false positives. 

How do false positives happen?

There are several reasons why false positives can happen, particularly if institutions have not taken all available steps to reduce their likelihood:

• Sensitive rules: If the rules that institutions use to detect suspicious behavior are too broad or sensitive, they may generate false positives by flagging legitimate transactions that meet their criteria.
• Context: Systems or analysts may lack the context to assess a transaction accurately. For example, a large one-time transaction from a new customer may trigger an alert, even if it’s legitimate.
• Inaccurate data: If data is incomplete, inaccurate, or outdated, the chance of false positives occurring becomes greater. 
• Human error: Manual review processes conducted by compliance officers can also lead to false positives. Transactions can be wrongly classed as suspicious if data is misinterpreted.

The role of technology and talent in transaction screening

Get actionable insights on optimizing screening processes to manamage financial crime risk.

Download your copy

The impact of false positives on financial institutions

False positives represent a challenge for FIs around the world for several reasons. They can: 

• Create negative customer experiences: Customers involved in transactions wrongly identified as suspicious will likely be inconvenienced, possibly through having their accounts frozen and being unable to manage their finances as normal for some time. Repeated on a large enough scale, this may lead to firms losing a significant number of customers.
• Slow down business growth: By causing firms to waste time and resources investigating alerts that turn out to be legitimate, false positives drive up the overall cost of compliance and threaten to restrict business growth. 
• Hinder firms’ efforts to address genuine money laundering threats: Given the administrative work that false positives generate, they can lead to damaging delays in firms reporting actual suspicious behavior. This is particularly true if false positives are wrongly escalated and reported to financial authorities. This not only means that criminal activity occurs undetected in the global fight against financial crime but can lead to overspending on compliance and regulatory consequences for institutions. 

While false positives can represent significant cost and efficiency drains, firms have little room to reduce the sensitivity of their screening and monitoring measures, given the financial, reputational, and even criminal penalties resulting from regulatory non-compliance. Finding a way to reduce false positives without compromising customer screening and transaction monitoring should be a priority for every compliance team. 

Four ways to reduce false positives in AML

While it is not possible to eliminate false positives, given the strict regulatory obligations firms have to meet, they can be reduced as much as possible with a few important measures: 

• Improve data quality 

Transaction monitoring and screening measures require firms to process and analyze a vast amount of data from various sources, often in an unstructured format. The more confusing the data collected as part of procedural compliance, the more challenging it is to discern false positives from true positive AML alerts. 

By carefully considering the data capture process and structuring that data clearly once acquired, firms can improve their false positive rate. Practically, one example would be organizing names into title, first name, and surname rather than listing each name as a single data point. The latter approach can create administrative ambiguity and hinder compliance teams’ attempts to resolve customer identities after AML alerts. To further clean data, duplicated or irrelevant data points should be removed, errors identified and corrected, and any missing data addressed. 

Criminals may seek to change their names or move between countries to thwart AML controls. With that in mind, while the quantity of AML data captured is important to building an accurate customer profile, the data’s relevance and recency are crucial to the verification process. For example, a false positive might be created if a firm fails to update a customer’s change of name or residence from a high-risk jurisdiction to a low-risk jurisdiction. The ability to update data in as close to real-time as possible represents a significant advantage to firms here. 

• Establish a risk-based approach

As with other elements of AML compliance programs, a risk-based approach is crucial in reducing false positives. Taking a blanket or one-size-fits-all approach to screening and monitoring is highly inefficient and likely to inflate false positive rates through unnecessarily broad or undeveloped rules used to identify suspicious transactions. Clear risk parameters based on contemporary money laundering typologies should be used while proceeding from an initial business-wide risk assessment. Rules should also be formed specifically with a given institution’s risk profile, regulatory requirements, and services or products in mind. 

• Conduct ongoing audits

An AML compliance program should evolve according to new regulations or the discovery of new criminal methods, not be a simple exercise in checking boxes. Accordingly, firms should conduct ongoing reviews of their screening and monitoring measures to ensure their continued accuracy. As regulations change and new financial crime patterns replace old ones, firms should ensure their place on the leading edge of AML/CFT and maximize their efficiency by adapting their compliance measures. 

• Choose a suitable AML software solution  

Smart technology is a crucial component of the modern financial compliance process. With the benefit of integrated artificial intelligence (AI) and machine learning models, firms can increase the efficiency and accuracy of their AML programs, building richer and more databases for use in risk assessments. 

While AI and machine learning models shouldn’t be seen as a replacement for human expertise within the AML process, they can help compliance teams address false positive alerts in a more structured and efficient way. Optimizing AI and machine learning integration of AI should become a foundational part of the AML process for financial institutions and an essential step towards reducing a firm’s false positive rate.

Reduce false positives with efficient AML solutions

ComplyAdvantage’s AI-powered matching algorithms screen new customers for sanctions, watchlists, adverse media, enforcement data, and PEP and RCA lists, drawing on dynamically updated databases to minimize false positives caused by inaccurate data. Our intuitive screening processes are also adapted to global customer naming conventions that could otherwise generate false positive hits on sanctions lists. At the same time, our ML models enable firms to perform semantic and statistical analyses on new alerts, quickly identifying duplicate or redundant data that might be creating the appearance of suspicious behavior. 

Our combination of consolidated data, real-time updates, and smart insights driven by natural language processing empowers FIs to make informed decisions and lower their false positive rate.