Minimising False Positives in Sanctions Screening

The occurrence of false positives while conducting the sanctions screening process is not unheard of. In fact, occasional false positives indicate that the sanctions screening mechanism is working just fine. Nevertheless, high false positive rates demand attention as they indicate the existence of certain lapses in the implementation of the sanctions screening process. This blog discusses, at length, the causes behind high false positive rates, the consequences of high false positive rates, and strategies that enable businesses to minimise the instances of false positives in the sanctions screening process.

Minimising False Positives in Sanctions Screening

The sanctions screening process helps businesses to identify if any of their prospective or existing customers, whether natural persons or legal entities, are:

• Sanctioned by any of the watchlists or sanctions lists published by a relevant regulator of any country or worldwide organisation, such as the United Nations
• Politically Exposed Persons (PEPs)
• Appear in any negative media, indicating involvement in a criminal activity

The process of conducting sanctions screening includes the following:

1. Collecting Key-Identifier Details from the Customer
2. Entering Customer Details into Sanctions Screening Software
3. Generation of Sanctions Screening Outcomes
4. Screening Outcome Analysis by Disambiguation of Matches
5. Disambiguated Matches can be Classified as:
   • Perfect Match: A perfect match is such a match result where the screening analyst conclusively confirms that through the screening process, they have found a profile of an individual or entity that completely, without a doubt, matches that of their customer’s profile. In a perfect match, the key identifier details of a customer match completely and perfectly with that of the profile found during screening.
   • Partial Match: A partial match is such a match result where customer profile details do not completely match with the profiles found during screening. In a partial match scenario, only certain parameters of a customer profile match with those of screening results, leaving the screening analyst unable to conclusively decide whether the match is a perfect match or a false match.
   • False Match or False Positive: A false match or false positive is when the screening exercise, either through software or sanctions list manual search, generates results, but none of those screening results match completely or partially with the customer profile. In case of a false positive, the screening analyst is sure that screening results generated through the screening exercise are false positives and none of the key identifier details of the customer correspond with the screening results derived.

• • No Match: A no match is a situation in which the screening exercise generates no result, and the number of potential matches found is zero.

What Are False Positives in Sanctions Screening?

• False Positives: Definition

False positives are sanctions screening disambiguation outcomes. They are screening results that initially give the impression that the screening match is probably related to that of the customer who is being screened, but upon analysis and disambiguation, the screening analyst decides that such a match is a false positive generated during the screening exercise.

• Difference Between False Positives, True Positives, and False Negatives

False Positives	True Positives	False Negatives
The screening outcome is false positive if, upon disambiguation, it is found that it has a similarity with the customer profile on a superficial level, and upon further analysis, the customer profile details do not match at all with the screening outcome.	The screening outcome is true positive if, upon disambiguation, it is found that a particular screening outcome completely matches with that of the customer on all parameters of key identifier details.	The screening outcome is said to be false negative if, upon disambiguation, it is found that a particular screening outcome, which initially was shown as negative or not matching with that of the customer profile, matches with all parameters of the customer profile.

Causes of False Positives in Sanctions Screening

Conducting name screening generates a wide range of results based on parameters set by the screening entity or screening analyst. Out of the screening results derived, false positives are bound to be found, indicating that screening software or screening mechanism is functional and operational to the extent that it generates results based on inputs supplied.

Nevertheless, a high number of false positives generated during screening is a cause for concern, and a high number of false positives could be the result of the following factors:

Similar Names:

• Running sanctions screening against customers whose names are common would naturally lead to having screening results with a high number of false positives.
• Even for customer names that are not so common, false positives are generated through a screening process; this happens because the name has a phonetic sound similar to that of the customer’s name, giving false positives.
• Spelling variation is also a factor in the generation of false positives.
• Sanctions screening tools often offer algorithms that enable fuzzy matching to avoid missing out on probable spelling and phonetic variations, which in turn, due to the setting up of low match thresholds, ends up generating a high number of false positives due to similarity in the name of the customer and those appearing in sanctions lists or watchlists.

Data Issues:

• There is always a possibility that the name screening tool generates false positives due to issues with the data it is accessing.
• False positives may also end up being generated due to substandard data quality, errors in rule setting, or uneven data or information patterns.
• Sanctions screening software runs through setting certain assumptions when executing the screening process. If data assumption rules are faulty, then screening results might generate a high number of false positives.

Lack of Contextual Information:

• Contextual information refers to any materially important information that impacts its users and their decision-making. With regards to sanctions screening, materially important or contextual information means customer information.
• When there is a lack of adequate and relevant customer information, such as their key identifier details, such as
  • last known location,
  • nationality,
  • date of birth,
  • gender,and so on, the screening analyst, to cast a wider net to avoid missing out on potential matches during screening, may set low match percentage thresholds, resulting in a large number of potential matches, where most of them would end up being false positives due to lack of contextual information. Having sufficient contextual information would enable screening analysts to set low match threshold limits.

• Lack of knowledge of cultural nuances which impact name spelling and script writing styles, such as Arabic or Cyrillic languages, also cause false positives.

Algorithmic Errors:

• Sometimes, a high number of false positives are generated due to errors in the screening tool algorithm. Algorithms also have their limitations in accurately interpreting subtle differences in any name spelling and considering language differences.
• Algorithmic error might result in identifying an obvious false match into a potential match, which, upon disambiguation, the screening analyst deems to be a false positive.
• This happens when algorithmic programs fail to identify the subtleness of cultural, phonetic, or spelling context in data available for screening.

What Are the Consequences of High False Positive Rates?

When screening outcome disambiguation leads to a high number of false positives, certain consequences are bound to follow, such as:

• Increased Workload: The screening exercise generates a high number of screening outcomes, which would lead to requiring disambiguation of each outcome. The disambiguation of all match results, particularly false positives, ends up consuming substantial amounts of time, increasing the workload of the screening analyst, which in the long run is resource-intensive on the organisation.
• Delays in Customer Onboarding: Until all screening matches are appropriately disambiguated and the fate of the screening exercise determined, the business cannot onboard a customer. The business cannot proceed with customer onboarding as knowledge of the customer’s sanction screening outcome is necessary to decide whether or not to onboard the customer.
• Delays in Concluding Business Transactions: Businesses are required to screen their customers on an ongoing basis throughout the business relationship. When a high rate of false positives is generated during the ongoing screening, it is obvious that the preceding process of match disambiguation would have consumed substantial amounts of time, leading to delays in concluding business transactions.
• Inefficient Screening Processes: When high false positives are usual in any organisation’s sanction screening process, it’s a clear indicator of the screening process being inefficient. Such inefficiency would be the clear outcome of a lack of fine-tuning the screening tool algorithms and various internal issues, such as a lack of close-match threshold settings.
• Fines and Penalties: When high false positives in sanctions screening are a norm, it is possible that a screening analyst mistakenly disambiguates true match as false match, resulting in missing a sanctioned individual and onboarding such individual without regulatory reporting. This lack of reporting of sanctioned individuals or entities leads to regulatory breaches, attracting fines and penalties.
• Reputational Damage: Attracting fines and penalties for regulatory breaches arising out of non-reporting of sanctioned individuals and entities in a timely manner due to high false positive rates would lead to reputational damage to the organisation.
• Obstruction of Real Threats: When a high number of false positives occur, the sheer volume of matches for disambiguation would be so high that it might obstruct real threats or the true match identification process.

Strategies for Minimising False Positives in Name Screening

Occasional high false positive rates are an indicator that some amount of fine-tuning needs to be done with sanctions screening tools or mechanisms. Nevertheless, frequent high false positive rates are cause for concern as their consequences act as impediments to business growth. Here are a few strategies that businesses can adopt to reduce false positives in the name screening process:

Ensuring Data Quality

• Data completeness while conducting name screening, as well as data quality of customer information, both must be up-to-the-mark and complete to aid with an accurate name screening process.
• Data preprocessing should be carried out by a business to standardise the available customer data into the ingestible format, which means that customer details should be organised in a manner that, when entered into the sanctions screening tool, generates accurate results, reducing false positives.
• Having the right approach to customer data uniformly across the business would help in reducing false positives when data quality is consolidated and accessible in alignment with the requirements of the sanctions screening tool.

Case Management

• Relying on case management tools, where centralised tracking along with Know Your Customer (KYC) and Know Your Business (KYB) processes are integrated with screening tools, would help reduce false positives as substantial information about the customer would be made available to the screening analyst. This would help them to set a low match type percentage or threshold in the screening tool and cast a precise net while screening customer details across screening lists.
• Most case management tools offer reconfigurability and customisability to suit individual business needs; leveraging a case management tool helps streamline the entire Anti-Money Laundering (AML) compliance process for any organisation.
• The case management tools with sanctions screening functionality come with a multifunctional User Interface (UI), enabling users to pivot from the name screening tab to the KYC tab to the Customer Relationship Management (CRM) tab seamlessly, enabling them to harmonise customer data available across various tabs to fine-tune and calibrate sanctions screening database and match percentage or threshold settings to reduce false matches.

Utilising Machine Learning and Artificial Intelligence

• The use of machine learning helps the user to train the tool to absorb data entered and give an output, accordingly, coupled with Artificial Intelligence (AI), which helps the user to automate repetitive human inputs which can be broken down into series of steps helps businesses to reduce false positives.
• AI and Machine Learning can be deployed by businesses to train their screening tool to consider various aspects, such as nuances of phonetic and spelling variations, similar names, and gender considerations, to name a few, prior to giving a match result, reducing the chances of false positives by eliminating obvious false positives.

Algorithm Fine-Tuning

• Usually, a high number of false positives is an indicator calling for algorithm fine-tuning. Businesses can reduce instances of false matches by calibrating their screening tools according to the type and range of screening data they require to screen their customers.
• Algorithmic fine-tuning should also consider proximity matching to avoid false positives arising due to location-based differences.
• While fine-tuning the algorithm for sanctions screening outcomes, setting up of dynamic fuzzy matching must also be considered as static or stagnant fuzzy matching would be unable to detect subsequent changes in customer information or sanctions screening database updating regarding listing and de-listing of names on the said sanctions list.
• While conducting algorithm fine-tuning, machine learning can be deployed to remove duplicate matches automatically using AI. This would tremendously reduce the de-duplication workload on the screening analyst.
• Lastly, the algorithm fine-tuning must be tested and validated using sample data to figure out if any discrepancies arise and ensure that the occurrence of false positives has reduced significantly.

Contextual Clarity

• When setting up sanctions screening tools according to business needs, it is important to understand the search data the entity would have to work with; understanding the type of sanctions list would help the screening analyst of the entity to develop contextual clarity, which would help reduce false positives when configuring sanctions screening tool by understanding the qualitative and quantitative aspects of search data.
• Sometimes, simply whitelisting frequently occurring false positives is a simple solution to avoid their repetition during ongoing screening. However, the whitelisting method must be used sparingly as it might result in obstructing true matches in future.
• Businesses need to develop contextual clarity regarding relevant and applicable sanctions list selections; screening customers across wrong or redundant sanctions lists would not give accurate results.
• Rescreening of customers with whom business relationships are continuous in nature must be ensured; there must be clear internal policies addressing the re-screening of customers with whom business relationships are occasional in nature; this would help in reducing false positives.
• Contextual and conceptual clarity regarding modelling customer data available according to its relevance and usage according to screening requirements would aid with reducing false positives.

Adopting a Risk-Based Approach (RBA)

• Adopting RBA, in simple words, means deploying measures or adopting practices in proportion to the financial crime risk faced by the business from various factors, such as customers, geographies, delivery channels, technologies used, and the mode of transactions, to name a few.
• Adoption of RBA would help in tuning the sanctions screening console according to the business’s due diligence needs, reducing overall false positives by focused and necessary screening only.
• Search filter calibration according to the geographical risk shall also help in minimising false positives.
• The screening tool should be periodically re-calibrated to conduct a weighted search and change in overall risk faced by business upon introduction of new products or services, based on the degree of risk posed by customers from each type of geography and sanctions list, the screening tool, for effective implementation of RBA, must have re-tuning and re-calibrating ability.
• The fundamentals of RBA require the selection of only sanctions lists which are relevant and applicable to a business, and accordingly, API integration of relevant sanctions lists must be ensured to minimise false positives.
• Businesses, upon adopting RBA and having a clear idea as to the causes of false positives, can also set automated clearance rules using AI and machine learning to resolve false positives. However, they should use this feature sparingly and with utmost caution while ensuring that true positives are not obstructed while minimising false positives.

Staff Training

• Organising and conducting staff training, particularly for those employees who are entrusted with the sanctions screening process, would help enhance their knowledge, which in the long run would help reduce false positives.

Human Intervention

• Businesses must understand that technology is no replacement for human input but merely a tool to simplify, streamline and automate repetitive human tasks. With human input, the name screening tool can be incorporated with noise word elimination, meaning that screening analysts or compliance personnel having experience in sanctions screening would know which words are mere ‘noise’ among volumes of data they handle every day and input such noise words into the screening tool, which would help in generating fewer false positives by being trained by human input to tune out noise words.
• Dynamic fuzzy is one such thing that can be improvised and fine-tuned only with the help of research and input by humans who have experience with the input and the desired outcome. Using human input to improvise and keep dynamic fuzzy up to date will help reduce false positives.
• It takes a keen human eye with an analytical and empathic mindset to aptly construct a customer profile and configure its attributes in a manner which can be used to conduct sanctions screening; this careful construction of a customer profile and tuning sanctions screening tool according to RBA, ultimately helps in achieving lesser false positives.

Continuous Monitoring and Improvement

• The continuous and dynamic updating of the sanctions screening tool according to changes and updates in sanctions lists helps in keeping the sanctions screening exercise in alignment with the RBA.
• Improvement measures such as doing away with redundant or obsolete sanctions lists, re-tuning match percentage thresholds that are no longer effective, and improving algorithmic settings help in reducing instances of false positives.

When all factors contributing to better screening results are implemented, the disambiguation of matches derived would definitely lead to generating fewer false positives.

Conclusion

Businesses relying on name screening software to meet their compliance requirements are often faced with the challenge of high false positive rates while disambiguating screening matches.

They must closely examine the causes of high false positive rates and attempt to understand the nuances involved while implementing strategies to minimise the occurrence of false positives. They can reduce instances of false positives by relying on technology such as AI and Machine Learning and make sure that they adopt RBA when conducting the name screening process.