Sanctions Screening – Threshold Setting

Key Challenges

SQA Consulting were engaged by a client to assist with the implementation of their new upgraded Screening Solution. In particular, they wanted advice and guidance concerning which screening threshold would work best for them.
Typically screening solution providers will provide a baseline threshold for their solution (if at all) but will require the client implementing the solution to assess this threshold and determine whether it needs to be amended to suit their risk appetite.

 

In theory, it is straightforward enough to run a sample customer file through the screening system at different thresholds. The likelihood is that at the different thresholds, different volumes of alerts will be generated – but what does this variance in the numbers of alerts generated mean?  Fewer alerts implies fewer false positives but at what cost? What types of names will not match at the higher threshold?

 

For example:
The impact of the matching capability in scenarios where there is British English spellings compared to US English spelling:
• ORGANIZATION FOR TECHNOLOGICAL INDUSTRIES V’s ORGANISATION FOR TECHNOLOGICAL INDUSTRIES
Or if numbers are replaced with the written word – will the name still match:
• Little Connemara 1 Farm V’s Little Connemara One Farm
Or for Personal Names, what will be the impact on the matching capability if there is a different middle name included:
• ABBUD AL-ZUMAR  V’s ABBUD Fred AL-ZUMAR

 

Multiple scenarios should be considered but these require a detailed test file and the knowledge to interpret the results.

SQA PROposed solution

SQA Consulting provides a Screening Assurance service based around our product The Screening Assessment Centre. The Screening Assessment Centre can create test files covering multiple scenarios for assessing the fuzzy matching capability for both personal and entity type names. Over 60 scenarios are considered by the system and the results are continuously benchmarked against peer FIs.

A single test file can be screened at multiple different thresholds and the results analysed and compared against each other as well as against the SQA screening benchmark.

solution benefits

The client was provided with a comprehensive report indicating how well the system performed at the various thresholds for both the false-positive rates and the fuzzy matching capability. The strengths and weaknesses in the fuzzy matching capability at the various thresholds were benchmarked across the different thresholds and also compared to other peer FIs included on the SQA benchmark. The Client was able to make an informed decision as to which threshold would be appropriate for their Sanctions Screening. The Graph below compares rates of Effectiveness and Efficiency for different Thresholds.

 


The green line and dots represent different screening thresholds.

Sample Conclusions and Graphs included in the report: In this example, the threshold of 75 proves effective at a rate of just under 70% but the associated level of false positives is high at 6.5%. The threshold of 91 realises a false positive ratio of approx. 2%. Whereas the threshold of 90 realises a false positive ratio of 4%, doubling the rate of false positives for a low improvement in effectiveness as the effective rate moves from 54% to just 56%. This example demonstrates the impact a minor change to the threshold can have.

A document such as this is exactly what a regulator is looking for. It demonstrates that there is a clear understanding of the adopted screening approach and that appropriate assurance steps have been taking as part of the implementation.
In addition, identifying the anticipated levels of false positives relative to the size of the customer base to be screened enabled the Client to set up adequate resourcing levels to process the alerts within their defined SLAs.

Technology Consulting Partners