Are you really concerned how your sanctions filter working?
The role and importance of strong sanctions controls cannot be underestimated. The challenge is multi-faceted: exploding data volumes; the constantly evolving sanctions landscape; differing customer, PEP and transaction screening requirements.
Let us discuss strategies to provide you peace of mind about the implementation and operation of your sanctions filter.
From regulators, Expectation is very high and till today Regulator is bit tolerant but tomorrow will not follow today as for as regulatory expectations are concerned. Now FIs really need to understand what Sanctions Filters are actually doing. So the black-box approach from vendors are no more valid, both FIs and vendors need to be open on their processes. So this is natural progression towards how well you understand requirements, how well you understand your data and how your resources are prepared & trained to use these filters.
It seems globally we are in same boat as for as quality standards for filtering settings are concerned. Additional pain we have in our side is you will find names of risk entities on sanction lists are embedded in our cities names, addresses, vessels and etc. So it becomes extremely difficult to deal with unstructured message filtering like SWIFT or private banking products.
In order to reach optimal settings for filtering, extensive testing, regression tests and model validations need to be done. Each cycle of data drops, make you learn lot of lessons related to:
Good guys list (which of course need to be monitored strictly with expiration criteria. Rule of thumb here is "do not assume a good guy will remain good guy forever" .
False Positive ratio (you need here learning algorithms that can adopt themselves to predict exact or near false-positives). Tagging each false positive with some standards reason to later help in data mining such as volume of transactions during Christmas, Year End and seasonal transactions varies culture to culture globally.
Segment the filtering criteria: Such as for USD traffic OFAC matters, for GBP/Euro EU list matters and UN sanctions are there for all currencies. Border areas to country requires stringent filtering settings compared to Center
Understand underlying mathematical model: Need to have grip on how underlying mathematical model is working. How it is treating misspellings, transpositions, initials, acronyms, synonyms, proximity.
Noise words dictionary (like bank, the, international and etc)
For instance: SEP is short for September and it is also risk entity in OFAC, EGP is Egyptian currency and also risk entity in OFAC. These are just few rats flush out of the woodpile of OFAC but do not forget you are in the forest of sanctions and it is growing. Latest trend is Panama papers leak & Russia… Who knows next except tomorrow's breaking news.
After going through multiple messages drops similar to one mentioned above, clients reach to optimal settings which depends also on external factors such as:
Input that you are providing to Sanctions filters need to be went through quality assurance steps. If it is difficult to go through this data quality cycle then there should be rules in sanctions filters which can accommodate your data weakness and still provide you a reasonable score to start investigation.
Good Case Management tool
Tool should be very much open to adapt to FIs workflow, provide comprehensive set of actions to make the investigation very much transparent in front of analysts. It should couple with document managements, can integrate cases from other solutions such as AML, Fraud & KYC if there is need to raise suspicious cases to authorities. It should be integrated with:
BO (Beneficial Owner) feeds [to identify ultimate Beneficial Owner – How long will you live with the excuse that UBO does not belong to your FIs?
Watch-list Management Interface
It is very crucial that how scanning algorithm deals with the risk entities found in different watch lists coming with different structures. Watch-list provider should also take one more step of quality assurance before publishing these Watch-lists.
Giving a sparse record with just two fields such as Last Name (Hassan) and then in Remarks field “Hassan is the son of Saddam Hussein's third wife and last time he was seen in Syria in 2006”. I believe such records will make sanctions filtering, a nightmare in the ME region. As I mentioned earlier contents of Watch-lists are very crucial for designing an efficient compliance program for sanctions filtering. Clients are reluctant to perform quality assurance on Watch-lists since they believe; this exercise may tamper the Watch-lists in such a way that in some scenarios the credibility can be lost. To the some extent, Clients are correct.
Second there should be a global common structure in XML for Watch-lists. All publishers should follow this structure strictly. Definitely it will streamline the performance of sanctions filters.
Compliance analyst head counts and training/awareness program
If you have these sanctions filter at various points in your FIs like in real-time it is integrated with payments, in ad-hoc it is integrated with compliance, legal department, credit cards, loans, risk department and in batch-mode it is integrated with your customer repository. Now these Watch-lists are updating frequently (OFAC on average has update two times in a week), so today’s bad guy is not necessarily be bad guy tomorrow or vice versa. So need to archive results of sanctions filters in some format where you can extract interesting trends related to financial crimes. Prepare golden data that you think that FIs will adopt future. Now point here is to validate whether your sanctions filter will work tomorrow for these upcoming changes. We know these changes are structural and are being adopted from other products. So it will be easy for you to portray interesting sets of trends related to data, sanctions filter results and exposing weaknesses and opening a channel for your sanctions filter to learn