Simon had a “false positive” problem. As the compliance chief of a fast growing financial services house, he was faced with a 14 day long queue of “flagged” transactions that the customer onboarding process had thrown up. Aside from the cost and risk involved in clearing this mound of transactions, there was also an angry CEO to contend with.
The queue meant delays in getting customers on boarded and transacting with the bank. Which meant a direct loss of revenue, fees and profits. Oh yes, the CEO was unhappy!
(Sounds familiar, compliance people?!)
Service providers were activated and emergency hiring done in the organisation’s capability centre in India. The target: get that queue down to 7 days within a quarter. Then, look to improve further from that point.
Simon also simultaneously embarked on a root cause analysis programme and realised that the challenge was both in the data quality as well as the software algorithms being used, most importantly, the name matching engine.
The programme examined the problem along 5 dimensions
- Diagnosing the problem
Simon chose to periodically review the output of their screening process, including hits and false positives (FP) . He was curious to understand where the FPs were coming from. He took a sample of 5000 transactions from his system and uploaded it to the @RZOLUT False Positive Analyser (www.falsepositiv.com) , a free tool that gave him the ability to pinpoint exactly what is driving the unacceptably high false positive rates.
The output of this exercise helped him immediately pinpoint the source of the problem and create an initiative to put a fix in place. In this instance, the single largest source of FPs was the name fields (and combinations thereof).
- Data Capture
Though it sounds simple, capturing data scientifically while wrestling with the tech debt of legacy platforms is complex. Simon recognized the need to capture customer data in a structured manner. For example, different fields for first, middle and last names. Provisioning for suffix titles such as “Jr” in the USA. This process would make the screening process against PEP, Sanctions, Watchlists and Enforcement data from providers such as @rzolut that much faster and more efficient.
He also realised that even the smallest data entry error would generate false positives, and given the volumes (millions of records), would create significant delays and costs.
- The Name Matching Algo
Simon realised that the efficacy of the name matching engine depends upon the quality of the underlying algorithm as well as the care with which it is configured and tuned for different scenarios. At @rzolut, we have built our Name Matching engine to account for 16 different combinations and variations of names (abbreviations, back to front, upside down and so on). We recognize that criminals use all kinds of tricks to evade detection and have built a world class engine so that our customers can rest easy during onboarding processes, or when they are doing their annual updates.
Having said that, it is key to remember that risk weighting is also important. Setting appropriate sensitivity levels at the time of screening and assigning appropriate risk weighting to different fields of the name are two key areas that customers must consider.
- Whitelisting as an efficiency tool
The easiest explanation for whitelisting is that it is the opposite of blacklisting! Once screened, Simon’s team started to create whitelists of targets that were otherwise recurring as false positives. This process can drive improvements in false positive reduction and up efficiency significantly.
However, a cautionary note is needed as whitelisting can also increase the risk in the opposite direction. In other words, they could cause you to potentially miss a ‘hit’. In case the target so missed happens to be a PEP, or is sanctioned, the organisation can get into hot water with the regulators. Use whitelists, but sparingly. Err on the side of caution.
- Leveraging the latest technology
Outdated screening software coupled with legacy data providers were hobbling Simon’s efforts for a long term fix to the false positive problem. He decided to take the bull by the horns and, with the support of the CFO and the CEO, embarked on a modernisation programme. In came a machine learning enabled Screening system and to it was married the latest, AI driven data platform. With new age scoring systems now possible, the journey towards a “zero false positives” world had truly begun!
Though Simon is a fictional character in a non-existent company, at @RZOLUT we are serving many real-world customers and helping them to build safer and more profitable businesses through the delivery of proven technology and the world’s most comprehensive and modern datasets. Do get in touch if our philosophy resonates with you and you’re looking for new partners to solve old problems.