Surging False Positives: The Problem of Finding the Needle in a Haystack
Thursday, October 5, 2017
Posted by: Brian Monroe
By Deepak Amirtha Raj
A research and strategy analyst specializing in compliance analytics through artificial intelligence, virtual reality and augmented reality.
October 5, 2017
Banks are prevented from allowing any illicit funds onto their accounts, or allowing criminal entities to access the broader international financial system through their physical or online portals.
Even so, in the recent past regulators have imposed heavy fines on financial institutions that did not implement proper controls to adequately monitor transactions to ensure they are not carried out by organized criminal or terrorist organizations.
This scenario has impacted the risk appetite of financial institutions when it comes to onboarding customers.
At the same time, with the European Union’s 4th Anti-Money Laundering Directive (4AMLD) coming into force across much of the bloc, the demands of anti-money laundering (AML) have also increased, with institutions required to perform more extensive KYC compliance processes to screen both individuals and corporations.
Large international banks must also screen current and potential customers – and any transaction passing through their institutions – against a bevy of blacklists to ensure they have not been designated by the United Nations, United States or other major economies as being tied to terror groups, criminal operations or are proxies for rogue nation stations.
Not surprisingly, these lists are fraught with challenges, from people having the same names, to geographic and language hurdles, to something as simple as the analysts reviewing the supposed screening hits not being familiar with jurisdictional quirks and customs.
In some cases, the real entities are hidden behind murky corporate shell companies or relatives and associates acting in their stead, further complicating an already perilous exercise.
So in a bid to better parse out the more risky individuals, financial Institutions conduct screening of customers and transactions against a long list of data points.
The chance of a near match or mismatch, called false positives, must be investigated and eliminated, and the process is complex due to the mechanics of screening data – it is like finding the needle in a haystack.
Or as one financial crime compliance professional once said: finding a needle in a stack of needles.
If you ask AML professionals what cause them the most headaches, the issue of “false positives” will no doubt be at the top of the list, particularly when it involves customer name screening and adverse news research. Each of these crucial processes runs into problems caused by false positives.
The result is perhaps inevitable in that these daily screenings produce a large number of false positives, or alarms that flag an issue that must be investigated.
In fact, a recent research report by AML technology firm Fortytwo Data, it states that banks are spending £2.7 billion per year investigating false positives due to outdated AML systems.
At the same time, spending by financial institutions and firms operating in other AML compliance-regulated industries is projected to increase to £6.4 billion in 2017 on a global scale and reach its peak at around 2020.
Investigating through each alert is time-consuming and complex at an average cost of £20 each time, cumulatively creating a hefty bill when a bank has millions of customers to screen.
Financial crime compliance departments have steadily increased in size and there are larger financial institutions that report their annual expenditure for AML screening alone exceeds £3billion – and that figure doesn’t even include voluntary or regulator-ordered remediation engagements to find any missed sanctions hits or suspicious activities.
However, this can be reduced by streamlining compliance process and implementing solutions with advanced matching algorithms. Here are some of the methodologies to reduce false positives:
1) You can use a partial matching algorithm which retrieves all the records that are similar to that on the watch list. Records show these types of relationships when few elements of the first record match few elements of the second record.
2) Phonetic Matching is another method which uses fuzzy logic algorithms to match records based on how they are pronounced rather than how they are spelled.
3) Another emerging method is to teach an intelligent machine to automatically learn from previously discounted transactions to refine the entire process, altering scoring criteria automatically to reduce false positives.
4) A search using your customer’s biometrics over a biometrically-enabled list of sanctioned individuals could also dramatically decrease false positives.
5) Blockchain and DLT can be used to create digital identities for the customers. This digital identity would store all relevant information about the customer from addresses, account details, director’s details, PEPs etc., which could be used during AML/transaction monitoring, thus increasing the accuracy of the monitoring and reducing the likelihood of false positives.
Financial Institutions should start tapping the potential of the Machine Learning, Artificial Intelligence and Blockchain technologies to mitigate false positives and adhere to the regulatory expectations at a reduced cost.
About the author:
Deepak Amirtha Raj is a Research & Strategy Analyst in the Risk and Compliance sector. He focusses on Business Strategy Research, Emerging Technologies and Advanced Analytics. He studied business at Saint Joseph’s College and had previously worked with Royal Bank of Scotland as Business Process Analyst.