Payment Screening in Banks: Managing the Dilemma of Cost vs Risk and how Machine Learning Helps

Inevitable trade-off between cost and risk, effectiveness and efficiency

Financial institutions find themselves caught between the conflicting priorities of effectiveness and efficiency. This means:

• finding as many risky transactions as possible (effectiveness = low risk but high costs)
• finding as few transactions as possible that, on closer inspection, turn out to be innocuous (efficiency = low costs but high numbers of hits).

Why is payment screening so important in the financial industry?

Financial institutions are under extreme pressure to comply with regulatory frameworks. The work of compliance departments is dominated by the need to prevent financial crime. Payment screening has a vital role to play in this respect. It is a key component of risk management that protects financial institutions from legal consequences, penalties and damage to their reputation.

Real-time payments and other digital payment options mean that risk decisions have to be made faster than ever before. Banks have to strike a balance between due diligence, customer expectations and the cost situation.

How important is efficiency and effectiveness in payment screening, and what exactly is the difference?

During the screening process, the compliance software checks payment data against sanctions lists, embargo lists and other blacklists. This data matching is based on criteria such as first and last name, company name, alias name, alternative spellings, countries involved, banks, BICs, accounts, amount, keywords and whitelist exceptions. Financial institutions are under increasing pressure to ensure their systems are both efficient and effective.

Effectiveness: How good is the system at identifying risky payments?

If a hit, i.e. a potential match to a sanctions list entry, is discovered during payment screening, this process has to be clarified manually. Banks can’t afford to miss anything – failure to spot a suspicious payment could entail serious legal and financial consequences.

Banks usually carry out effectiveness tests every year using predefined test cases. These check whether the system is identifying the hits it has to find in order to ensure compliance. These tests are usually done using test data rather than real data to avoid confidentiality issues.

They are normally conducted by independent third parties such as management consultancies and auditing firms. They use benchmarks from several institutions to compare a bank’s payment screening performance against that of their peers.

Efficiency: How accurate is a hit? Is it a false positive or a true positive?

The efficiency test examines how much work a hit generates when screening a payment. Every hit has to be reviewed by a member of the compliance team, who decides whether the name that appears on the payment is actually the person on the sanctions list. The percentage of true positives, i.e. hits that represent a real threat, should be as high as possible. This is because unnecessary hits (false positives) increase the workload of compliance staff without posing an actual risk. The efficiency of a system depends on finding the fewest possible false positives.

The fact that sanctions lists are growing longer and banks are regularly screening transactions increases the likelihood of false positives, for example when customers have an identical or similar name. Processing every hit requires time and money, which is why banks are working hard to improve the efficiency of their payment screening systems.

Striking the right balance between effectiveness and efficiency – a tricky business

Achieving the twin goals of effectiveness and efficiency in payment screening is rarely easy because it is always a trade-off between risk and cost. On the one hand, the system has to identify every risky transaction for compliance reasons. But on the other, banks want to save money by handling as few clarifications as possible.

Machine Learning is the best solution to this dilemma: The best possible blend of effectiveness and efficiency is achieved by combining different matching algorithms with Machine Learning. Find out more in the whitepaper.

Why compliance teams are feeling stressed

All over the world, financial institutions are investing in compliance staff to investigate the growing number of suspicious cases – but this comes at a price. Costs are soaring, and so is the danger of making wrong decisions. A study by Lexis Nexis reveals that the projected total cost of financial crime compliance across the global finance sector amounts to around $180.9 billion. According to the study, Europe and the US are hardest hit by financial compliance costs because they have significantly higher numbers of financial institutions. On average, the global distribution of compliance costs stands at 57 percent for labour, 40 percent for technology and 3 percent for other factors. The study also finds that compliance teams feel stressed, and 67 percent of compliance decision-makers are concerned about the job satisfaction of their workforce.

The best solution: Using machine learning in payment screening

Customers are increasingly communicating with their banks via digital channels, so they expect to have a fast and secure digital experience. This means that the payment screening system must be powerful enough to handle this shift in consumer behaviour and deliver the speed and reliability that is needed.

Fuzzy matching makes it possible to analyze transactions more effectively. It also identifies anomalies in the event that names are reversed, falsified or misspelled, a letter is added, abbreviations are used, etc. But there is another side to the coin: a fuzzy search always produces more hits than an exact search. Therefore, the screening system has to be intelligent enough to flag only risky payments (true positives) and keep false positives to a minimum in order to avoid unnecessary clarification costs.