Emarketer projected that e-commerce is set to grow another 13.7% in 2021, hitting $908 billion in e-commerce sales. While e-Commerce is enjoying an unstoppable force of growth, fraud is also quickly gaining ground.
Juniper Research released a study, “Online Payment Fraud: Emerging Threats, Segment Analysis & Market Forecasts 2021-2025,” where they estimated that the losses due to e-Commerce fraud will rise 18% to more than $20 billion by the end of 2021. Up from last year’s $17.5 billion.
When The Solution Is Worse Than The Problem
To counterattack these threats, merchants have implemented solutions such as compliance software like anti-money laundering and countering the financing of terrorism (AML/CFT) programs.
However, these programs are known to increase the risk of boosting the number of false-positive alerts. This leads to “misidentifying customers” and erroneously flagging transactions as money laundering or terrorist financing risks.
False positives are certainly not an issue to be ignored. According to research from Forter, a leader in e-Commerce fraud prevention, retailers stand to lose 75 times more revenue to false positives than they do to fraud itself.
Further research suggests that as much as 15% of card-not-present (CNP) transactions are “falsely flagged as fraudulent”, leading to a yearly revenue loss of $118 billion.
Reduce The Risk Of False Positives Through AI and ML Solutions
Machine learning is a particularly useful AML tool due to its data collection requirements of “false positive remediation.” Furthermore, it takes shape based on the speed and efficiency of the AI compliance software. Simultaneously, it informs the remediation process with the depth of added information.
Machine learning systems use historical data to make annotations about customers and their buying behavior over a period of time. Based on that data, they then make “intuitive decisions” about developing alerts. They can even predict future outcomes.
Even more crucially, machine learning systems are programmed to modify their outputs automatically. This generates new data points for the false-positive remediation process, without the need for compliance employee intervention.
As part of a large-scale compliance solution, machine learning tools accelerate the remediation process for AML alerts, detecting false positives faster than other forms of analysis. This allows it to escalate real positives where it’s necessary.
According to SAS, a provider of anti-money laundering solutions, “machine learning enabled-compliance solutions” may lower false positive alerts by about 55%.
The way they do it is the following:
1. Structuring Data:
Machine learning systems help companies structure their data better by learning to both prioritize and categorize information that is based on the relevance to a particular type of alert.
2. Statistical and Semantic analysis:
Machine learning systems can be trained to spot redundant data by “semantic context” to streamline alert remediation. They can also be programmed to conduct statistical analysis on both historic and “emergent” data to help set up the likelihood of a positive false alert.
The greatest advantage of using both AI and ML-based systems is the ability to look at each transaction “holistically”, taking every component into account and combining them to create a more specific “fraud risk measurement”.
Teradata, in a case study of Danske Bank, reported that implementing AI decreased false positives by 60%. Concurrently, it improved real fraud detection by 50%. As the system continues to learn and fine-tune its detection model, the reduction in false positives could go up to as much as 80%.
Striking a balance between customer convenience and fraud detection will always be an ongoing challenge. If things go off-kilter, it can mean a damaged reputation and potential financial losses. Both AI and machine learning solutions offer a viable way to manage risk, enrich the customer experience, and mitigate false positives.