Fraud detection and monitoring concepts are the working ideas that banks, payment firms and merchants use to decide which transactions and customers to flag, review or block. Three of them recur across fraud and anti-money-laundering (AML) teams: the false positive, a legitimate event wrongly flagged as suspicious; alert fatigue, the loss of analyst attention that follows when alerts are too numerous and mostly wrong; and FRAML, the practice of running fraud and AML monitoring together.
They are linked by one trade-off. Lowering the thresholds of fraud detection and transaction monitoring catches more genuine cases but produces more alerts for staff to review, while raising them reduces workload but lets more crime through. Institutions use these concepts to judge whether their AML and fraud controls are effective rather than merely busy.
False Positive
A false positive is a legitimate payment or customer that a monitoring system wrongly flags as suspicious. In card payments it appears as a declined purchase or a blocked account, which costs the cardholder convenience and the merchant a sale; in AML it appears as an alert that an analyst reviews and closes without a report. It is the counterpart of the false negative, a genuine fraud that goes undetected, measured by the False Negative Rate. The usual measure is:
Here is the number of legitimate items flagged and the number of legitimate items passed. If 5,000 of 1,000,000 legitimate payments are flagged, the rate is 0.5%. Even a low rate can bury analysts: if those alerts sit beside 500 real frauds, only 500 of 5,500 alerts, about 9%, are genuine. Teams reduce the problem by tuning rules, adding a risk score and reviewing declines against later outcomes, as in rule tuning.
Alert Fatigue
Alert fatigue is the loss of attention and care that analysts show when they face a steady flow of alerts, most of them false. Causes include rules that are too broad, the same event generating alerts in several systems, poor customer data and thresholds set long ago and never revisited. The effects are backlogs, rushed or box-ticking reviews and the risk that a real case is closed along with the noise; the burden also contributes to staff turnover.
Regulators generally expect monitoring to be effective and risk-based, not simply to produce volume, so many institutions respond by prioritising alerts with a machine learning model or scoring scheme, applying a risk-based approach to focus on higher-risk customers, grouping related alerts into one case in a case management system and testing activity that fell below thresholds to see whether real cases were missed. Well-run investigations feed outcomes back to the rules.
FRAML (Fraud and Anti-Money Laundering)
FRAML is an approach that merges fraud detection and AML monitoring into one team, platform or shared set of data. The two disciplines overlap because the proceeds of fraud must be laundered, often through a money mule network, and because both rely on the same customer, account and transaction records. Separate teams can raise duplicate alerts on the same customer and miss patterns visible only when the data are combined, and a fraud loss may also require a suspicious activity report.
The benefits claimed include a single customer view, fewer duplicate investigations and quicker detection. Challenges include different objectives, since fraud work is largely real-time loss prevention while AML work is driven by regulatory reporting, along with data-governance, privacy and cultural barriers. Sharing between institutions is helped in the United States by section 314(b). Supervisors have generally encouraged better analytics while still expecting each programme to meet its own legal duties. Vendors market the idea through fraud management systems.
Ref 1050000824