South African banks are losing significant amounts of interchange revenue as legitimate card transactions are increasingly incorrectly declined by outdated fraud detection systems.
This is one of the key findings of a new report by payment technology provider BPC.
It examines how outdated fraud detection and payment authorisation systems are causing banks to wrongly decline legitimate card transactions, resulting in hidden revenue losses and poorer customer experiences.
The study, titled: “Modernisation without disruption,” estimates the revenue impact of avoidable false payment declines across five markets – SA, Ecuador, Europe, Malaysia and Morocco.
According to the study, local banks could be losing an estimated $160 000 (R2.6 million) annually in interchange revenue as a result of legacy systems failing to identify legitimate transactions accurately.
It points out that SA experiences one of the highest revenue impacts among the five markets analysed because of its interchange structure.
While banks have traditionally focused on preventing fraud losses, the report argues that avoidable false declines have become an overlooked financial drain, particularly in markets such as SA and Ecuador, where interchange rates are relatively high.
Khurram Ahmed, senior product consultant for fraud solutions at BPC, tells ITWeb that the underlying problem lies in ageing payment infrastructure rather than the local market itself.
"South Africa has relatively high interchange rates compared to regulated markets like Europe, namely 0.44% on debit and 1.48% on credit, according to the South African Reserve Bank's interchange tables,” notes Ahmed.
“That means every legitimate transaction that gets incorrectly blocked carries a higher revenue cost for the issuer. But interchange is only the surface issue. The deeper question is why those transactions are being declined in the first place."
In many cases, the answer comes back to the authorisation platform, he adds.
“Legacy fraud detection systems that rely on rigid, rules-based logic tend to flag normal spending patterns as suspicious, rather than applying the kind of real-time, behavioural risk scoring that modern platforms use. That is a technology infrastructure problem, not a market-specific one.”
According to the study, Europe records the lowest losses due to regulated interchange fee caps rather than fewer false declines.
Malaysia and Morocco fall between these two extremes, illustrating that although the model assumes the same transaction volumes and false decline rates across all five markets, the financial impact varies significantly depending on each market's interchange structure.
The report stresses that this does not mean banks in Europe, Malaysia or Morocco experience fewer false declines, only that the revenue lost from each incorrectly declined transaction differs according to the interchange fee.
False negatives
For the research methodology, BPC used an illustrative model based on publicly available interchange rate data across the five markets.
The model assumes a representative mid-sized issuer processing 10 million attempted debit transactions a month.
Using a conservative false decline rate of 0.5 percentage points, it estimates that around 50 000 legitimate transactions could be incorrectly rejected every month by each bank, resulting in about $1.5 million in lost approved spending before interchange revenue is applied.
While the study does not single out individual local financial institutions, it highlights that the issuers most exposed to avoidable false declines, in SA or any market, are those still running legacy authorisation platforms built around static, rules-based fraud logic. These systems were designed for an earlier era of payments.
According to Ahmed, these platforms rely on predefined thresholds instead of analysing transaction behaviour in real-time, increasing the likelihood that legitimate spending will be incorrectly blocked.
"The result is that normal spending gets flagged as suspicious, legitimate transactions are blocked and revenue is lost, without the bank necessarily knowing the scale of the problem.
“Legacy infrastructure does not just affect fraud performance; it creates what we describe as hidden cost areas across authorisation leakage, operational drag, and opportunity cost and fraud management. Banks that have not modernised their core payment stacks tend to experience all of these simultaneously."
Beyond the immediate loss of revenues, the report warns that falsely declined payments can have dire consequences on customer experience and in turn longer-term consequences for customer retention.
According to research cited by BPC, conducted by Datos Insights, 78% of financial institutions believe failed payments have a critical impact on customer experience, while one-third report losing between 2% and 5% of customers because of payment failures.
Hidden cost
According to the study, banks have traditionally focused their fraud management efforts on preventing easily measurable financial losses; therefore, the cost of false declines often goes unnoticed because it leaves no obvious financial trail.
"Fraud losses are directly visible. They appear in scheme reporting, dispute records and financial statements. They have defined measurement frameworks and regulatory reporting requirements," Ahmed says.
"False declines are different. A declined transaction does not generate a loss entry in the same way. The revenue that would have come from that approval simply never materialises."
Ahmed describes this as an infrastructure visibility problem rather than simply a fraud management issue.
"The customer either retries elsewhere or switches to a different card, and the bank rarely connects that behaviour back to an authorisation decision. This is fundamentally an infrastructure visibility problem. Legacy authorisation systems were not designed to measure the cost of what they block."
According to the report, modern fraud management platforms increasingly use artificial intelligence (AI) and rich transaction context to evaluate payments in real-time, allowing issuers to reduce false declines without compromising fraud controls.
"By scoring transactions in real-time using machine learning models trained on richer transaction context, issuers can meaningfully reduce false declines, while maintaining fraud protection.
“Real-time decisioning allows issuers to assess risk at the individual transaction level rather than applying blanket rules. Modern systems also provide the data visibility to actually measure false decline rates and the associated revenue impact."
SA’s major banks − Standard Bank, First National Bank, Absa, Capitec and Nedbank – previously told ITWeb they are entering a new phase in their AI journey in 2026, shifting from pilot projects and isolated experiments, to large-scale deployment across core operations.
After years of testing machine learning and data-driven tools, the banks said they are now accelerating investment and integration of AI into fraud detection, customer service, risk management and digital platforms, signalling the technology is becoming central to competitiveness in the financial sector.
Share
