The TSB Case Shows How A 25-Cent Fee Became A $1.73 Million Conduct Problem New Zealand's Financial Markets Authority warned TSB Bank on September 21, 2026 after the bank charged a $0.25 fee on certain business cheque transactions that were not subject to the charge under its own terms and conditions, an error that affected 7,309 customer accounts and resulted in TSB paying back $1.73 million covering the fees, related interest and charges and compensation for the use of customers' money. The FMA said the mismatch persisted from November 30, 2017 to November 20, 2023, illustrating how a small automated charging error can become a substantial conduct issue when a bank's systems continue doing something that its contractual documentation says should not happen.
The episode is important precisely because the individual amount was so small. A 25-cent charge is unlikely to trigger an immediate complaint, especially when it appears among numerous account transactions. The same fee repeated across thousands of customers over several years, however, produces a very different financial and regulatory outcome.
The case illustrates one of the less visible risks created by increasingly automated banking systems. Low-value errors can become high-value institutional problems when they are repeated at scale. The FMA said TSB changed the wording of its Business Cheque terms and conditions in November 2017 to clarify fees, but the change did not fully align with the bank's systems. The systems continued charging the fee on transactions that should not have attracted it. TSB identified the issue in August 2023, stopped the charging in November that year and subsequently remediated affected customers.
The crucial issue is the gap between what the bank promised and what the system delivered. That gap is becoming increasingly important across financial services. Products are now defined partly by software. Fees, interest rates, eligibility conditions, transaction limits and customer permissions are implemented through technology systems that may operate across several platforms.
The system can therefore function exactly as programmed while the financial product still produces an incorrect customer outcome. This creates a governance challenge. A bank may have accurate documentation, compliant staff procedures and functioning software, but if those components are not synchronised, the customer experience can still become inconsistent with the institution's contractual obligations.
The problem is particularly difficult to detect when the financial impact on each customer is small. A customer may not notice a 25-cent charge, and even if the customer notices it, the time required to investigate may exceed the amount involved. The institution, by contrast, has access to system-wide information and can identify patterns if it has the appropriate controls. That creates an important responsibility around monitoring.
Financial institutions increasingly need systems that test actual customer outcomes against product terms rather than simply testing whether the software is operating as intended. If a bank changes a fee schedule, the change should be tested against representative customer transactions. If the product terms change, the underlying system should be reconciled against the new contractual position.
The TSB case also highlights the importance of legacy systems. Banks often operate technology environments that have evolved over many years. A change made in one part of the system may not automatically flow through every other component. The result can be a mismatch between legal documentation, customer communications, pricing engines and transaction-processing systems. Modernisation therefore becomes a conduct issue as well as a technology issue.
The FMA's intervention also highlights the importance of remediation. Returning the money is essential, but it does not eliminate the underlying question of how the error remained undetected for so long. A financial institution needs to demonstrate that the cause has been identified and that controls have been strengthened.
“A small financial error can become a significant conduct issue when it is repeated systematically and remains undetected,” FSCL said. “The important control is not simply checking whether individual transactions appear technically correct. Institutions need mechanisms that reconcile their contractual promises, product configurations and actual customer outcomes, especially when changes to fees, terms or product features are implemented through automated systems.”
The customer dimension is equally important. “A refund addresses the financial consequence, but customer trust is influenced by the entire remediation process,” FSCL said. “The institution has to explain what happened, correct the underlying system, identify all affected customers and demonstrate that the same type of error is unlikely to recur. Conduct risk increasingly sits at the intersection of legal obligations, technology and customer experience.”
That intersection is becoming more important as financial institutions introduce artificial intelligence and automated decision-making. Automation can reduce inconsistency, but it can also make errors less visible to individual employees. A human employee might identify an unusual transaction by chance. A system can process millions of transactions according to a flawed rule without raising an obvious alarm. The answer is not to abandon automation. It is to build stronger controls around it.
Exception reporting, independent testing, reconciliation between systems, customer-outcome monitoring and post-implementation review become increasingly important as the number of automated decisions grows. Institutions also need clear escalation mechanisms so that a small anomaly can be investigated before it becomes a systemic problem.
The TSB case is therefore relevant beyond banking fees. Similar issues can arise in insurance premiums, investment charges, loan interest calculations, foreign-exchange spreads and digital-payment fees. Any financial product in which software translates contractual terms into transactions can produce a mismatch if the two are not continuously reconciled.
The episode also illustrates the changing nature of financial conduct supervision. Conduct risk was once commonly associated with sales behaviour, misleading communications or unsuitable products. Increasingly, regulators are looking at whether the systems delivering financial products produce outcomes consistent with what customers were promised. That changes the role of compliance teams. Compliance can no longer operate entirely through policy documents and periodic reviews. It needs access to data, technology teams and customer-outcome information.
For customers, the lesson is less about the size of the fee than about the difficulty of detecting systemic errors. Individual customers cannot reasonably audit every transaction. Financial institutions therefore have a greater responsibility to identify problems that customers cannot realistically discover themselves.
The $1.73 million remediation figure puts the scale of that responsibility into perspective. A fee worth a quarter of a dollar became a multi-million-dollar customer-outcome issue because it was repeated and persisted for years.
The TSB case consequently provides a useful warning for every financial institution operating increasingly automated systems. The largest conduct problem may not begin with a major transaction or a dramatic failure. It may begin with a tiny number that nobody thinks is important enough to investigate.
The regulatory significance lies in what happens next. As banking becomes more automated, institutions will increasingly be judged not only by whether their systems work, but by whether those systems consistently deliver what customers were promised.