By Nicole Govender
A bank can restore an unauthorised debit and still leave its customer poorer. The gap between correcting an account and repairing the harm is where South Africa’s bank-conduct framework is now being tested, one viral video at a time.
I call this the Reversibility Gap: the difference between reversing an erroneous entry and restoring the customer to the position she would have occupied had it never occurred.
There is a particular kind of dread reserved for the modern banking customer. It is not the discovery that an account is empty; that, at least, one usually sees coming. It is opening an app to find that yesterday’s credit balance has become today’s overdraft, with no intelligible explanation of how the numbers changed. In recent weeks, customers of one of the country’s major banks have taken to TikTok and other platforms to describe unexplained deductions, apparent duplicate transactions and balances that no longer reconcile. The comment sections have done what comment sections do best: confirmed that nobody suffers alone.
A word of discipline before the analysis. This article does not determine any individual complaint, identify the institution or any customer, or make any finding of wrongdoing. The volume of similar accounts does not prove that they share a cause, and public reporting that a technical incident was acknowledged and corrected is context, not proof of what happened in any particular account.
The question it asks is narrower and more useful: when a bank’s system takes money it was never owed, is giving it back enough?
What customers are describing
The accounts in circulation share a recognisable shape. A customer opens her banking app and finds a debit she does not recognise, a purchase apparently processed twice, or an overdraft she says she never used. She contacts the bank and receives a reference number, an incomplete explanation and an uncertain timeline. Sometimes the amount is substantial. Often the amount matters less than what follows it: a debit order bounces, a payment fails, and she can no longer tell which balance is safe to spend.
Media reports describe an incident involving duplicate card transactions and a settlement problem elsewhere in the payments chain. Statements attributed to the bank indicated that affected entries were being reversed, balances corrected, associated charges addressed and the process monitored, with an apology. Those are the right steps. But a general announcement cannot establish that every account has been correctly reconciled, that every consequential charge has been found, or that every customer has been put back where she was. That is the difference this article is about.
Two problems wearing the same face
The public accounts should not be collapsed into one explanation, still less into the word “scam”, which is easy to type and much harder to prove. At least two different card-processing events can produce the same sinking feeling.
The first is the duplicate presentment or settlement. A customer makes one purchase; her account reflects two debits. If the second was not authorised and cannot be justified under the account agreement or the applicable payment rules, the bank must explain on what basis it reduced her credit balance. If that entry also tipped the account into overdraft and attracted interest, fees or returned-payment charges, those consequences need their own examination. A system-generated entry does not become a customer mandate merely because it appears in a ledger.
The second is the expired authorisation hold. When a card is used, the merchant typically reserves the amount before submitting the transaction for final settlement. Depending on the scheme, merchant, issuer and timing, the hold may lapse before settlement arrives. The reserved amount then reappears as “available”; the customer, reasonably trusting her screen, spends it; and the merchant’s settlement later collects money that has already gone. Nothing need have malfunctioned. The conduct question is whether she was told, clearly and in time, the difference between a pending authorisation, a released hold, a posted transaction and the money genuinely available to her.
One may reflect a processing failure; the other may arise from ordinary card-processing mechanics. Both nevertheless raise questions about the accuracy of the information supplied to customers and the allocation of financial risk. In both, the customer makes financial decisions on information her bank supplied and may end up worse off for having done so.
Behind both lies the same sequence: authorisation, when the issuer approves the transaction and may reserve funds; clearing and settlement, when the transaction is transmitted, reconciled and settled between institutions; and posting, when it reaches the customer’s account. The Reversibility Gap opens at the last stage, where an error between institutions becomes a loss to the customer.
The balance on the screen is the bank’s statement of what the customer can spend. Where it is presented as available, she should be able to rely on it.
A bank account is a debt, not a vault
Popular imagination pictures a bank account as a box of notes with the customer’s name on it. South African law sees it differently. Where an account is in credit, the relationship is one of debtor and creditor: money deposited becomes the bank’s property, and the customer acquires a personal contractual claim to payment of the balance. The Supreme Court of Appeal reaffirmed the general rule, while recognising its limits, in FirstRand Bank Ltd v The Spar Group Ltd [2021] ZASCA 20. The customer’s “money” is, in truth, a claim against the bank.
That does not give the bank a free hand to reduce the claim. The relationship is contractual and includes a mandate governing the payment services the bank undertakes. In Absa Bank Ltd v Hanley [2013] ZASCA 183, the Supreme Court of Appeal explained that a bank transferring funds on a customer’s instructions acts as mandatary and must perform timeously, in good faith and without negligence. A debit the customer did not authorise, and which neither the contract nor the law otherwise justifies, is in substance the bank reducing its own debt to her without a basis for doing so.
It follows that an overdraft created or increased by such a debit is a curious creature: an apparent debt the customer did not incur, arising from a payment she did not authorise, on which she is then charged interest and fees. Even where she has an agreed overdraft facility, the facility does not supply the missing authority: the bank must establish the legal and contractual basis on which it treats the resulting amount as owing and charges interest or fees on it. The bank must be able to establish the legal basis for every rand it claims. Reversal is the first step in correcting an unsupported entry. It is not, without more, the answer to everything that entry set in motion.
What the regulatory framework requires
South Africa does not leave this to goodwill. The framework combines contract and mandate with a statutory conduct standard, an industry code and an independent ombud. Each does different work, and precision about which does what is the difference between analysis and rhetoric.
The Conduct Standard for Banks. Conduct Standard 3 of 2020 was issued by the Financial Sector Conduct Authority under section 106 of the Financial Sector Regulation Act 9 of 2017, with its complaints provisions in force since 3 July 2021. Section 2(4) requires banks to conduct their business in a manner that prioritises the fair treatment of financial customers. Section 4(7) requires a financial product or service to perform, or be executed, as the bank or its service provider has led customers to expect. Section 8 requires an effective complaints-management framework. Most tellingly, the Standard defines “compensation” as money, a benefit or a service provided for proven or estimated financial loss caused by the bank’s non-compliance, action, failure to act or unfair treatment, where the bank accepts liability. The definition expressly excludes a goodwill payment, a payment contractually due to the complainant, and a refund of an amount paid to the bank that was not contractually due. That last exclusion matters: reversing the principal debit and compensating a customer for the additional financial loss caused by the error are not necessarily the same exercise. The regulator, in other words, has already written consequential loss into the vocabulary of bank conduct.
An important qualification concerns the Standard’s scope. Section 2(3) provides that sections 4 to 10 do not apply to a bank in relation to the provision of credit under a credit agreement. The distinction matters where an erroneous transaction creates an overdraft or affects an existing credit facility. The Standard’s application must be assessed by reference to the particular financial service and contractual arrangement involved. A payment-processing service should not automatically be characterised as the provision of credit, although a credit-card transaction or an overdraft facility may raise additional questions about the credit relationship. Where the consequences engage a credit agreement, the National Credit Act and its dispute-resolution mechanisms may also become relevant. The exclusion does not remove the bank’s contractual obligations or any other remedy available in law.
Where the disputed entry sits on a credit agreement, such as an overdraft facility, section 111 of the National Credit Act 34 of 2005 supplies a statutory mechanism of its own. The consumer may dispute all or part of a particular credit or debit by written notice to the credit provider, which must then either explain the entry in reasonable detail or confirm the error and set out the revised entry. Until it has done so, the credit provider may not begin enforcement proceedings based on a default arising from the disputed entry. If the dispute is not resolved directly or through alternative dispute resolution, section 115 allows the consumer to apply to the National Consumer Tribunal. These provisions do not govern every transactional account. Where an erroneous debit pushes a customer into overdraft, however, they give her a right to an explanation and bar enforcement based on a default arising from the disputed entry until that explanation is given, and while the dispute is before alternative dispute resolution or the Tribunal.
The Code of Banking Practice. The Banking Association South Africa’s updated Code took effect on 1 July 2025. It is an industry commitment rather than an Act of Parliament, but a consequential one: participating banks accept the jurisdiction of the National Financial Ombud Scheme over unresolved complaints, and the ombud may decide a matter on the law where the law is reasonably certain, or make a recommendation in other circumstances, including on grounds of equity.
The National Financial Ombud Scheme. Operating since 1 March 2024, the NFO absorbed the former banking ombud. Its Banking and Credit Division, formed in July 2025 when the scheme’s banking and credit divisions merged, offers customers a free route once the bank’s own process has run its course. The scale of its banking work is instructive.
The NFO’s annual report for 2025 records 21,555 banking case openings, comprising 12,641 premature cases and 8,914 formal cases. These figures include matters that progressed from the premature to the formal stage: since 6,870 formal cases originated as premature cases, the total number of registered banking cases was 14,685, with each matter counted once. The report records 14,171 banking cases closed during the year and R53.0 million recovered for consumers through formal processes, with a further R22.3 million recovered at the premature stage, before complaints progressed to formal disputes. The report also records 17% of banking findings as being in favour of complainants, down from 21% in 2024. That proportion should be read with care. It should not be taken as the proportion of all complaints received, nor as evidence that banks are usually right or usually wrong. Whatever lies behind it, the figure is a reminder that complaints are decided on evidence, and that a customer’s best protection is the record she keeps.
The architecture also divides labour. The FSCA supervises conduct; the Prudential Authority supervises soundness; the NFO resolves individual disputes; and the Ombud Council oversees the ombud system. None of them is a complaints desk for a single customer’s bad Tuesday. Together, they are meant to ensure that bad Tuesdays are rare, and properly remedied when they happen.
When a TikTok is a complaint
There is a temptation, in boardrooms at least, to treat a social-media storm as reputational weather: noisy, temporary and best waited out. The Conduct Standard does not permit that luxury. But precision matters: not every post about a bank automatically becomes a complaint received by that bank. The legal question is whether the expression of dissatisfaction has been communicated to the bank or has come to its knowledge through a service provider, and whether the requirements of the Standard’s definitions are met.
The Standard defines a complaint broadly, as an expression of dissatisfaction relating to a financial product or service which alleges non-compliance, maladministration, harm, prejudice, distress, substantial inconvenience or unfair treatment. A complainant must have a direct interest, or act on behalf of someone who does. Section 8 requires accessible channels and an effective framework to receive, investigate, categorise, record and analyse complaints. A public post that is brought to the bank’s attention may therefore trigger a response obligation, depending on its content and circumstances. It is not sound to say that every comment or viral video, without more, is automatically a formally lodged complaint. Nor, once it reaches the bank and meets the definition, can it be dismissed as mere noise: the channel through which dissatisfaction arrives does not change the bank’s obligation to deal with it.
The FSCA’s 2025 report on its 2022 banking-sector desktop review found that the majority of the 23 banks assessed initially displayed unsatisfactory complaints-management processes against the requirements of section 8. The report also records subsequent improvements by banks, but its findings remain instructive about the weaknesses identified in the sector. That is context, not a finding about any institution in the present incident. It reinforces a narrower point: complaints management is not an administrative afterthought. It is a regulated control function through which a bank must receive, investigate, categorise and resolve dissatisfaction through an accessible and effective process. Its wider lesson is that complaints data is itself a conduct-risk resource: each complaint, properly recorded, is not only a grievance to be closed but evidence about how the bank’s products, systems and controls are performing. The implications of recurring complaints extend beyond the handling of individual cases and warrant separate consideration.
Why reversal looks like a complete remedy
If the gap is so apparent, why has it persisted within the legal and regulatory framework? Part of the answer lies in how the law frames the bank account. Because an account in credit is a debt rather than a vault, an unauthorised debit that lacks a mandate does not, in law, reduce what the bank owes. From the bank’s accounting and contractual perspective, reversal is therefore not a concession but a correction: the number returns to where it always should have been, and the relationship appears repaired.
The structure of remedies reinforces that impression. To recover the correct balance, the customer must establish her contractual entitlement and the incorrect debit, but she need not prove consequential damages merely to establish what the bank owes. A claim for what the error set in motion is different: it is a claim for damages, carrying the full weight of causation, remoteness, quantification and mitigation, with the burden resting on the customer. The remedy that is easiest to obtain is the one that leaves the gap open; the remedy that would close it is the hardest to pursue.
Scale completes the picture. Consequential losses from banking errors may be individually modest and widely dispersed: a returned-debit fee here, a day’s bridging interest there. Losses of that size are rarely worth litigating individually. Inter-institutional settlement and loss-allocation mechanisms within payment schemes do not necessarily provide the customer with a remedy for downstream costs, and complaints systems are built to measure closure rather than restoration. Each element is rational on its own terms. Together, they produce a system in which correcting the ledger can pass for repairing the harm.
“Resolved” is not the same as “remedied”
Here lies the heart of the matter. A reversal returns the principal. It does not necessarily return the customer to the position she would have occupied had the error never happened.
Why is the bank’s definition of a resolved complaint potentially narrower than the customer’s actual financial loss?
Consider a customer with R5,000 to last the month. An erroneous R2,000 debit leaves her with R3,000, and her rent debit order bounces. Her insurer’s premium fails and cover is at risk. A school payment attracts a penalty. She borrows from a relative, or from a lender at a price, to bridge the gap. The next day, the R2,000 reappears with an apology. The ledger is now correct. Her month is not. The returned debit, the penalty and the cost of borrowing remain exactly where the error left them. That difference is the Reversibility Gap. It is not a new cause of action. It is a way of testing whether a bank that has corrected an error has also repaired its consequences.
The common law has long had a vocabulary for this, though not a blank cheque. A claimant must establish a cause of action, prove breach and causation, quantify the loss and meet the rules on remoteness and mitigation. Our courts distinguish between loss flowing naturally from a breach and special loss recoverable where it was within the parties’ contemplation when they contracted, an approach most familiar from Holmdene Brickworks (Pty) Ltd v Roberts Construction Co Ltd 1977 (3) SA 670 (A) at 687D–F. In the context of a transactional banking relationship, the failure of an essential debit order is a potentially foreseeable consequence of an erroneous reduction in available funds. Whether the resulting loss is recoverable will nevertheless depend on the applicable contractual principles, causation and the circumstances of the particular claim.
The conduct framework points the same way without overriding the law of damages. Beyond the definition of compensation, section 8(12) requires that, where a complaint is upheld, any commitment by the bank to make a compensation payment or goodwill payment, or to take any other action, be carried out without undue delay and within any agreed timeframes, and section 8(15) requires it to keep data on compensation and goodwill payments. None of this creates an automatic entitlement to whatever is claimed. All of it makes it untenable to treat the return of the original debit as the only possible measure of redress where further loss is proved and liability is accepted or established.
Returning what was taken settles the account. It does not necessarily settle the debt.
The third-party alibi
An explanation that places the error elsewhere in the payments chain (an acquirer, a processor, a scheme) may be commercially accurate. Legally, it is unlikely to carry much weight with the customer, who dealt with her own bank, used its card and relied on its app. She did not choose the acquiring institution, cannot see the settlement files and has never read the agreements between them.
The Conduct Standard anticipates this. Its definition of a complaint extends to the conduct of a bank’s service provider, and section 8(3)(i) requires the complaints framework to deal with complaints involving service providers, including referral, monitoring and communication. That does not make a bank automatically liable for every act of every independent participant. It does mean that the involvement of a third party is no excuse for failing to investigate, explain the entries or correct an unsupported debit. The bank may well have recourse against whoever erred; that is a matter between institutions. Risk in a payments chain ought to sit with the parties who design it, price it and can insure against it. It ought not to settle, by default, on the one party who cannot even see it. A payment may have many participants, but accountability must never become a collective responsibility that, in practice, belongs to no one.
What other jurisdictions already require
South Africa need not invent an answer from scratch. In the United Kingdom, regulation 76 of the Payment Services Regulations 2017 (SI 2017/752) requires a payment service provider to refund an unauthorised payment transaction as soon as practicable, and in any event no later than the end of the business day after it becomes aware of it. Where applicable, the provider must also restore the debited account to the state it would have been in had the transaction not taken place. Article 73(1) of the European Union’s second Payment Services Directive (Directive (EU) 2015/2366) imposes the same next-business-day obligation. Both obligations carry an exception where the provider has reasonable grounds to suspect fraud and reports those grounds to the appropriate authority, and the UK duty also operates subject to regulations 74 and 75. In Tukowiecka (Case C-70/25, N.O. v PKO BP S.A.), the First Chamber of the Court of Justice held on 15 October 2026 that a provider may not refuse that immediate refund on the ground that the payer enabled the transaction through gross negligence (ECLI:EU:C:2026:852; see especially paragraphs 33–35, 42 and 46–47, and the Court’s answer at paragraph 52, repeated in the operative part). The provider must refund first. It may subsequently seek reimbursement where it establishes that the customer acted fraudulently or failed, with intent or gross negligence, to comply with the relevant obligations under Article 69 of the Directive. The distinction between ordinary negligence and gross negligence is important under Article 74(1).
An important qualification applies. These rules, and the Court of Justice’s judgment, concern unauthorised payment transactions. They do not establish that every duplicate settlement or processing error falls automatically within the same regime, still less that a valid transaction settling after an authorisation hold has expired must be refunded. Whether a particular debit counts as unauthorised remains a question of characterisation in each case.
Two features stand out. The first is the clock: a refund measured in hours, not in reference numbers. The second is the standard of restoration: the account must be put back as if the transaction had never happened, which reaches the interest and charges the error generated on the account itself, though not every loss suffered beyond it. Neither rule is a panacea. But where a debit is unauthorised, both convert what South African customers must currently argue for into something they can simply expect.
Recurrence and the regulatory question
Not every unexplained deduction is a breach. Some may involve third-party fraud; others may reflect legitimate authorisation and settlement events, delayed presentment or a misunderstanding of the distinction between ledger balance and available balance. The legal character of an entry cannot be determined from a screenshot or a cluster of social-media posts alone. It requires transaction-level evidence: the original authorisation, any reversal or release, the clearing and settlement records, the account terms, notifications sent to the customer and the final ledger entries.
Recurrence nevertheless matters. Where materially similar incidents are reported repeatedly, the question is no longer confined to the correction of one account. A pattern may point to weaknesses in reconciliation, outsourced-service oversight, customer communication or complaints management. Section 8(16) of the Conduct Standard requires complaints information to be analysed to identify risks and trends and to prevent the recurrence of errors and poor outcomes. This is a regulatory obligation directed at learning from complaints, not merely closing them one by one.
The sharper question is therefore not whether every customer who reports a missing amount is automatically entitled to damages. It is whether South Africa’s bank-conduct framework, together with ordinary contractual remedies, provides an effective route to identify the error, restore the account, remove charges that have no lawful basis, correct adverse reporting and compensate proven consequential loss where liability is established. The answer depends on the interaction between mandate, contract, the Conduct Standard, the Code and the jurisdiction of the relevant ombud or court. If systemic patterns become visible only after customers amplify them publicly, regulators should ask whether ordinary complaint channels and internal controls are detecting problems early enough.
The point is not only a regulatory one. Mr. Justin Pravda of Pravda & Knowles Attorneys puts the commercial logic plainly: “Banks need customers to buy in, and customers buy in only where they trust that the system will treat them fairly. Fair conduct is not a constraint on the banking business; it is the very thing the business is selling. An institution that will not hold itself to that standard is not protecting its margins. It is eroding the confidence on which every deposit, every transaction and every product depends, and confidence, once lost at scale, is the one thing a bank cannot survive without. Trading that way is self-defeating.”
Closing the Reversibility Gap: four targeted changes
The existing framework has the right vocabulary. What it lacks is specificity at the point where errors become losses. Four targeted changes would close most of the gap. They are proposed safeguards, not a description of rights customers can enforce today. Most could be achieved through amendments to the Conduct Standard or the Code rather than new primary legislation, although some, such as incident reporting and scheme-level timelines, would also depend on the FSCA’s existing reporting powers and the payment-system rules overseen by the South African Reserve Bank.
1. A refund clock, with a clear line between confirmed and disputed debits. Where a bank confirms that a debit was unauthorised or erroneous, it should reverse it by the end of the next business day, value-dated to the original debit date. Where a customer disputes a debit and the bank cannot yet confirm its validity, the bank should act within five business days of the complaint. It should either produce the records establishing a proper basis for the entry (the authorisation, presentment and settlement data) or provisionally credit the disputed amount while it completes its investigation, which should conclude within a defined period, such as 30 days. An unresolved dispute is not a confirmed error; but nor should the customer fund the bank’s investigation while it runs.
A provisional credit must not become a second source of dispute, so its reversal should be hedged with four safeguards. First, reasons and evidence: a bank seeking to reverse a provisional credit should give written reasons and disclose, in intelligible form, the evidence on which it relies. Second, an opportunity to respond: the customer should have a reasonable period, such as ten business days, to challenge that conclusion before the reversal takes effect. Third, no collateral harm: neither the provisional credit nor its reversal should itself generate interest, fees or adverse credit reporting, and where a reversal would leave the account overdrawn, the customer should have a reasonable period to regularise it at no cost. Fourth, a clear path where the parties disagree about the evidence: a challenge should go to an independent internal review within the bank’s complaints framework, concluded within a fixed period, after which the customer may refer the matter to the NFO, with no adverse credit reporting while the referral is pending.
None of this removes customer liability where it properly exists. If the bank establishes that the customer authorised the transaction, acted fraudulently, or is otherwise liable under the account terms and the applicable law, it may recover the provisional credit. A provisional credit is an interim allocation of risk, not a finding of liability or a waiver of a valid debt; but a bare assertion that the bank’s records show the transaction as valid should not, on its own, be enough to reverse it. The structure draws on the logic of Tukowiecka: refund first, decide liability afterwards, and place the burden of proof on the party holding the records (see para 40). That judgment, however, applies a specific European statutory refund regime. The framework proposed here is a policy recommendation for South Africa, not a legal consequence of that decision.
2. Restoration, not merely reversal. Interest, fees and charges generated on the account by the error should be reversed automatically, without the customer having to identify each one. Third-party costs, such as returned-debit penalties and borrowing costs, should be payable on reasonable proof.
3. A freeze while the bank investigates. No collection demands, overdraft interest or adverse credit-bureau reporting should flow from a disputed entry while the bank is investigating a credible error report.
4. Incident transparency. Processing incidents affecting retail customers at scale should be reported to the FSCA, with a remediation report confirming how many accounts were affected, how many were restored and what was paid.
These proposals are legally and institutionally feasible because they extend architecture that already exists rather than building new structures. The FSCA’s power to make conduct standards for financial institutions derives from section 106 of the Financial Sector Regulation Act. Conduct Standard 3 of 2020 was itself made under section 106(1), read with sections 106(2)(b) and 108(1), and it already contains the definition of compensation, the complaints framework and the duty to implement remedial commitments on which the proposals build. Where a disputed entry sits on a credit agreement, however, the section 2(3) exclusion means the corresponding safeguards would need to be secured through the National Credit Act framework rather than the Conduct Standard. Amending the Standard would follow the ordinary statutory process for a non-urgent regulatory instrument. Under section 98, the draft must be published with a statement explaining its need and intended operation, a statement of its expected impact as section 98(1)(a)(iii) requires, and an invitation for public submissions, with at least six weeks allowed for comment under section 98(2). Under section 103, it must then be submitted to Parliament for at least 30 days while Parliament is in session, and under section 104 a report on the consultation process must follow. That process is demanding by design, but it is an established statutory mechanism. The Code of Banking Practice offers a faster, complementary route: the industry updated it as recently as July 2025, and because the NFO already considers complaints by reference to the Code, an amended Code could provide a basis for the ombud to assess complaints and make determinations or recommendations within its applicable jurisdiction and rules, without waiting for a new regulatory instrument. An industry-code commitment is not, however, the same as a statutory obligation or a court order. Scheme-level matters, such as timelines for resolving duplicate presentments between issuers and acquirers, may also require attention within the payment-system rules overseen by the South African Reserve Bank.
None of these would penalise a bank that gets it right. They would simply ensure that, when it gets it wrong, the cost of the error lands where the control over it lies.
From analysis to action: what each institution should be asking
The legal analysis matters only if it changes what institutions do on a Monday morning. Each participant in the system has a distinct set of questions, and the most useful thing an incident like this can produce is a list of them.
Banks: reconciliation controls. Operations and technology teams should test whether duplicate-detection logic runs before a settlement entry posts to a customer’s account, matching on card, merchant, amount, authorisation code and time window, rather than after the customer notices. They should examine how inbound settlement files from acquirers and processors are validated, whether volume and value tolerances would flag an anomalous file, and whether anyone has the authority to halt one. They should review the ageing of unmatched items in exception queues and who owns them. They should also check whether the rules for releasing authorisation holds, and the way pending and settled amounts are displayed in the app, give customers an accurate picture of what they can actually spend.Finally, an entry flagged as a probable error should automatically suppress the interest and fees it would otherwise generate.
Compliance officers: complaints management. Compliance functions should ask whether dissatisfaction that reaches the bank through social media is captured, categorised and investigated as section 8 contemplates. They should check whether closure codes distinguish a reversal-only outcome from full restoration, because a system that records both as “resolved” cannot tell the board whether customers were actually made whole. They should confirm that compensation and goodwill payments are recorded as section 8(15) requires, that root-cause tagging is granular enough to support the trend analysis section 8(16) demands, and that complaints involving service providers are referred, tracked and reported back under section 8(3)(i). Sampling closed complaints for fairness of outcome, not merely speed of closure, is the test of whether the framework works.
Risk committees: what to monitor. Boards and their risk committees should receive a small set of leading indicators rather than a summary written after the event:
- the rate, value and ageing of reconciliation breaks and duplicate transactions;
- error-related complaint volumes alongside spikes in social-media sentiment;
- the elapsed time from error report to reversal, and from reversal to full restoration;
- consequential-cost payments and ombud referrals and outcomes;
- service-provider incidents and service-level breaches, especially where the same root cause recurs.
A recurring root cause is the single most important signal. Put rhetorically rather than as any regulatory threshold: one incident is operational risk. Two is a control failure. Three is a governance question.
Customers: what to request. A customer disputing a transaction should ask for a full reconciliation of the entry: the authorisation, the presentment, the settlement, any reversal and the dates of each. She should ask for the written contractual basis of any overdraft, interest or fee that followed, the reversal of charges generated by the error, and written confirmation that no adverse credit information has been or will be reported. Where the entry sits on an overdraft or other credit facility, she can send a written dispute notice under section 111 of the National Credit Act. She should keep the complaint reference, insist on a written outcome and, if the answer is unsatisfactory, take the file to the NFO’s Banking and Credit Division. A specific, dated and documented complaint is far harder to close with a generic apology.
Regulators: where complaints recur. When similar complaints cluster, the supervisory questions go beyond the individual account. Was the incident a single event or the recurrence of a known root cause? Were reconciliation and outsourcing oversight controls adequate? Did remediation proactively identify every affected customer, or only those who complained? Did it restore accounts or merely reverse debits? Did the bank’s own complaints analysis detect the trend before the public did? Does the presentation of “available” balances meet the expectation in section 4(7) that products perform as customers have been led to expect? Where the same pattern appears across institutions, a thematic review of duplicate-settlement handling and remediation outcomes would serve the whole market better than a series of individual inquiries.
If this sounds familiar
Somewhere reading this is a person who knows exactly what that Tuesday morning felt like: the app opened, the number checked, the screen refreshed twice in the hope that it would correct itself. Perhaps the money has since come back. Perhaps the fees have not. Perhaps she is still trying to establish whether a payment was merely pending, whether a debit was duplicated, or whether a separate transaction simply settled late.
It is worth saying plainly that she is entitled to an intelligible answer, and that the requests listed above are hers to make. The statements she downloaded, the screenshot of her “available” balance, the notice that her debit order bounced and the SMS demanding immediate repayment are not merely the record of a bad week. They are the evidence on which a fair investigation, and any later claim, will be decided.
An individual complaint is a grievance. A cluster of materially similar complaints is a pattern, and patterns are how banks find weaknesses, how ombuds recognise systemic disputes and how regulators decide where to look. The legal questions raised here will not be settled by a single apology or a single reversal, and the author will continue to follow the response as reliable information emerges. So, it seems, will a great many South Africans who have discovered that they share more than a bank.
Banks have spent a decade persuading South Africans that the app is the bank. They were right. Which means that when the app shows money that is not there, or takes money that was never owed, “we have reversed the transaction” is not an answer. It is the beginning of one. The rest of the answer lies in closing the Reversibility Gap.
________________________________________________________________________________
Ms. Nicole Govender
Independent Legal Researcher & Commentator
LLM (Commercial Law) | University of Johannesburg
Digital Lending | Banking and Credit Law | Financial Regulation | Fintech | Consumer Protection
Copyright and licence
© 2026 Nicole Govender. This article is published under a Creative Commons Attribution-NoDerivatives 4.0 International licence (CC BY-ND 4.0). It may be shared and republished in full, including by media organisations, provided the author and the original publication are appropriately acknowledged, and the text is not altered. Brief quotation for purposes of comment, criticism, review or news reporting is welcome, with attribution. Any adaptation, abridgement or translation requires the prior written permission of the author. Where this work is republished, quoted or relied upon, it should be attributed to the author with a reference to the original publication and its DOI.
How to cite
Govender, N. (2026). Reversed but not repaired: who bears the loss when the bank gets it wrong? Zenodo. https://doi.org/10.5281/zenodo.24337329
________________________________________________________________________________
D I S C L A I M E R
This article is published as commentary on a matter of public interest and is for general information only. It does not constitute legal advice and does not create an attorney–client relationship. Professional advice should be sought on the specific facts of any matter before action is taken. The article refers to matters that have been publicly reported and attributes statements to their reported sources. It makes no allegation or finding of wrongdoing against any institution or individual and does not identify or refer to the circumstances of any particular customer. The opinions expressed are the author’s own, are honestly held and are based on information publicly available at the date of publication. They do not necessarily reflect the views of any organisation with which the author is associated or on whose platform the article is republished. The law and facts are stated as at the date of original publication. The author does not warrant the accuracy, completeness or continued currency of information derived from third-party reports. To the fullest extent permitted by law, the author and any publisher accept no liability for loss arising from reliance on this article.