Chargebacks

49.0 What this chapter gives you#

1. You will be able to explain why the money leaves a merchant’s account before anybody has decided who was right, and why that ordering is deliberate rather than a defect.
2. You will be able to name the four parties in a card dispute and say which of them the cardholder actually has a claim against.
3. You will be able to read a dispute’s reason code as the question being asked, and choose evidence that answers that question rather than the customer’s grievance.
4. You will be able to work out, for a given sale, whether contesting a dispute pays, using the dispute fee, the countered fee, staff time and a realistic win rate.
5. You will be able to explain why a merchant refunds a case it would almost certainly win, and why that is arithmetic rather than cowardice.
6. You will be able to compute a VAMP ratio and a Mastercard Excessive Chargeback Programme ratio, and say why the two produce different answers for the same business.
7. You will be able to distinguish a fraud report, filed as TC40 or through SAFE, from a dispute, and explain why an event that moved no money can still end a merchant account.
8. You will be able to say what chargeback gives a United Kingdom consumer that section 75 does not, and what section 75 gives that chargeback cannot.
9. You will be able to explain why a disputed sale can come back at a different amount from the one that settled, and name at least three causes.
10. You will be able to state what a merchant must record at the moment of sale in order to use Compelling Evidence 3.0 eighteen months later.

The plain version#

Picture a school summer fete where nobody carries cash. Twenty stalls, and every one is paid the same way: you tell the desk by the gate what you bought and for how much, the desk writes it down, and at the end of the week the desk pays each stall what it is owed. You settle up with the desk separately.

The desk is not a bystander. It is the thing that actually moved the money. So if you come back on Monday and say the cake stall took four pounds off you and handed you an empty box, the desk has a lever nobody else at the fete has. It can decline to pay the cake stall, or, having already paid, take four pounds out of next week’s takings.

That is a chargeback, and your bank is the desk. When you tap a card, your bank pays the shop and bills you afterwards. Your money never travels to the shop; your bank’s money does. Your bank therefore has both the standing and the plumbing to reverse the payment.

The part that surprises people is the order of events. The money moves first and the argument happens afterwards. You ring your bank, your bank forms a view in a few minutes, and the shop’s money is taken. It is not a hearing followed by a verdict; it is a seizure followed by an appeal.

Here is what that looks like with real numbers. Dee runs a small online shop in Leicester selling headphones and turntables. In March she sells a pair of headphones for £145 to a customer in Swansea and posts them the next morning. Her cost was £85, so on that sale she made £60 before overheads.

Eleven weeks later, in the last week of May, an email arrives from her payment provider. The £145 has been disputed. The reason given is that the cardholder says he did not make the payment. Dee’s account has already been debited: £145 for the sale, and £20 on top as a fee for having received a dispute at all. That £20 figure is what a mainstream UK payment provider charged at the time of writing in August 2026, and it does not come back whatever happens next.

So before Dee has read the second paragraph of the email, she is £165 down and the headphones are in Swansea.

Now she has a decision to make, and it is not the one most people expect.

She can fight. The industry word for fighting is representment, which literally means presenting the transaction a second time: here is the sale again, and here is why it was good. To do that she gathers what she has. The order, the delivery address, the courier’s tracking record, the time and date the parcel was signed for, the customer’s email address, and the fact that the same email address and the same card bought a turntable from her in January and never disputed it. She uploads all of it. Filing costs another £20 at the time of writing, refundable only if she wins.

And then she waits. Not days. The bank that raised the dispute typically takes two to three months to read the file and decide.

Two things about that wait matter. Dee never speaks to her customer again; she is submitting a file to a stranger at a bank she has no relationship with, who will read it alongside a few hundred others. And the question that stranger answers is not “was Dee treated fairly?” It is a much narrower question, fixed at the very beginning by the box the customer’s bank ticked.

That box has a number, and the number is the most important thing in this chapter.

If the box says “I did not make this payment”, the only question is whether the cardholder made the payment. Delivery proof helps only insofar as it suggests the real cardholder received the goods at his own address. If the box says “the goods never arrived”, the only question is delivery, and a signature at the door ends the argument in a sentence. If the box says “the goods were not as described”, a signature at the door is worth nothing at all, because nobody disputes that the parcel arrived; they dispute what was in it.

Merchants lose disputes they should have won by answering the grievance instead of the box.

Suppose Dee wins. The £145 comes back, and so does the £20 filing fee. The £20 dispute fee does not. She has spent three hours and waited eleven weeks to end up £20 and three hours worse off than if the sale had gone through.

Suppose she loses. She is out £145, out both £20 fees, and out the £85 the headphones cost her: £270 in total. On a net margin of four and a half per cent, which is realistic for a small online retailer, £270 is the entire profit on about £6,000 of sales. One disputed pair of headphones eats a fortnight.

If the story ended there, chargebacks would be an expensive nuisance. It does not end there, because the card companies count them.

Every month, Visa and Mastercard divide the number of disputes a shop received by the number of sales it made, and compare the answer with a threshold. Cross it and the shop is put into a monitoring programme. That means monthly fines starting at around a thousand US dollars and climbing, at the top of the schedules in force in August 2026, to two hundred thousand US dollars a month. It means a written remediation plan. And if the numbers do not come down, the shop’s bank ends the relationship, at which point its name goes onto a shared list that other banks consult before taking on a new customer. For a business that sells online, being unable to accept cards is not a setback. It is the end of the business.

Dee sells about 9,000 items a month. On Mastercard’s threshold in force at the time of writing, a shop of that size is at risk once it takes 135 chargebacks in a month and its ratio reaches 1.5 per cent. Not 135 lost chargebacks. 135 chargebacks. Whether she wins them makes no difference whatsoever to the count, because the count is taken when the dispute is filed, months before anyone knows who was right.

Which produces the most counter-intuitive behaviour in the business, and the thing this chapter exists to explain. A shop with a strong, defensible, almost certainly winnable case will very often refund the customer instead of fighting. Not because it thinks it would lose, but because winning costs a non-refundable fee, several hours of somebody’s week, an eleven-week wait, and does nothing at all to the number that can get its ability to take payments withdrawn. Refunding the customer before a dispute is raised costs the sale price and keeps that number down. The shop is not conceding the argument. It is paying to keep a counter from ticking.

That is why the industry has quietly built an entire layer in front of the dispute process: services that let a bank ask the shop about a suspicious transaction before raising anything, and services that let the shop pre-authorise an automatic refund the instant a dispute is about to be filed. Both exist for the same reason. A refund that arrives before the dispute is a refund. A refund that arrives afterwards is a chargeback, and chargebacks are counted.

Where the plain version stops being true#

Chargeback is not your legal right, and you are not a party to it. The plain version made it sound as though Dee’s customer was pursuing a claim. He was not. He asked his bank to pursue one, and his bank chose to. In the United Kingdom, at the time of writing in August 2026, chargeback has no statutory basis at all: it is a private contractual mechanism created by the card schemes and binding only on their licensees. UK Finance states it in exactly those terms: unlike section 75, chargeback is not a legal right. The Financial Ombudsman Service is as plain, saying a bank does not have to raise a chargeback, though where valid reasons exist it can be good practice, and that what it examines is whether it was fair not to raise one or to discontinue one. That is a duty owed by the bank to its customer, not a right of the customer against the shop. The legal instruments sit alongside chargeback and behave differently: section 75 of the Consumer Credit Act 1974 makes a credit card issuer jointly and severally liable with the supplier for breach of contract or misrepresentation, applying only to credit cards, only where the cash price of the item is over £100 and no more than £30,000, and with a six-year window rather than a scheme deadline; and regulation 76 of the Payment Services Regulations 2017 requires a payment service provider to refund an unauthorised transaction no later than the end of the business day following the day it becomes aware of it, which is a statutory obligation on the bank that exists whether or not any chargeback is ever raised or ever succeeds. The bank refunds because the law tells it to. The chargeback is how the bank goes looking for the money afterwards.

A chargeback is not the reversal of the original payment, and it does not always come back the same size. The fete analogy invites you to picture the same four pounds travelling backwards. It is not the same four pounds and it is often not four pounds. A chargeback is a new financial transaction with its own message, its own clearing cycle, its own value date and its own currency conversion, and any system built on the assumption that the disputed amount equals the original amount will eventually produce a reconciliation break nobody can explain. The commonest cause is foreign exchange: a payment converted at January’s rate and disputed in April comes back at April’s rate, so a sale that settled as 100 euros can return as 107.86 euros. Three other causes are routine. An issuer can bundle several months of a disputed subscription into one chargeback filed against a single charge. A cardholder can dispute part of an order. And a cardholder can dispute the full amount of a charge already partly refunded, leaving the merchant out the refund and the chargeback at once and having to prove the overlap.

Winning is neither final nor free, and it does not undo the count. The plain version implied a scoreboard with two outcomes. There are more. A successful representment can be met by the issuer filing a further stage, so cases marked won are routinely reversed weeks later and the money leaves the merchant’s account twice. Money also moves outside the lifecycle altogether when an issuer credits or debits to correct an amount, which is why a lost case occasionally turns into a late win months after everyone stopped looking. The fee for receiving a dispute is normally non-refundable however the case ends. And the monitoring counters are indifferent to outcomes by design, not by oversight: waiting months for every outcome would make monitoring useless, and what the schemes measure is how well a merchant prevents disputes rather than how well it litigates them. Refunds do not remove a dispute from the count either, so a merchant who refunds after the dispute exists has paid twice for nothing.

Most disputes filed as fraud are not fraud, and the reason code is frequently the wrong one. The plain version treated the box as a neutral label chosen by an informed person. It is neither. A dispute filed as card-absent fraud asserts that the cardholder did not authorise the transaction, and that is the assertion the merchant must answer, regardless of what the customer’s actual grievance was. Very often the actual grievance is that the subscription was harder to cancel than to start, or that a family member used the card, or that the statement descriptor was a holding-company name nobody recognised. The industry calls this first-party misuse, and it is a large enough share of card-absent disputes that both major schemes have built dedicated programmes to attack it. Two consequences follow. A merchant holding perfect delivery proof can lose a dispute where delivery was never in question, because it answered the customer instead of the code. And merchants get undeserved wins in the other direction, when a case really about quality is filed as non-delivery and defeated by a tracking number, leaving the customer’s genuine complaint unadjudicated by anybody.

The technical version#

Everything in this section is stated as accurate at the time of writing, in August 2026, and this is the most perishable chapter in the most perishable volume of this book. Scheme rulebooks are revised at least twice a year, one threshold below moved in April 2026 having moved in 2025, and the fee schedules are contractual documents not published in full. Check the current rulebook and your own acquirer’s schedule before relying on any figure here.

Vocabulary, and who the parties actually are#

Visa’s rulebook calls the process a dispute; Mastercard’s calls it a chargeback. Most acquirers treat the words as synonyms, and so does this chapter, while keeping one distinction sharp: a pre-dispute event is one where no money has moved and no counter has incremented, and a dispute proper is one where funds have been debited from the acquirer and passed to the issuer.

The parties are those of the four-party model. The cardholder complains. The issuer decides whether to raise the case and carries the statutory obligations towards its customer. The scheme provides the message formats, the deadlines, the reason codes and, at the end, the adjudication. The acquirer is debited by the scheme and passes the debit on to the merchant. That last link matters more than merchants realise: the acquirer’s right to take money out of a merchant’s account comes from the contract between them, not from the scheme, and the acquirer carries the credit risk if the merchant has no money left. That single fact explains rolling reserves, delayed settlement for high-risk sectors, and why an acquirer exits a merchant long before the scheme forces it to.

Two data feeds run alongside the dispute system and are constantly confused with it. Visa’s TC40 and Mastercard’s SAFE are fraud reports: an issuer that believes a transaction was fraudulent must report it, and the report reaches the acquirer as what most gateways call an early fraud warning. A fraud report is not a dispute and no money moves, but it counts in Visa’s monitoring ratio, which is the most misunderstood fact in merchant risk management. Analysis published by one large processor at the time of writing indicates that around 80 per cent of early fraud warnings become a fraud dispute if the merchant does nothing.

Stage zero: the pre-dispute layer#

Before a dispute exists there is a layer designed to stop one existing, and it explains most of the merchant behaviour that looks irrational from outside. Visa’s is operated by Verifi. Order Insight pushes enriched transaction detail, the merchant name a human would recognise, the item, the delivery address, into the issuer’s call-centre screen and banking app at the moment the cardholder queries the charge, so that “I don’t recognise this” is answered before it becomes anything; an April 2026 update lets merchants share Compelling Evidence 3.0 data at this stage. Rapid Dispute Resolution goes further: the merchant writes rules in advance, up to ten scenarios keyed on attributes such as amount and dispute condition, and a matching case about to be raised is refunded automatically in near real time. Mastercard’s equivalent alerting network is Ethoca.

The commercial logic is one sentence: cases resolved through these pre-dispute products are excluded from the merchant’s dispute count for monitoring purposes, as Visa’s own programme documentation states. A merchant paying an alert fee and refunding a sale in full is buying the removal of a tick from a counter, and that is often worth several times the value of the sale.

American Express and Discover retain a formal inquiry stage, sometimes called a retrieval or request for information, where the issuer asks before it takes; Visa and Mastercard no longer use inquiries this way. Failure to answer an inquiry is treated as implicit acceptance and produces a chargeback that is, in practice, unwinnable.

The lifecycle, scheme by scheme#

Visa has run two workflows since Visa Claims Resolution took effect in April 2018. The allocation workflow covers fraud, dispute condition group 10, and authorisation, group 11: Visa allocates liability up front from data it already holds, such as whether the transaction was authenticated or authorised, so there is no classic representment and the merchant’s evidence is carried into a pre-arbitration case filed by the acquirer. The collaboration workflow covers processing errors, group 12, and consumer disputes, group 13, and works the way most people picture it: chargeback, representment by the acquirer, then, if the issuer is unpersuaded, an issuer-initiated pre-arbitration, then arbitration.

Mastercard runs one flow: first chargeback, second presentment by the acquirer, pre-arbitration by the issuer, arbitration case filed with Mastercard’s Dispute Resolution Management team. The mechanics are executed in Mastercom, and a party that fails to respond within its window is treated as having accepted liability.

American Express is a three-party scheme and its own issuer, so there is no inter-bank adjudication to have: inquiry, chargeback, a single opportunity to respond, and the decision is final. There is no arbitration stage.

Reason codes, and what each one actually demands#

The reason code is the question. Evidence that does not answer it is not weak evidence but irrelevant evidence, and submitting a great deal of it is the commonest self-inflicted loss in the trade.

Visa’s conditions and issuer time limits, as published at the time of writing in August 2026:

Mastercard’s principal codes, with the windows published at the time of writing:

The evidence that answers each family is different in kind.

For fraud in a card-absent environment, Visa 10.4 and Mastercard 4837 and 4863, the merchant must show that the person who benefited was the cardholder: delivery to the cardholder’s own billing address, a signature, an account with a login history predating the order, prior undisputed purchases from the same device or address, and any usage log showing the account was used after the purchase. Nothing about product quality is relevant, and a successful 3-D Secure authentication usually disposes of the case before any of it matters.

For non-delivery, Visa 13.1 and Mastercard 4853, the merchant must show delivery, or that the promised delivery date has not yet passed. A courier record placing the parcel at the cardholder’s address, ideally with a signature or a photograph and a geolocation, ends it; for digital goods, an access log with timestamps and IP address.

For not as described or defective, Visa 13.3, the burden is descriptive: the listing as it stood on the day, photographs, specifications, the terms the customer accepted, and evidence that what shipped matched the listing. Proof of delivery is worthless here.

For cancelled recurring or subscription disputes, Visa 13.2 and Mastercard 4841, the merchant must produce the cancellation policy as displayed, the record that the customer agreed to it, the absence of a cancellation request before the billing date, and, if a trial converted, evidence of the reminder sent before conversion. For credit not processed, Visa 13.6, the answer is either the refund record with its date and amount or the published policy showing no refund was due. For processing errors, Visa group 12 and Mastercard 4834, the merchant is defending its own systems: authorisation and settlement records, the currency indicator submitted, and proof that a supposed duplicate was two genuine purchases.

The clock#

Deadlines are calendar days, and they are enforced by software rather than by people. A missed deadline is a loss with no appeal.

Two corrections to that table. The merchant almost never gets the scheme’s window, because the acquirer or gateway must reserve time to format and transmit the response, so the internal deadline a merchant sees is typically 7 to 21 days by network and provider. And the issuer’s own decision takes far longer than any response window: 60 to 75 days is normal, and the full lifecycle commonly runs two to three months, with no way to accelerate it other than to give up.

Compelling Evidence 3.0, and its Mastercard counterpart#

The single most useful development for merchants in this decade is Visa’s Compelling Evidence 3.0, effective 15 April 2023 and still in force at the time of writing. It applies to dispute condition 10.4, fraud in a card-absent environment, and it lets a merchant defeat the dispute with history rather than argument.

To qualify, the merchant must identify two previous transactions by the same cardholder that were not disputed, that are at least 120 days old and no more than 365 days old measured from the dispute date, and that share with the disputed transaction at least two of four data elements: user or account identifier, IP address, delivery address, and device identifier or fingerprint. At least one of the two matching elements must be the IP address or the device identifier. Meet that and liability moves back to the issuer.

Two design points matter. This is a rule about instrumentation rather than advocacy, so a merchant that does not record device fingerprints and IP addresses against orders cannot use it however honest its business; and Visa excludes qualifying fraud reports from the monitoring ratio, so the mechanism protects the counter as well as the money.

Mastercard’s answer is the First-Party Trust programme, launched in the United States and, per Mastercard’s announcement of June 2025, expanding to Canada, Latin America and the Caribbean, and Asia Pacific. Merchants supply enriched evidence either during authorisation or afterwards during the dispute, including purchase history, device details, delivery information, identity elements and location, so that an issuer can test a cardholder’s claim at the point it is made. Mastercard’s framing of the problem, citing its 2025 State of Chargebacks report, is that the global cost of chargebacks is forecast to reach 42 billion US dollars by 2028, with nearly half reported as fraudulent.

Who pays, at each stage#

Follow the money for a single £145 card-absent dispute at a UK merchant, on the published fee schedules in force at the time of writing in August 2026.

The £20 received and £20 countered figures are the published UK dispute fees of one large payment provider, effective 17 June 2025 and unchanged at the time of writing; the same provider charges 15 US dollars for each in the United States and 20 euros in the euro area. Traditional UK acquirers typically charge £15 to £25 per chargeback. The countered fee is returned on a win; the received fee, at that provider and most others, is not.

Arbitration is where the economics become absurd on purpose. The schemes’ fee schedules are contractual documents and are not published, which is easy to demonstrate: Visa’s own public rulebook, the Visa Core Rules and Visa Product and Service Rules in the edition of 18 April 2026, sets out the whole dispute resolution machinery, arbitration and compliance included, without stating a single amount, referring the reader instead to the applicable fee schedule, and Mastercard’s Chargeback Guide reaches members through its technical resource centre rather than open publication. The shape of the charging is nonetheless consistent across every account of it: a filing fee paid by the party that escalates, a substantially larger fee borne by the party that loses, and a separate penalty for technical violations such as filing outside the rules or in the wrong format. Industry sources consistently put the minimum exposure on a losing party at several hundred US dollars, with case filing and ruling amounts commonly quoted between 500 and 600 US dollars and technical violation penalties of 100 to 250 US dollars, and those sources disagree with one another at the edges. Treat the range as the settled view of the trade rather than as a verified figure, because no publicly available primary specification confirms it.

The consequence is not in dispute even where the numbers are. Spending six hundred dollars to contest £145 is irrational, so arbitration is almost never used at ordinary retail values. It exists as a threat that disciplines the earlier stages, and as a remedy for large-ticket disputes: travel, freight, professional services, luxury goods.

Ratios, and the programmes that end merchant accounts#

This is the part that decides whether a business survives, and the arithmetic differs between schemes in ways that catch people out.

Visa replaced the old Visa Dispute Monitoring Programme and Visa Fraud Monitoring Programme with a single Visa Acquirer Monitoring Programme, VAMP, which took effect on 1 April 2025 with an advisory period that Visa’s published fact sheet ended on 30 September 2025.

The VAMP ratio is the count of TC40 fraud reports plus the count of TC15 disputes, divided by the count of settled transactions, measured monthly on card-absent volume. Three features of that formula matter enormously. Fraud reports count even though no money moved and no dispute was ever raised. A transaction appearing in both the TC40 and TC15 feeds is counted twice. And the denominator is settled transactions, so declining suspicious authorisations does not flatter the ratio, though it does keep fraud out of the numerator. Excluded from the count are cases resolved through pre-dispute products and fraud reports qualifying for Compelling Evidence 3.0.

On the thresholds in force at the time of writing in August 2026, a merchant is identified as excessive at a VAMP ratio of 1.5 per cent in Asia Pacific, Canada, the European region, the United States and Latin America and the Caribbean, and at 2.2 per cent in Central Europe, the Middle East and Africa. The 1.5 per cent figure is recent: Visa cut it from 2.2 per cent on 1 April 2026. A minimum count applies before a merchant is assessed at all, 1,500 combined items a month in most regions, or 150 items and 75,000 US dollars in CEMEA. At acquirer portfolio level the thresholds are far tighter: 0.5 per cent above standard, 0.7 per cent excessive.

Visa monitors enumeration, meaning card-testing traffic, separately: excessive monitoring applies at an enumeration count of 300,000 and a ratio of 20 per cent, with no fines attached.

Visa assesses a per-item fee on identified merchants but does not publish the schedule in its public fact sheet. Industry reporting at the time of writing puts the fee at 8 US dollars per fraud report or dispute at the excessive tier and 4 US dollars at the above-standard tier, with a three-month grace period for a first identification in a rolling twelve months. Treat those as industry reports rather than primary citations.

Mastercard operates the Excessive Chargeback Programme in two tiers, and calculates its ratio differently: chargebacks received in the current month divided by the transaction count of the preceding month. That definition punishes sharply seasonal merchants, whose January disputes are divided by a December that has ended.

The assessment schedules published at the time of writing, in US dollars or euros per month, count from the first month above threshold:

From month four an issuer recovery assessment is added: 5 US dollars or euros for each chargeback above 300 in the month, so a merchant identified in month four with 400 chargebacks pays 5,500 on top of nothing else. Exit requires three consecutive months below the threshold, and the schedule then resets to month one rather than resuming where it stopped. Mastercard counts a chargeback regardless of any prior refund, of liability shift, and of the outcome.

Mastercard’s separate Excessive Fraud Merchant programme targets a different failure: high fraud combined with an unwillingness to authenticate. On the criteria published at the time of writing, all of the following must hold in a month: at least 1,000 Mastercard e-commerce payments; net fraud chargeback volume above 50,000 US dollars, or 15,000 Australian dollars in Australia; a fraud chargeback rate above 0.5 per cent, or 0.2 per cent in Australia; and 3-D Secure used on no more than 10 per cent of Mastercard payments in non-regulated countries or 50 per cent in regulated ones. Fines run from 500 US dollars in month two to 100,000 US dollars from month nineteen. Where both are breached, this programme takes precedence but the Excessive Chargeback Programme clock keeps running underneath.

Two regional regimes deserve naming. In the United States only, Visa’s Secure Excessive Fraud Programme captures merchants whose fraud on domestic 3-D Secure authenticated Visa transactions reaches 75,000 US dollars and 0.9 per cent in a month; there is no fine, but the merchant loses the liability shift on domestic authenticated transactions until it fully exits, which is heavier than any fine. In Australia, the AusPayNet card-not-present framework works quarterly from 50,000 Australian dollars of fraud chargebacks and a 0.2 per cent ratio, escalating to a mandate that every card-not-present transaction be sent for strong customer authentication.

The lists, which are the real sanction#

Fines are survivable. The terminated merchant files are not, and they operate on thresholds far below the monitoring programmes.

Mastercard’s MATCH, the Mastercard Alert to Control High-risk Merchants system, is a shared database of merchants whose acquiring relationship was terminated. Every processor must check it when onboarding, and must add a terminated merchant meeting the criteria within one business day. Eleven of the reason codes are qualitative, covering laundering, collusion, illegal transactions, PCI DSS non-compliance and the rest. Two are quantitative, and as published at the time of writing they are stark. Reason code 4, excessive chargebacks, means Mastercard chargebacks in any single calendar month exceeding 1 per cent of Mastercard sales transactions in that month and totalling 5,000 US dollars or more, with no minimum chargeback count. Reason code 5, excessive fraud, means a fraud-to-sales dollar volume ratio of 8 per cent or more in a calendar month with ten or more fraudulent transactions totalling 5,000 US dollars or more. Note the 1 per cent: below every monitoring threshold in this chapter, and indifferent to whether the chargebacks were later reversed or won. Records remain for five years and are then purged automatically; a processor can remove an entry only if it listed the merchant in error, or where the listing was for PCI DSS non-compliance since remedied. Mastercard does not assess the accuracy of listings.

Visa’s equivalent is the Visa Merchant Screening Service. Its two quantitative reason codes, as published at the time of writing, are excessive disputes, meaning a dispute count of 1,000 and a 1.8 per cent dispute-to-sales amount ratio in any single month, and excessive fraud, meaning 250,000 US dollars of fraud at the same 1.8 per cent ratio; both are qualified by the merchant having failed to remediate adequately.

Neither list is supposed to be a blacklist. Both function as one, because most acquirers decline an application that produces a hit.

Why a merchant refunds a case it would win#

Everything above assembles into two decisions taken at different moments, and only the second is about winning. Confusing them is why merchant behaviour looks like cowardice from outside.

The first decision comes before any dispute exists, when an early fraud warning or a pre-dispute alert arrives. Refunding costs the whole sale, £145, with certainty. Doing nothing risks the dispute, and on the processor analysis cited earlier roughly 80 per cent of ignored early fraud warnings become disputes: a non-refundable £20 fee, the £145 if lost, the stock, and one tick on a counter that no later victory removes. Hence the published guidance of one large processor at the time of writing, which is to refund proactively where the charge is roughly at or below your dispute fee and generally not to bother above about 135 per cent of it. For a UK merchant paying £20, that means refunding the small ones and standing and fighting on the rest.

The second decision comes after the dispute exists, and by then the counter has already moved. Refunding now achieves nothing: the merchant cannot refund outside the dispute process while it is open, and a refund would not remove the dispute from any monitoring count. So the only remaining question is money. Fighting the £145 case costs the £20 countered fee, returned on a win, plus perhaps £30 of loaded staff time to assemble the evidence properly. At a 40 per cent win rate, realistic for a well-evidenced delivery case, the expected recovery is £66 against about £42 of expected cost, and fighting pays. At a 15 per cent win rate, realistic for a subscription dispute with no usage logs, the expected recovery is £25 against £47 of cost, and it does not.

The behaviour looks irrational because observers watch the second decision while the merchant is making the first. A business inside a monitoring programme, or near one, is not optimising individual cases at all. Its objective is the three consecutive months below threshold that end the programme, and the only lever on the numerator is stopping disputes before they are filed. Hence alerts that cost more than the sale they cancel.

Three circumstances justify fighting regardless. Where the value is large enough that the fee stack is immaterial. Where the case is a template, a cancellation policy or a delivery process that will generate hundreds of identical disputes, so winning once fixes evidence practice for all of them. And where the merchant is comfortably below every threshold, so the counter is not binding and money is the only consideration.

The United Kingdom layer#

British merchants and issuers operate under scheme rules and statute simultaneously, and the two do not align.

Regulation 76 of the Payment Services Regulations 2017, quoted earlier for its next-business-day deadline, also requires the account to be restored to the state it would have been in, with a value date no later than the original debit. That obligation is owed by the bank to its customer and is independent of the chargeback. Regulation 77(6) allocates the loss further along: where strong customer authentication was required under regulation 100 but the payee or the payee’s payment service provider did not accept it, that party must compensate the payer’s payment service provider for the sums it paid out under regulation 76. That is the statutory backbone of what merchants experience as the 3-D Secure liability shift, and it is why a UK merchant that suppresses authentication is not merely accepting scheme liability but a regulatory one.

Section 75 of the Consumer Credit Act 1974 is the other statutory route, and often the better one: joint and several liability of the credit card issuer with the supplier, on the terms set out earlier, with a six-year limitation period rather than a 120-day scheme deadline. Chargeback reaches debit cards and small values that section 75 does not; section 75 reaches consequential loss and long-dead purchases that chargeback cannot. A bank handling a complaint properly considers both.

The Financial Ombudsman Service sits above all of it. Its approach for consumers is described earlier in this chapter; for a merchant complaining about its acquirer, its stated position is that where the acquirer acted in accordance with the merchant services agreement the complaint is unlikely to be upheld, which returns the merchant to the contract it signed.

For scale, UK Finance’s Annual Fraud Report published on 15 June 2026 recorded £1.28 billion of payment fraud losses in 2025, up 4 per cent, of which £703.4 million was unauthorised fraud, down 5 per cent. Remote purchase fraud, the card-absent category that generates most fraud disputes, was £423.5 million across 3.2 million cases: a 3 per cent rise in value on a 13 per cent rise in cases. Falling average values on rising case counts is precisely the pattern that pushes dispute counts up while losses hold steady, and it is counts that the monitoring programmes measure.

What is moving under this chapter#

Three things will date this chapter first, and a reader in 2028 should check all three before quoting any of it.

The thresholds are moving downwards. Visa’s merchant excessive threshold fell from 2.2 per cent to 1.5 per cent on 1 April 2026, having been set only a year earlier, and the acquirer-level figures moved on their own schedule. There is no reason to expect that to be the last revision.

The adjudication is being automated. On 1 April 2026 Visa announced six dispute services: the Visa Dispute Resolution Network for pre-dispute handling, in pilot with general availability planned for late 2026; Visa Dispute Recovery Manager, which automates representment with generative responses and win-prediction scoring; Dispute Intelligence and Dispute Doc Analyzer, which summarise and score cases for issuers and acquirers; Visa Dispute Case Manager; and the Order Insight update that carries Compelling Evidence 3.0 into the pre-dispute conversation. Visa’s stated context was 106 million disputes processed globally in 2025, a 35 per cent increase since 2019. When both sides of a dispute are assembled and assessed by models, the evidence that wins will be structured data captured at the time of the order rather than prose written afterwards, which is a change in what merchants must build, not merely in what they must write.

And the ground is shifting under the word fraud. Compelling Evidence 3.0 and First-Party Trust both let history speak for the merchant, and both reward instrumentation over argument. The merchant that survives the next revision of these rules is not the one with the best dispute writer. It is the one that recorded the device, the address, the login and the delivery at the moment of sale, and can prove eighteen months later that the same person came back.

49.98 Common wrong ideas#

Wrong: A chargeback is a legal right the cardholder exercises against the shop. Right: In the United Kingdom it has no statutory basis at all; it is a private contractual mechanism of the card schemes, the cardholder asks their bank to raise one, and the bank decides whether to.

Wrong: A chargeback reverses the original payment, so the same amount comes back. Right: It is a new financial transaction with its own clearing cycle, value date and conversion, so a sale that settled as 100 euros can return as 107.86 euros, and issuers can also bundle, split, or overlap with a refund already given.

Wrong: Winning the dispute puts the merchant back where it started. Right: The fee for receiving a dispute is normally non-refundable, the issuer can escalate and take the money a second time, and no victory removes the tick from the monitoring counter.

Wrong: Good delivery proof wins most disputes. Right: Proof of delivery decides non-delivery cases, helps in card-absent fraud only insofar as it places the goods with the cardholder, and is worth nothing at all under “not as described”, where nobody disputes that the parcel arrived.

Wrong: Refunding the customer once the dispute exists gets the chargeback off the books. Right: The count is taken when the dispute is filed, the merchant cannot refund outside the process while it is open, and a refund at that stage pays twice for nothing.

Wrong: Disputes filed as fraud are fraud. Right: A large share of card-absent fraud disputes are first-party misuse, a subscription harder to cancel than to start, a family member, an unrecognised statement descriptor, which is why both schemes have built programmes aimed squarely at it.

Wrong: Declining suspicious authorisations improves the VAMP ratio. Right: The denominator is settled transactions, so declines do not flatter the ratio; they only keep fraud out of the numerator.

Wrong: The fines are the thing to fear. Right: Fines are survivable; the terminated merchant files are not, and MATCH reason code 4 bites at 1 per cent of monthly Mastercard sales totalling 5,000 US dollars, below every monitoring threshold and indifferent to whether the chargebacks were later won.

Wrong: Arbitration is where a merchant with a strong case gets justice. Right: With case filing and ruling amounts commonly quoted between 500 and 600 US dollars, it is irrational at ordinary retail values and exists as a threat that disciplines the earlier stages.

Wrong: The merchant gets the scheme’s full response window. Right: The acquirer or gateway must reserve time to format and transmit the file, so the internal deadline is typically 7 to 21 days, and a missed deadline is a loss with no appeal.

49.99 Chapter summary in 20 lines#

1. When you tap a card your bank pays the shop and bills you afterwards, so it is the bank’s money that moved and the bank that has both the standing and the plumbing to reverse it.
2. A chargeback therefore runs backwards from the ordinary intuition: the money is seized first and the argument happens afterwards.
3. For the merchant the first news is a debit of the sale plus a non-refundable dispute fee, arriving weeks or months after the goods were posted.
4. Contesting a dispute is called representment, costs a further fee refundable only on a win, and is decided by a stranger at the issuing bank two to three months later.
5. The question that stranger answers is fixed by the reason code the issuer chose, not by the merchant’s account of what happened.
6. Evidence that does not answer the reason code is irrelevant rather than weak, and submitting a great deal of it is the commonest self-inflicted loss in the trade.
7. Visa runs allocation and collaboration workflows, Mastercard a single flow ending in arbitration, and American Express, being its own issuer, decides once and finally.
8. Deadlines are calendar days enforced by software, and the merchant’s internal window is shorter than the scheme’s because the acquirer must reserve time to transmit.
9. Chargeback in the United Kingdom is a scheme mechanism with no statutory basis; section 75 of the Consumer Credit Act 1974 and regulation 76 of the Payment Services Regulations 2017 sit alongside it and behave differently.
10. Because a chargeback is a new transaction, foreign exchange, bundled subscriptions, partial disputes and earlier refunds routinely make the returning amount differ from the original.
11. Winning is neither final nor free: issuers escalate, fees stay spent, and money moves outside the lifecycle through issuer credits and debits months later.
12. Every month the schemes divide a merchant’s disputes by its transactions and compare the answer with a threshold.
13. Visa’s VAMP ratio adds fraud reports to disputes over settled card-absent transactions, so events that moved no money count, and a transaction appearing in both feeds counts twice.
14. Mastercard divides this month’s chargebacks by last month’s transaction count, which punishes sharply seasonal merchants.
15. The counters are indifferent to outcomes by design, because what the schemes measure is prevention and waiting months for every result would make monitoring useless.
16. Crossing a threshold brings escalating monthly fines, a written remediation plan, and eventually the end of the acquiring relationship.
17. The real sanction is the terminated merchant lists, MATCH and the Visa Merchant Screening Service, whose quantitative triggers sit below every monitoring threshold and whose entries last five years.
18. That is why a merchant refunds a case it would win: a refund arriving before the dispute is a refund, and only one arriving afterwards is counted.
19. Hence the pre-dispute layer, Order Insight, Rapid Dispute Resolution and Ethoca, whose commercial value is the removal of a tick from a counter rather than the rescue of a sale.
20. And hence the direction of travel: Compelling Evidence 3.0 and First-Party Trust reward data captured at the moment of the order rather than prose written afterwards.

Sources, all consulted in August 2026: Visa’s Acquirer Monitoring Programme fact sheet, its Compelling Evidence 3.0 merchant readiness paper and its press release of 1 April 2026 on dispute resolution services; Verifi on Order Insight and Rapid Dispute Resolution; Mastercard’s Chargebacks Made Simple guide, its Excessive Chargeback Programme guide and its June 2025 First-Party Trust announcement; Stripe’s public documentation on disputes, monitoring programmes and high-risk merchant lists, which quotes the MATCH and VMSS standards verbatim, and its June 2025 dispute pricing notice; regulations 76, 77 and 100 of the Payment Services Regulations 2017 and section 75 of the Consumer Credit Act 1974 on legislation.gov.uk; UK Finance on chargeback and section 75 and its Annual Fraud Report of 15 June 2026; Financial Ombudsman Service pages on disputed transactions and goods bought on credit; and trade references from Chargeback Gurus, Chargeflow, Braintree and the Merchant Risk Council where scheme documents are not public, used only where consistent and flagged where not.