The refund you paid on day 12 cost more than the refund you paid on day 3
2026-09-23
A chargeback gets filed under fraud. A good share of them are caused by the returns policy, and this year they cost considerably more than they did last year.
- Ecommerce
- Payments
- Returns
- Chargebacks
A chargeback is usually filed under fraud. In most of the ecommerce businesses I have worked in, a large share of them were filed under fraud and caused by the returns policy.
The customer returned an item, and then, for a whole week, no action was taken. The tracking information showed that it had been delivered to the warehouse, the portal still indicated that it was awaiting receipt, the money had not been returned, and the help centre provided a form with a five-day response time. Therefore, they contacted their bank, and the bank asked them if they recognised the transaction; the way of getting the money back most quickly was to say that they did not recognise it.
That transaction is now a dispute. It costs more than the refund would have, it costs it in places the returns team does not see, and the cost has gone up sharply this year.
Why the customer picks the bank
Start with the choice the customer is actually making, because it is a rational one and treating it as bad faith leads to the wrong fixes.
They have two routes to the same outcome. One is your support queue, which has an unknown length, an unknown probability of success and requires them to make the case. The other is a single conversation with their bank, in which someone whose commercial interest is in keeping their account happy takes the case off their hands and provisionally credits the money while it is investigated. The effort asymmetry between those two routes is enormous, and it points one way.
Procedural justice research, the line of work associated with Tom Tyler and Allan Lind, found that people's willingness to accept an outcome depends heavily on whether the process that produced it felt fair, and that a process judged unfair produces escalation even where the eventual outcome is identical. This is old literature, and I flag that, as I do with any source of that age. It has held up, and it describes what happens in a returns portal with unnerving accuracy. The customer who escalates is frequently not disputing the outcome at all. They are disputing being made to wait without information.
If loss aversion is taken into account, money which has already been paid and is now held by the retailer is not perceived as a merely pending administrative issue; instead, it is seen as an ongoing loss, the discomfort from it increasing as the delay lengthens rather than remaining constant. The longer the refund stays unresolved, the more value the customer places on having it settled and the less effort they are willing to expend on your channel in order to achieve it.
The reactance then completes the loop: the conditions that come after the purchase, the restocking fee noted at the bottom of the page, and the 14-day period which is only discovered on day sixteen, all count as restrictions on the freedom of action which had not been disclosed at the time the freedom was sold. The natural response in such a situation is to find a way round the restriction, and the bank is that way round.
What the dispute actually costs
The smallest item on the bill is the refund.
The Visa Acquirer Monitoring Programme combined the previous dispute and fraud monitoring programmes into a single measure based on the number of incidents. The VAMP ratio is obtained by summing the number of fraud reports and the TC40 figure together with the number of disputes and the TC15 figure, after which this total is divided by the number of settled card-not-present transactions and the TC05 figure. The excessive threshold for merchants was reduced from 2.2% to 1.5% in the United States, Canada, Europe, the Asia Pacific region and Latin America on 1 April 2026. Acquirers have their own limits, namely 0.7% for excessive and 0.5% for above standard, and in practice many set an internal threshold below the figure published by Visa as a buffer, so that your actual ceiling is determined by what your acquirer states rather than by the figure Visa has published.
Merchants who are at the Excessive level have to pay a fee of $8 for each disputed transaction and are given a three-month grace period in the case of their first breach within any twelve-month period. After the fee come reserves, compulsory remediation and, at the other end, the acquirer deciding that the relationship is not worth the portfolio risk.
For anyone who prepares returns, two aspects of that arithmetic are relevant.
The ratio is based on the number of disputes. Thus a dispute involving £19 is treated in the same way as one involving £900. A business which has a low average order value and a slow refund process can still be in breach simply on the basis of volume, even though its fraud losses appear entirely unremarkable.
Second, the fraud side of the numerator comes from issuer-submitted TC40 reports, which are generated whether or not a chargeback ever follows. A customer who told their bank they did not recognise the transaction has put a mark on your ratio even if they later drop it, and even if you win the representment.
Put numbers on it. A business taking 60,000 card-not-present orders a month reaches the 1.5% Excessive threshold at 900 combined fraud reports and disputes, which is 30 a day. That same business at a 12% return rate is handling about 240 returns a day. One in eight of those customers reaching for the phone during a slow peak is enough to put the acquiring relationship, rather than the refund policy, at the centre of the conversation. The $8 fee is the part that surfaces in a management account. The reserve the acquirer then asks for is the part that surfaces in the cash flow.
So the true cost of that refund paid on day 12 is the refund, plus the representment work, plus the scheme fee once you are over threshold, plus a permanent entry in the numerator of a ratio that decides whether you keep your acquiring relationship. Against paying it on day 3, which costs the refund.
The legal floor, which is lower than most policies think
The Consumer Contracts (Information, Cancellation and Additional Charges) Regulations 2013 set the reimbursement deadline for a cancelled goods contract at 14 days, running from the earlier of two events: the day the trader receives the goods back, or the day the consumer supplies evidence of having sent them back. The reimbursement must use the same means of payment the consumer used, and the trader may not charge a fee for making it.
Go through the trigger once more since it is the section in which policies make a mistake. The clock starts when evidence of the item being sent back is provided. If a customer has a Post Office receipt from day 1, then their 14 days have already begun on day 1, regardless of what the warehouse receipt date indicates.
The 14-day period is the target that most of the retailers I've observed set for themselves, but it is only a maximum, and a business seeking to achieve this maximum will go over it during the peak times, since it is then that the number of disputes is highest and the VAMP ratio is most sensitive owing to the fact that the number of settled transactions is still catching up.
Where the actual fraud sits, and why it is worth separating
Card fraud is still a problem. UK Finance's Annual Fraud Report 2026 shows that losses from remote purchase fraud, where criminals use stolen card details to buy things online, rose to £423.5 million in 2025. This is a 3% increase from the previous year and involved 3.2 million cases, up 13%. While total unauthorised fraud losses fell by 5% to £703.4 million, the number of cases went up by 11% to 3.81 million. This suggests that fraud is shifting to more frequent but lower-value cases, rather than going away.
That divergence between falling losses and rising cases is the same pattern that makes a count-based ratio dangerous. The industry is getting better at stopping the money leaving and is being hit by more attempts, and a merchant's monitoring ratio counts attempts.
The practical consequence is that a dispute reduction programme aimed only at fraud tooling will miss most of what is actually filling the numerator, and a returns programme that never looks at the dispute data will never learn that it is generating them.
What I would actually change
There are three, and they all involve sequencing and not spending.
The refund is based on proof of despatch rather than on the goods being received at the warehouse, at least in the case of the various categories and customer groups for which the return rate is known. The inventory risk can be measured and is generally less than the dispute cost that it eliminates. Moreover, this situation brings the policy in line with the legal trigger rather than going against it.
Fill in the information gap in the middle. The stage between the carrier picking up the package and the warehouse scanning it is the one in which the customer has no insight and experiences the most anxiety, since it is also when the phone call to the bank is made. A tracking status indicating that the package has been received by the carrier, is expected at the warehouse on Thursday, and will be refunded within two working days is almost negligible to create and eliminates most of the reason for having to escalate.
Capture the checkout data your representatives will need before you need it. Visa's Compelling Evidence 3.0 framework lets a merchant defend a first-party misuse claim using matched data elements from prior undisputed transactions, device identifier and IP address among them, and disputes resolved that way do not count toward the ratio. That data has to exist at the moment of the transaction. It cannot be reconstructed once the dispute arrives, which makes this a checkout instrumentation decision rather than a disputes team decision.
And you should measure the link since very few people do. Refund latency and the dispute rate are kept in two separate systems by two different teams and are never plotted against one another. To combine them, you have to spend a whole morning dealing with an order ID, and the only thing that will result in a change to the refund policy is the output, since it transforms a complaint from a customer experience perspective into a point that the finance director already cares about.
The pattern underneath all three is that the disputes team is measured on win rate, the returns team is measured on cost per return, and the person carrying the consequence of both is whoever owns the acquiring relationship. In most retailers, that is finance, who see the fee line and not the returns portal.
Which brings it back to the same place as Tuesday. The cost lands on a seam, and the only person positioned to see both sides of it is the one whose remit covers the whole customer journey rather than one department's slice of it.