Chargeback or refund: what the difference costs you

Both end with the customer getting their money back. That is where the similarity stops. A refund is a payment you make; a chargeback is a payment taken from you, through a process you do not control, with a fee attached and a counter that the card schemes watch. Treating the two as interchangeable is how a merchant ends up in a monitoring programme while believing its service is generous.

The two paths

A refund is a new transaction that you initiate, in the opposite direction to the original. You decide the amount, you decide when, and the customer sees the money come back on the card that paid — usually within a few working days, depending on their issuer. In your ledger it is a credit against the original payment. The scheme is not involved beyond moving the money.

A chargeback starts with the cardholder contacting their issuer rather than you. The issuer raises a dispute under a reason code, debits the acquirer, and the acquirer debits you. The funds leave your balance before anyone has looked at your evidence. You then have a fixed window to respond, and the decision at the end of it is made by the issuer and, if it goes that far, by the scheme.

The difference that matters operationally: a refund is a business decision with a cost you set, and a chargeback is an administrative event with costs, deadlines and consequences you do not set.

What a chargeback costs beyond the amount

Four things, and only the first is obvious.

The transaction value. Taken back in full, including the part that was shipping, tax or a third-party fee you have already paid onward.

The chargeback fee. Charged per case by your provider, generally whether you win or lose, because the acquirer incurs cost handling it either way. On a low average order value, the fee can exceed the transaction.

The original processing cost. The interchange and scheme fees on the initial authorization are not always returned, and the acquirer's margin on it usually is not.

The staff time. Assembling evidence for one dispute takes a person between fifteen minutes and an hour, depending on how well your systems record delivery, access and consent. At any volume this is the largest of the four.

Compare that with a refund, which costs you the transaction value, the processing fee on the original payment, and in most schedules a small refund fee. No deadline, no evidence, no counter.

The dispute lifecycle

The names differ slightly between the schemes, but the shape is the same.

1. Retrieval or inquiry. Some issuers ask for information before raising a formal dispute. Not every case has this stage, and where it exists, answering it well is the cheapest possible outcome.
2. Chargeback. The funds are debited. You are notified with a reason code, a case identifier and a deadline.
3. Representment. You submit evidence that the payment was valid. The deadline is measured in days from the chargeback, not from when you noticed it, and a missed deadline is a loss with no appeal.
4. Second chargeback or pre-arbitration. The issuer can come back once more with new information.
5. Arbitration. The scheme decides. It charges a fee for doing so, and that fee is usually larger than the transaction in dispute. This stage is for a case with a principle behind it, not for one order.

Two clocks matter throughout: the cardholder's window to raise a dispute, which is long and measured in months from the transaction or the expected delivery date, and your window to respond, which is short. Build your process around the short one.

How the schemes count you

Visa and Mastercard each run dispute and fraud monitoring programmes. Both work the same way in outline: your dispute count and fraud count are measured against your transaction count over a calendar month, per merchant identifier, and once a ratio crosses a published threshold you enter a programme with escalating remediation requirements and per-dispute fees.

The specific thresholds are set by each scheme, published in their rules, and revised from time to time. Do not work from a number you read in a blog post, including this one. Ask your provider for the current published threshold that applies to your merchant identifier and your region, and ask it to show you your current ratio against it in your dashboard, monthly, without you having to request it.

Three details about the arithmetic are worth knowing, because they explain outcomes that otherwise look unfair.

Refunds do not reduce the ratio. The denominator is transactions, not net revenue. Refunding generously does not lower your dispute rate; it lowers your revenue.

The month of the dispute is not the month of the sale. A dispute raised in March against a January sale counts in March's numerator against March's transaction count. A merchant with falling volume and a stable dispute count has a rising ratio, which is why a quiet month after a busy one is the classic way to cross a threshold.

Fraud and non-fraud are counted separately. A merchant can be inside the dispute threshold and outside the fraud threshold, or the reverse, and the remedies differ.

When to refund instead of fighting

A representment costs you staff time and the chargeback fee, and it does not remove the dispute from the numerator even when you win. The dispute is counted when it is raised.

That leads to a rule most experienced merchants converge on: prevention beats representment, and representment is worth it only where the evidence is strong and the amount justifies the work.

Prevention, in order of effect:

• A descriptor the customer recognises on their statement. A large share of disputes coded as unrecognised are the customer genuinely not recognising a trading name that appears nowhere in their purchase.
• A refund path that is easier than a dispute. If cancelling takes three emails and a dispute takes two taps in a banking app, you have chosen for them.
• Delivery and access records attached to the transaction, automatically, at the time. Evidence assembled six weeks later from four systems is weak evidence.
• Answering retrieval requests, where the issuer offers that stage.
• Authentication on the risky slice of your traffic, which moves fraud-coded liability to the issuer.

Evidence that wins

A representment is read quickly by someone with a queue. It should be short, specific to the reason code, and made of records created before the dispute existed: the authorization and authentication result including the cryptogram and ECI, the IP address and device at purchase, the delivery confirmation with a signature or a tracked timestamp, the access or usage log for a digital product, the terms the customer accepted with the version and the timestamp, and the full history of your correspondence with them.

What does not win: a long narrative about your business, a screenshot of an internal system without context, or an argument that the customer is acting in bad faith without a record that shows it.

The operational summary

Refund when the customer is right, when the evidence is weak, or when the amount does not justify the work. Fight when you have contemporaneous records and the amount is worth the fee. Watch the ratio monthly rather than the count, because the denominator moves. And keep the dispute rate in the same report as the approval rate: tightening one at the expense of the other is the most common way to make a payment operation quietly worse.