Why a card refund is slower than the purchase was
A customer who paid in two seconds and waited nine days for the refund is not being lied to by the merchant, which is the assumption everyone reaches first. The asymmetry is real and structural: the two movements of money are not the same operation reversed. One is an authorisation, the other is a fresh transaction that has to clear and settle before it can appear anywhere.
The confusion comes from the word refund, which sounds like undoing. Almost nothing in card payments is undone. Understanding the sequence makes the delay predictable, which is most of what an unhappy customer actually wants.
The purchase was fast because nothing moved. The refund is slow because something does.
What happened when you paid
At the moment of purchase, an authorisation request travelled from the terminal or the checkout to the acquirer, through the scheme, to the issuer, and back. The issuer checked the account and placed a hold. That round trip is the two seconds everyone remembers, and no money moved during it: the balance available to the cardholder fell, the merchant received a promise.
The actual movement came later, in clearing and settlement, on the acquirer's cycle. Transactions are batched, submitted, netted between institutions and settled, typically over following business days. By the time funds reach the merchant's account the customer has long since walked out, which is why nobody ever notices this half of the process.
What happens when you are refunded
A refund enters that same pipeline from the start, as a new transaction in the opposite direction. The merchant submits it; it joins a batch; the batch clears; the issuer receives it and applies it to the account. There is no hold to release, because the hold was consumed when the purchase settled. Each hop runs on business days, and consecutive weekends and public holidays add to the total in a way no customer service script explains well.
Three further things stretch it. Merchants often submit refunds in a nightly batch rather than immediately, so a refund agreed on Monday morning may not leave until Monday night. Acquirers apply their own processing windows. And issuers differ in how quickly a received credit becomes visible, which is why two people refunded by the same merchant on the same day see it on different days.
The cancellation that is genuinely fast
There is one case where the fast path exists, and it is worth asking for by name. If the original transaction has not yet cleared, the merchant may be able to cancel the authorisation rather than issue a refund. Releasing a hold is quick, and the amount simply stops being reserved. The window for that is short — it closes when the batch goes — which is why the first hour after a mistaken payment is worth more than the next three days of emails.
After clearing, the authorisation no longer exists and only a refund is possible. A merchant saying so is not being obstructive; a merchant saying so twenty minutes after the purchase might be.
What the merchant controls and what it does not
The merchant controls when the refund is submitted, and that is the part worth pressing on, because it is the only part where a human decision is being made. Everything after submission is the clearing cycle, and the merchant has no more influence over it than the customer does. A merchant that can give the date it submitted the refund has given you a verifiable fact, and from that date the remaining wait is a matter of ordinary business days.
There is a commercial footnote that explains some of the reluctance. In most card arrangements the merchant pays fees on the original transaction that are not returned when the sale is refunded, and in some arrangements a refund carries its own small charge. Refunds cost merchants money beyond the sale value. That is a reason some firms are slow to submit them, and it is a different problem from the clearing cycle — which is exactly why it helps to know which part of the delay is physics and which part is choice.