Refund

Concept

Money a company sends back to a customer on its own initiative, in full or in part, under reasons and deadlines fixed in advance by policy.

Why it exists

Once software started being paid for up front, there was always some amount of unused time sitting on the books. A physical product goes back in its box and travels the other way. A service paid for a month or a year ahead has no object to return. A customer who paid for twelve months, used two, and wants out leaves behind a question about how much of that money is theirs.

Early on, the answer depended on whoever picked up the ticket. Two customers making the same request in the same week get different answers — one gets the remaining ten months back, the other is pointed at the terms of service and refused. If both of them are in the same community, that difference is public within a day. The money handed back costs less than the discovery that there was no rule.

There is also a line the company doesn't get to draw. Consumer protection rules give individual buyers a right to withdraw from a contract inside a set window, so a request that arrives in that window has to be honored even if the terms of service say refunds are unavailable. That is why refunds ended up as a defined process rather than a series of judgment calls. Separating what the law decides from what the company decides is what makes the answer repeatable.

How it works

A refund means checking the reason, settling the amount, sending the money back along the path it arrived on, and taking that much revenue off the books.

  1. Sort the reason. A duplicate charge or an incorrect invoice is the company's error. A customer who expected something else and wants to stop is a request. A withdrawal inside the legal window is neither. The three run on different tracks: the first is refunded in full as soon as it's confirmed, the second within whatever the policy allows, the third regardless of what the policy says.
  2. Settle the amount. Policy decides whether the answer is everything or only the unused remainder. A customer who paid $900 for a year and used two months gets $750 back if the rule is the remaining ten months, and the whole $900 if the request lands inside a 14-day window where usage doesn't matter.
  3. Send the money back. The default is a reversal through the payment gateway to the card that was charged, which typically takes three to five business days to appear on the customer's statement. When the charge is too old to reverse, it goes out as a bank transfer and that exception gets recorded.
  4. Correct the books. Revenue already recognized comes down by that amount, and if a tax invoice was issued it has to be reissued so the value-added tax matches. Returning the money and leaving the paperwork alone shows up at the next filing.
  5. Record the reason as a code. One at a time these are support notes; in aggregate they are product signal. If half of a month's refunds trace to the same feature, the problem isn't in billing.

Fees are the part that surprises people. The card fee paid on the original charge usually does not come back. Collect $75 and pay $2 in processing, then refund it in full, and the company sends out $75 while keeping negative $2. At a few refunds a month that rounds to nothing; at several hundred it becomes a line worth watching on its own.

Where it shows up

Refund policy gets pulled up in three places.

  1. The support queue. When a request arrives, the rep reads the policy rather than deciding. How many days the full refund window runs, what happens after it, and how already-used features are treated all have to be written down for the answer to come out the same twice.
  2. Reversing charges that shouldn't have gone out. Duplicate approvals from repeated dunning attempts land here, as does a period billed while an outage made the service unusable. This category is better found by the company than reported by the customer — a refund that arrives before the complaint reads as competence, and the same refund a week after an angry ticket does not.
  3. Reading the numbers. Refunds pull back money already collected, so gross charges and what the company actually keeps diverge. Most teams watch refunds as a percentage of charges, weekly. Under one percent is usually routine; a jump to two percent means the cases are clustered in one plan or one feature, so the reason codes come first. Monthly recurring revenue (MRR) has to exclude refunded charges too, or next month's projection starts off wrong.

How it sounds at work

"The duplicate charge was refunded in full today — three to five business days to show up on the card."

A confirmed company error, processed without a judgment call.

"We're past the 14-day mark, so not the full amount, but I can refund the unused portion."

The full-refund window has closed, so this goes out as a partial refund.

"Refunds were 1.8 percent this week. We need to see which plan they came from."

The rate is above normal, which suggests the cases are concentrated somewhere.

Pagedeck runs a two-line policy: full refund within 14 days of payment, unused portion after that. Sofia knows both lines by heart and answers on the call. Only one case needs a decision — how much to return for time lost to an outage — and there Theo confirms the downtime and Grace approves the amount.

Last month a checkout bug charged twenty-two customers twice. The list was pulled and refunded before any tickets arrived, with the notification email going out first.

What people get wrong

The first assumption is that a refund and a canceled payment are the same event. The customer gets money back either way, so the distinction looks academic. But voiding an authorization on the day of the charge removes the transaction from the statement entirely and returns the processing fee. A refund issued after the card network has settled is a separate transaction: the statement shows a charge and a credit as two lines, and the company eats the fee. Explain it wrong and the customer sees two debits and writes back.

The second is that refusing a refund keeps the money. On a ledger it looks that way. But a refused customer still has the option of disputing the charge with their card issuer, and once it goes that way the decision moves from the company to the issuer. The company ends up assembling evidence, paying a per-case fee, and if the rate crosses a threshold, facing higher processing fees or a canceled contract. The $75 kept comes back priced higher.

Easily confused with

TermWho starts itHow it differs
ChargebackThe customer, by disputing the charge with the card issuerThe issuer decides and the company argues its case. Each one carries a fee, and a high rate puts the payment processing contract itself at risk

Where to go next

Last updated 2026-09-16