Running it
Most reversals start as confusion, not fraud
An unclear scope and a vague delivery are the two ingredients of nearly every dispute.
Guides on Running it: The admin that turns income into a trade
A chargeback is a buyer asking their card issuer to reverse a payment, and it is not the same thing as a refund request. It goes around you, it costs a fee whether or not you win, and a rate high enough to notice can cost you the ability to be paid at all.
The useful thing to know is that outright fraud is a minority of them. The bulk are a buyer who did not recognise a charge, did not get what they thought they were buying, or could not find a faster way to complain.
The four causes, in rough order of frequency
Non-recognition. The buyer sees a line on a statement they cannot place, three weeks after a small purchase, and disputes it. Card networks have a category for this and it is consistently one of the largest.
Expectation mismatch. The delivery was real, and it was not what the buyer pictured. Almost always a scope failure at the start rather than a quality failure at the end.
Non-delivery. Sent to the wrong place, sent late, or sent somewhere the buyer never looked - a message thread they had stopped reading, or a link that expired.
Genuine fraud. A stolen card, or a buyer who intended this from the beginning. Real, and much rarer than the first three.
The first three are all yours to reduce. That is the whole reason this subject is worth a post rather than a shrug.
The descriptor does more work than anything else
The payment descriptor is the text that appears on the buyer's statement, and it is the single highest-leverage thing in this list because it is the only defence against the largest category.
A descriptor that is a company name the buyer has never heard of will generate disputes from honest people. A descriptor that carries the platform's consumer-facing name, ideally with a support contact, will not. Card networks have pushed for clearer descriptors for years, and processors document the field precisely because it moves dispute rates.
You usually do not control it, and you can find out what it says. Ask the platform, or make one small purchase yourself and read your own statement. If it is unrecognisable, say so on your listing: a line telling the buyer what will appear on their statement is worth more than any amount of dispute-handling skill.
There is a second-order effect in this sector. A descriptor that is explicit is a descriptor a buyer will dispute rather than explain to whoever else reads their statement, and that is a category of reversal no evidence will win. Discretion in the descriptor is a commercial feature, not a nicety.
Reducing the other three
| Cause | The intervention | Where it happens |
|---|---|---|
| Non-recognition | Tell the buyer what the statement will say | Listing, and the confirmation message |
| Expectation mismatch | One scope sentence agreed before starting | Before the work |
| Non-delivery | Deliver in the platform's own channel, and say you have | At delivery |
| Fraud | Nothing you do; the platform's controls | Not your problem to solve |
The scope sentence is the intervention with the best ratio of effort to effect in this entire trade, and it takes twenty seconds - the two-sentence confirmation covers exactly what to write.
Delivering in the platform's own thread rather than by email or a file-sharing link matters twice: the buyer finds it, and the timestamped record exists when someone later asks whether you delivered. Evidence that actually wins a dispute is mostly a list of things that only exist if you did them at the time.
Offering a refund quickly when a complaint is legitimate is also a chargeback-reduction measure, not a soft-heartedness. A refunded order has no dispute attached, no fee and no effect on your rate, since Stripe notes the monitoring programmes do not count refunds as disputes, whereas a defended reversal costs the fee even when you win - the judgement about when to do it is in refunds, when to give one.
What a rate actually means
Card networks and acquirers monitor disputes as a ratio of transactions, and exceeding a threshold moves a merchant into a monitoring programme with higher fees and, eventually, loss of processing. Stripe's summary of the card monitoring programmes gives the current shape. Visa's acquirer monitoring programme counts disputes plus fraud reports against transactions, with a non-compliant threshold of 0.5% and an excessive threshold of 1.5% in most regions. Mastercard's excessive chargeback programme starts at 100 chargebacks and a 1.5% rate in a month. The tolerance is measured in fractions of a percent, not in "a few", and the programmes change, so the networks' own documentation is the current word.
On a platform, the merchant of record is usually the platform rather than you, so its ratio is what the networks watch. Yours still matters internally. Platforms track disputes per seller because they inherit the cost, and a seller with a visibly worse rate than the population attracts holds, reserves and eventually removal - which is the account risk described in not putting everything on one platform.
Small transaction sizes make this harder, not easier. The fixed dispute fee can exceed the value of the order, so a single reversal on a cheap job wipes out several of them, which is one more entry in the argument that very low prices are expensive to serve - why the cheapest tier attracts the hardest buyers has the rest.
The part that is about clarity, not payments
Nearly everything above reduces to describing accurately what a buyer will receive, and then sending exactly that.
A delivery that names its own contents - the axes covered, the length, what was and was not assessed - is very hard to dispute, because the buyer has to argue against a document that describes itself. A delivery that is a paragraph of impressions is disputable by anyone who was expecting something else, and buyers frequently were.
That is also why buyers' own expectations are worth understanding directly rather than inferred: the commissioning-side account of what people think they are ordering explains most expectation mismatches before they happen. Where the disagreement is about a number rather than a judgement, it is usually a method disagreement, and the documented conventions settle it faster than any argument about intent. And where the buyer expected the specificity of an automated readout and got prose, the mismatch was in your listing - what an automated score presents and how is the format they had in mind.
Platform dispute processes differ, and knowing yours before you need it is the difference between a reply within the window and a reply after it - Rate Cock's judges page is the entry point for how orders and disputes are handled there.