Losing quickly is sometimes the cheap outcome

Concede a dispute whenever the fee is smaller than the hours and the likely bad review a contest will cost, which in this trade is most of the time. The decision is arithmetic, not principle, and running it honestly means conceding more often than feels comfortable.

Being right is not one of the inputs.

The four costs of contesting

Put a number on each before you decide.

The fee. Known, and usually the smallest number on the list. Card disputes add their own charges on top: Stripe's dispute documentation says a countered dispute incurs a separate countered fee, returned only if you win, while the dispute received fee is not returned either way, and the bank's decision can take up to three months.

Your hours. Assembling a chronology, waiting, replying to a follow-up question, re-reading the thread at midnight because you are annoyed. Two to four hours is typical for a case you take seriously, and it is not billable to anyone.

The probability of winning. Not your confidence in the merits. Your honest read of whether the record shows a stated scope and a delivery inside the platform - what evidence carries weight is the whole of this variable, and if the answer is no, the probability is low regardless of what happened.

The review. A contested dispute frequently produces a public review from a person who has just lost an argument. A refunded one frequently produces silence.

That last cost is the one people leave out, and it is often the largest. A visible negative review sits on the listing for months and suppresses bookings from people who never mention it.

The arithmetic, with a worked illustration

These figures are an illustration, not data, and the point is the shape rather than the numbers.

Take a fee of 40 in whatever currency you work in, and value your own hour at 25 because that is roughly what the work pays you.

At a 70 percent chance of winning, contesting is still worse than conceding before the review effect is counted at all. The break-even is not a probability, it is a fee: contesting only pays when the amount is large enough to cover several hours of your time, which for most rate cards means it almost never pays on a single job.

Two things change the answer.

A large or bundled job, where the fee is a multiple of the hours involved. A buyer whose claim is provably false in the platform's own record, where the case takes twenty minutes rather than three hours because the evidence is already assembled.

Everything else is a refund you are delaying.

The cases where you contest anyway

Three, and they are narrow.

When the pattern matters more than the case. A buyer who has disputed you twice is not a customer, and letting the second one pass invites a third. Contest, and stop taking their work.

When conceding admits something specific. If the claim is that you failed to deliver at all and you did, an uncontested refund is a record of non-delivery sitting against your account. Platforms track dispute outcomes internally and use them, so a false claim about performance is worth twenty minutes even at a small fee.

When the chargeback clock is still running. Refunding inside the platform after a card dispute has been opened can leave you paying twice, because the two processes do not know about each other. Holds, chargebacks and the order they happen in covers the sequencing; the short version is that once the card network is involved, act inside the platform's guidance and not on your own initiative. The window is short and final: Stripe gives usually 7 to 21 days to respond depending on the card network, allows one submission of evidence, and treats no response as a loss.

Conceding well

If you are refunding, refund fast and say almost nothing.

One sentence, no justification, no restatement of who was right. "Refunding this in full - sorry it was not what you were after."

The temptation is to attach an explanation of the misunderstanding. Resist it, because an explanation is an invitation to continue the argument, and the entire value of conceding is that the matter ends today. When a refund is the right instrument generally sits next to this, and the two decisions differ only in whether a formal process has already started.

Then close the loop on the cause. Almost every dispute in this trade traces to one of two things: a scope nobody wrote down, or a deliverable that did not match what the buyer pictured. Both are fixed before the job, not during the argument.

A specific version of the second one is worth naming. A significant share of disputes are not about the work at all but about a number - a buyer who expected a score to mean something other than what it meant, and reads a low one as a failure of service. That expectation is set long before you: the presentation of scores in the free tools is where most buyers learn to read a number, and what an automated score can actually be held to is the honest bound on it. Rate Cock's own explanation of what a score out of ten means is the page to send someone who is arguing with the number rather than with you, and sending it is frequently the whole resolution.

The rest is buyer expectation formed before the booking, which the buyer-side account of how commissions are placed describes better than any post written from this side can. Reading it occasionally is cheaper than winning arguments.