Running it
The party with timestamps usually wins
Whoever can produce the agreed brief and the delivery record wins, largely regardless of who was right.
Guides on Running it: The admin that turns income into a trade
Disputes are won by a timestamped record of what was agreed and what was delivered, not by the quality of the work. A dispute is a clerical exercise in which someone with a queue of forty cases decides which party has produced a verifiable record of what was agreed and what was delivered.
That is good news if you keep records and very bad news if you do not. It also means the dispute was mostly won or lost before it started.
How the process actually runs
The shape is consistent across platforms even where the names differ.
A buyer raises a claim, usually against a specific order, sometimes with a category attached: not delivered, not as described, unauthorised payment. Funds are held if they have not already been paid out. Both sides are asked for a statement and any evidence, generally with a deadline in days. A moderator reads the platform's own message history, the order metadata and whatever you attached, then decides.
Three things follow from that description. The moderator is not the person who read your messages last week; they are seeing the whole thing cold, in about four minutes. Nothing outside the platform's own records is independently verifiable to them. And a missed deadline is a loss by default, which is the single most avoidable way to lose one.
A card chargeback is a different animal with the same evidence requirements. There the decision is made by the buyer's bank, the platform is merely relaying what you send, and the timescale is months. Stripe's dispute documentation gives the shape: usually 7 to 21 days to respond depending on the card network, an automatic loss if you miss the deadline, a single submission you cannot amend, and up to three months for the bank to decide. The wider mechanics are in what drives chargeback rates.
The evidence hierarchy
Not all evidence is equal, and the ranking is about verifiability rather than persuasiveness.
| Evidence | Weight | Why |
|---|---|---|
| Platform message confirming scope and date | Highest | Timestamped, unedited, already in front of them |
| Platform delivery record | Highest | Proves what was sent and when |
| The deliverable itself | High | Lets them check it against the confirmation |
| Buyer's own messages accepting the work | High | Their words, on the platform |
| Screenshots of off-platform chat | Low | Unverifiable, trivially edited |
| Your account of what was agreed verbally | None | Not evidence |
The pattern is obvious once written out. Everything in the top half exists automatically if you confirm scope in the platform's messaging before starting, which takes twenty seconds and is the whole subject of the two-sentence confirmation.
One addition that punches above its weight: a delivery message that restates what was delivered against what was agreed. "Attached: the 500-word assessment across the four axes, as confirmed on the 14th." It gives the moderator the comparison in a single line rather than making them build it.
Tone is a tactic, not a virtue
Write your statement as if the reader is bored, neutral and has no context, because they are.
Facts in order, dates attached, no adjectives about the buyer, no history of how unreasonable they have been. Three short paragraphs: what was agreed, what was delivered, what the buyer is now claiming and why the record contradicts it.
Do not argue with the buyer inside the dispute thread. Everything you write there is read by the moderator, and an exchange of accusations makes both parties look equally likely to be the problem, which favours whoever is claiming rather than defending.
Do not send new evidence in five separate messages. One statement, attachments named clearly, done.
The generally decent behaviour that makes disputes rare in the first place is unglamorous and mostly consists of answering quickly and confirming things in writing; the buyer-side account of how a commission is normally conducted is worth reading precisely because it shows how low the bar is.
When to concede early
Conceding is a pricing decision and should be made with a calculator rather than a sense of injustice.
Concede when the fee is smaller than the hours the dispute will take, when your records are thin enough that you would probably lose anyway, when the buyer has a plausible-sounding claim you cannot cleanly disprove, or when the dispute is heading somewhere that will produce a public review whatever the outcome.
Contest when the amount is material, when your confirmation and delivery record are both clean, or when conceding would establish something with a repeat buyer that you do not want established.
There is a rate effect too. Platforms watch dispute frequency rather than dispute outcomes, so a habit of contesting small claims to win them is still a habit of accumulating disputes. The cheaper instrument for a small contested job is usually a refund, and the comparison is set out in when a fast refund is the cheapest outcome.
The disputes that are really about expectations
A recognisable subset of claims are not about delivery at all. The buyer received exactly what was agreed and disagrees with the conclusion, or expected a number to mean something more definite than it can.
These are winnable, because the record is on your side, and they are still worth preventing. Prevention is one line at delivery stating what the assessment is and is not claiming - which is the same sentence that stops the equivalent revision request.
Where the disagreement is about a measurement rather than an opinion, quoting a stated convention ends it faster than reasoning does, and the conventions live on the measurement side rather than being yours to invent per job. Where it is about a score not matching a tool the buyer ran first, the honest account of what automated accuracy actually means is the reference, and one sentence pointing at it is more effective than defending your own judgement at length.
Platforms differ in deadlines, in whether payouts are reversible mid-dispute, and in who can open one. Rate Cock's judges page is where its own order and resolution flow is documented, and the part worth knowing in advance is which of your evidence it already holds for you.