Timestamps, in the platform, in order

A dispute is not decided on who was right about the work. It is decided by whoever can show, from the platform's own records, what was agreed and what was delivered, with dates attached.

Everything else you might want to submit is context, and context loses to timestamps.

What a reviewer is actually doing

The person or team reading your submission is working through a queue with minutes rather than hours per case, and they are not qualified to judge the assessment itself.

So they resolve the case on things they can verify without leaving their own system. Was there a stated scope, and where. Was something delivered, and when relative to the agreed date. Did the conversation stay in the platform, or does half of it exist only as an assertion.

This is why the same evidence wins repeatedly and the same evidence fails repeatedly, regardless of the merits. A reviewer who cannot verify a claim treats it as unproven, not as false, and unproven is close enough to losing that the distinction does not help you.

The hierarchy, roughly

The gap between the first two rows and the last two is the whole subject. It is also why the confirmation habit matters more than any dispute technique: the two-sentence confirmation exists to create row one before you need it, and no amount of skill at writing submissions makes up for its absence.

Deliver inside the platform, always

The single most common self-inflicted loss is delivering the work by email, a file-sharing link, or a message on another service because it was more convenient.

From the reviewer's position, that job was never delivered. There is no record, the buyer has no incentive to confirm one, and your explanation of where the file went is the weakest category of evidence there is.

If the deliverable genuinely has to leave the platform - a large video file, a format the messaging will not carry - the fix is to post the delivery notice inside the platform at the moment you send it, naming what was sent and where. That creates a timestamped record of performance even though the file travelled elsewhere.

The same logic covers revisions, extensions and every agreed change. A verbal agreement to a two-day extension that exists nowhere is a missed deadline in the record.

Assembling the submission

Write it as a chronology, not as a case. Payment processors give the same advice for card disputes: Stripe's dispute evidence guidance says to present evidence chronologically, keep it relevant to the dispute reason, and include only the relevant excerpts of long message threads.

Date of booking and the agreed scope, quoted. Date of the confirmation message, quoted. Date of delivery, with the platform's own record referenced. Date and content of any change the buyer requested, quoted. The specific outcome you are asking for.

Three rules make the difference between a submission that reads well and one that reads badly.

Quote rather than characterise. "On the 4th I wrote: the fee covers one submission" is evidence; "I made clear from the start" is not.

Include the things that hurt you. A reviewer who finds an omitted message in the thread stops trusting the rest of your account, and they will read the thread.

Say what you want, in one sentence, at the end. Submissions that argue at length without naming a requested outcome get resolved by whatever is easiest for the queue.

Keep the tone flat. The buyer's submission being emotional is an advantage you throw away by matching it.

Timing, and what runs out

Dispute windows are shorter than people assume, and card chargeback windows run on a separate and longer clock that the platform does not control. Stripe's account of how disputes work puts numbers on that clock: cardholders typically have 120 days from payment to dispute, the seller usually gets 7-21 days to respond, and the full cycle can take 2-3 months. That mismatch is why a job can be settled internally and reappear as a chargeback months later - the piece on holds, chargebacks and walking away covers what happens to money in that gap, and what actually drives chargeback rates covers preventing the second event entirely.

Because of the second clock, retention is part of the evidence strategy. Export the thread and the order record when a job goes sideways, not when a notification arrives - by then the account may be restricted and the records unreachable. Which windows apply and what the platform itself retains is set out in its own terms; Rate Cock's rewards and payout page is where the money-side timings for that platform are described, and every platform states its own somewhere similarly dull.

Two things from outside are worth borrowing. The reason a documented method beats a confident conclusion is general, not specific to disputes, and the measurement side's insistence on stating conditions before producing a figure is the same instinct applied earlier in the process. When a buyer's complaint is that a score was wrong rather than that nothing arrived, it helps to know what a number can and cannot be held to: the account of what automated scoring reliably does sets the realistic bounds, and how scores are presented in the tools buyers have already used explains where their expectation came from in the first place.

Reviewers do not resolve that argument on the substance either. They resolve it on whether the scope said what would be assessed, which returns the whole subject to a message you either sent before starting or did not.