Evidence That Actually Wins Disputes

Listing snapshot before payment

A claim is settled on what can be read again months later, not on what a buyer remembers. The page that carried the terms can be rewritten, swapped for another item or taken down within days of payment, and a file assembled from memory afterwards argues against nothing. What holds is a capture taken while the listing is still the listing: title as written, option selected, price line, gallery the seller was showing, and the date. Half a minute of housekeeping at that moment decides how much of the case survives.

Desk practice for that capture stays deliberately plain. One full-window image, uncropped, with the title, the option selector and the price line all legible in the same frame; a second image of the gallery opened at full size; a third of the specification block. Written beside them in the same note, the date, the storefront name as displayed, and the item description copied word for word. The copied text usually outlives the pictures, because text can be searched and a photograph of text cannot.

Captures fail in three familiar ways. They are taken after dispatch, when the page has already moved, so the file shows a listing nobody bought from. They are cropped to the product, which removes the option list and the price line that carry the terms. Or they live only inside a chat thread, where anyone reading the claim has to scroll past conversation to reach them. One dated file, kept outside the thread, is the difference between a document and a remark.

A snapshot has a second use that only appears later. Comparing the capture taken at payment with the page as it stands during a claim tells a reader whether the listing changed, and roughly when. Two dates and two versions of one page are more persuasive than any sentence about what the page used to say, and they cost nothing to keep.

Chat log export

Written answers are the one place where a specification turns into a promise. Ask inside the platform's message channel rather than a side conversation, and ask in a form that can be answered with a fact. The working questions name the thing and the reply wanted: which batch marker sits on the box, which material the upper uses, which day the parcel leaves, whether the size chart on the page is the one to trust.

Keep the reply that commits and the reply that dodges, because both are evidence of process. A seller who writes that the item ships within four working days has created a line a later reader can date. A seller who answers a precise question about batch codes with a warm sentence about quality has created a different line, and it belongs in the file with its date attached rather than in a summary written afterwards.

Threads move, get truncated, or sit behind an account that a reader of the claim does not hold. Export the exchange at the end of each round of messages instead of at the end of the order, so a thread that closes early is already saved. The desk keeps these exports as plain text where possible, with the message dates left in place; a rewritten transcript loses the one quality a claim relies on, which is that the words were somebody else's.

Unboxing sequence shots

Sequence is what a single close-up cannot supply. Six frames taken in order — outer carton, courier label, seal before cutting, contents laid out, the part in question, the defect at reading distance — link a specific flaw to a specific parcel and a specific order. A photograph of a scratch on its own proves that a scratch exists. It does not connect that scratch to the box which arrived that afternoon.

Shoot the whole set in one session, under one light, without moving the parcel out of frame between shots. Continuous numbering from the camera roll keeps the order and the time stamps intact, and originals matter more than a tidy edit: a cropped or brightened copy invites an argument about what was changed, while the untouched file ends it. Where a fault is mechanical rather than visible, one unbroken clip showing the fault happen beats five stills of the same part.

Resist tidying up before the camera comes out. Packaging is evidence, and flattened cartons, discarded filler and a label torn off at the corner are the three things most often missing from a file. Where the parcel was left at a door or collected by somebody else, a photograph of where it was found, taken before it was moved, is the only record of that half of the story.

Measurement evidence

Numbers end arguments that adjectives prolong. A flaw stated as eleven millimetres long, on a part whose listing promised a smooth edge, is a claim a reader can check. The same flaw described as unacceptable is a preference. Measure the part itself rather than its photograph where possible, and where the part cannot be measured directly, put an instrument in the frame so the picture carries its own scale.

Three measurements earn their place in most files: length of a seam, gap or scratch against a steel rule laid flat beside it; weight of the item on a kitchen scale, set against the figure in the specification block; and, for anything worn, the flat measurement of the garment or strap against the size chart the seller published. Each one pairs a reading with a promise made on the page.

State the error band with the reading. A photograph taken square-on with a rule in frame holds roughly a two millimetre tolerance at the distances most buyers use, and a kitchen scale drifts by a few grams. Publishing that band weakens nothing; it stops a dispute from being built on a difference the instrument cannot support. The desk records both numbers — what the page said and what the instrument showed — and leaves the gap to speak for itself.

Instrument choice matters less than consistency. A cheap steel rule and a domestic scale are enough, provided the same ones are used for both the measurement and the comparison shot. Where a second opinion is wanted, a second measurement taken on another day, with the first reading left unrevised, is worth more than a more precise instrument introduced halfway through.

Timeline of messages

A file is read as a chronology long before it is read as an argument. Lay the events out as dated lines: order placed, dispatch notice, first scan, delivery scan, first message to the seller, each reply, each silence, each deadline. Written that way, the record shows where the process stopped rather than what anybody felt about it, and a reader who was not present can follow it in a minute.

Silence deserves a line of its own, with the date a reply was expected and the date it had still not arrived. Recording an unanswered message is not an accusation; it is a fact about the process, and it is usually the fact that decides which route is left. Use the platform's own published response windows as the expectation, and write down where those windows are stated, because they change.

Keep the whole chronology to one screen. Each line holds four things and nothing else: the date, who acted, what they did, and what changed as a result. Lines carrying opinion get moved to a separate note, where they cannot contaminate the record. The desk rebuilds timelines from raw messages rather than from an earlier summary, because a summary written in the middle of a dispute inherits the mood of the week it was written.

Deadlines belong on the timeline as calculated lines rather than remembered ones. Write the date a claim window closes, the date a reply fell due, and the date a return scan had to appear, each with the rule it came from. A reader who can see three closing dates at once understands why the next message has to go this week, which is exactly what a stalled case needs.

Escalation note

When the first route stalls, the file has to be lifted out of the thread and onto a page. The escalation note the desk uses runs to roughly four hundred words: what was ordered, what arrived, what was asked, what was answered, what remains unresolved, and what outcome is being sought. It carries dates and document references rather than adjectives, and it is written so that a reader with no stake in the case can act on it.

Name the remedy in one of three forms and put a figure on it: a refund of a stated amount, a partial refund of a stated percentage, or a replacement of a stated specification. A note asking for fairness invites another round of messages; a note asking for one of three defined outcomes can be accepted or refused. Attach the smallest set of documents that carries the case — snapshot, thread export, arrival sequence, measurement — and list them by name.

Two boundaries hold. The desk describes process and evidence; it does not give legal advice, and a case raising a question of law belongs with a qualified adviser in the reader's own jurisdiction rather than in a message thread. The second boundary is restraint: a note that overstates the file costs more than it gains, because the first thing a reader checks is whether the strongest claim survives contact with the documents attached. The protocol page sets out how each item is weighted before a case goes up a level.