The window is the deadline
Two people comparing the cost of returning an item usually disagree because they are counting different things. Three return paths set side by side, each with its own definition of cost, its own clock and its own blind spot.
Three return paths, three different cost definitions
Most return arguments are not arguments about money. They are arguments between two people using different definitions of cost, one counting the postage already paid and the other counting only what is still in front of them. The desk return bench exists to make that disagreement visible, and it does it by pricing three paths separately rather than producing one figure called the cost of returning.
The send-back path counts the return postage, unless the seller covers it, plus the goods value while the item is away. The exchange path counts postage in both directions, because an exchange is two journeys over the same item. The keep path counts the share of the purchase price that cannot be recovered locally, which is usually the smallest of the three figures on a low-value order and the largest on a valuable one.
Postage already paid to the international line is excluded from all three, deliberately. It is gone whichever path is chosen, so including it would make every path look equally bad and hide the only difference that can still be acted on. Anyone comparing return costs should confirm which definition the other side is using before comparing numbers.
The send-back path and what it counts as cost
Sending the item back is the only path that returns the goods price itself, and on a genuine fault it is usually the strongest claim available. Its evidence requirement is also the simplest: a photograph of a fault that can be seen, attached to the item and the order it belongs to. Where the item has not yet been packed, the seller relationship is still open and the fix is cheaper for everyone.
Its strengths come with three dependencies. The path runs on a handling window that the buyer does not control and cannot extend; it depends on the seller accepting the fault, because postage normally falls on the buyer when the fault is not accepted; and it takes the item out of reach for a period while the money is still not back, which the bench models as roughly three and a half weeks of planning time rather than days.
This path wins when the item is genuinely faulty, the fault is visible rather than felt, the order is worth more than the postage, and there are enough days left both to raise the return and to complete it. It loses when any one of those four is missing, and it loses hardest when the last one is missing, because a return raised on the final day of a window is a return that may not finish inside it.
The exchange path and its second leg of postage
An exchange answers a narrower question than a refund does: the item is right and the size is wrong. It keeps the purchase rather than reversing it, which is often what the buyer actually wanted, and it is the only one of the three paths that ends with the reader holding the thing they originally set out to buy. On a model that disappears from listings quickly, that is a real advantage rather than a consolation.
The cost definition is what makes it the most expensive path on paper. Both legs of postage fall on the same item, and the bench models the round trip as the longest of the three, because it is a return plus a fresh dispatch plus an inspection in between. It also depends on the size still being available, which is a condition the buyer cannot check in advance and cannot influence once the return is moving.
Where a measurement is available the claim is much stronger, because the argument stops being about the label. The desk footwear tally records insole length and widest point across 46 rows from a run of three pairs per model, and the garment run adds 38 rows measured as chest width and body length rather than by labelled size. An exchange supported by a measurement is a factual request; one supported by a description of how the item fits is a preference.
The keep path and why it has no request to file
Keeping the item is the path with no counterparty. Nothing is filed, nothing is approved, and no window applies, which makes it the only option that cannot be closed by the calendar. In the bench model it is also the quickest to complete, because there is no journey and no inspection, only a decision and whatever local effort the reader chooses to spend.
Its weakness is the input it depends on. The cost of keeping is defined as the share of the purchase price that cannot be recovered locally, and that share is an estimate the reader supplies. Overstate the local recovery and keeping looks better than it is; understate it and a perfectly sensible decision looks like a loss. Unlike the other two paths, this one cannot be checked by anyone else, because nobody else knows the local market.
The path wins on low-value items with cosmetic faults, and on any item where the return postage quoted is a large fraction of what the goods cost. It also wins when the window has already closed and the reader is deciding what to do rather than whether to file. That is not a defeat: it is the same arithmetic with two of the three rows removed.
Choosing a path against the window you have left
The window decides before the money does. In the desk model two of the three paths stop existing when the handling window closes, and the closure is silent, which is why the bench prints the days remaining above the comparison instead of beneath it. Window length is a policy figure rather than a carrier promise, so the desk records it as the policy states it and keeps the policy date beside the number.
Once the days remaining are known, the choice usually collapses to two questions. Is the item faulty, or is it correct but wrongly sized? A faulty item points at the send-back path and a correctly sized-up item points at an exchange, while an item that is both faulty and wrongly sized is the case where an exchange quietly becomes the most expensive of the three. The second question is whether the postage is a large fraction of the price, which is what moves a low-value order onto the keep path.
The table below sets the three paths against six dimensions rather than against one number, because the dimensions are what people actually disagree about. Read the cost definition row first, then the closed-by row, then the recover row, and the path that fits usually selects itself.
| # | Dimension | Send back for a refund | Exchange | Keep and move on locally |
|---|---|---|---|---|
| 01 | Cost definition | Return postage unless the seller covers it | Postage in both directions on the same item | The share of the price that cannot be recovered locally |
| 02 | What it recovers | The goods price | A correctly sized version of the item | Whatever the local market pays for it |
| 03 | Planning days | About 24, as a two-journey round trip | About 32, because it is a return plus a fresh dispatch | About 7, and mostly spent deciding |
| 04 | What it depends on | The seller accepting the fault | The size still being available | An honest local recovery estimate |
| 05 | Closed by | Window expiry, or a refusal | Window expiry, or a sold-out size | Nothing on the platform side |
| 06 | Strongest when | A visible fault on an order worth more than the postage | A measured size error on an otherwise correct item | A cosmetic fault on a low-value item |
| ∑ | Planning days are desk model figures for two journeys and an inspection, not carrier commitments. | Postage already paid to the international line is treated as sunk on all three paths. |
Window length is quoted as the policy states it, with the policy date recorded beside the figure.
What this comparison leaves out on purpose
Three things are missing by design. The first is the reader time, which is not priced on any path and is genuinely large on the exchange row. The second is the possibility that a seller accepts a fault late, after the reader has already chosen a different path, which is a real outcome that no comparison can schedule. The third is the case where the item cannot travel back at all because the line refuses its contents, which removes the send-back and exchange rows entirely and leaves only the local option.
Duty and tax are also outside the comparison. A refunded purchase can still leave a border charge standing, and a replacement item can attract a fresh assessment, because the two events are assessed separately and by different parties. Where a border authority is involved, the notice that governs is the one that authority issues, and the desk does not model what that notice will say.
What the comparison is good for is the decision in front of the reader today: which of three paths is cheapest given what is still recoverable, how many days remain, and what evidence exists. It is a planning tool rather than an assurance, and nothing in it commits any party to an outcome.
What this note comes down to
- 01Three return paths use three different definitions of cost, so comparing one number against another rarely settles anything.
- 02Postage already paid is sunk on every path; only the money still in front of you can be acted on.
- 03The send-back path recovers the goods price and depends on the seller accepting the fault.
- 04The exchange path pays postage twice and depends on the size still being available, which is why it is the most expensive row on paper.
- 05The keep path has no counterparty and no window, but it depends on a local recovery estimate that nobody else can check.
- 06Read the days remaining before the money: two of the three paths stop existing when the window closes.
The data point behind this note
Window lengths are recorded as the platform or agent states them, with the policy date written beside each figure, and no window is inferred from a forum post; that rule was set during batch KB-2026-05 and is what keeps the day counts in the comparison dated rather than remembered.
KB-2026-05· 2026-09-30 Batch ledger · Challenge a figure
Notes in the same cluster
| # | Note | What it argues | Words |
|---|---|---|---|
| 01 | A refund that was actually worth taking | A refund is worth taking when the money coming back beats the money the return itself removes. Three completed files show where that line actually fell, and how close two of them came to falling the other way. | 1625 |
| 02 | What a rejection letter gets you, and what it does not | A rejection notice answers one question well and leaves four others untouched. This is a composite exchange, followed from the first question through to the point where the answer stopped changing. | 1591 |
| 03 | Five ways a parcel gets held, and how each one ends | A parcel that stops at a border is usually stopped by paperwork rather than by suspicion. Five holding events, scored on how likely each is and how much of the order it can consume, with the first move for every one. | 1856 |
Related entries and gauges
- What can you still do after the handling window expires?
- What can you do when a seller refuses a return?
- Return path bench
What each path costs from where you are standing now.