A customer opens a brand-new order and finds somebody else's used item inside. Trace it back through your returns process and tell me where it went wrong.

Receipt and grading are two events, and most systems only have one

Work backwards from the used item in the box and the path is short. That unit was picked from a sellable location, which means something wrote it back into sellable stock. In almost every case the write happened at the moment the parcel was received, triggered by a scan of the return label or of the item barcode, because someone modelled "the return came back" and "the stock is available again" as the same fact. They are not, and conflating them is the whole defect.

Receipt is an acknowledgement that a physical object has arrived at your building. It tells you the customer has done their part, which is why it is a reasonable trigger for the refund. It says nothing about the object's condition, its completeness, or whether it is even the item claimed. Grading is a separate decision, made by a person or a station with the item in hand, that establishes what you now own. Disposition is a third thing again: a routing decision about where the graded unit should go. Only the last of those may increment sellable stock, and only for the grades that justify it.

Once the three are separated the fix is structural rather than procedural. Received returns land in a holding location that is excluded from availability by construction, not by convention. Nothing in that location can be allocated, so no amount of pressure on the floor can put an ungraded unit into a customer's order. Moving stock out of it requires a recorded disposition, which gives you the audit trail when this happens again.

The other ways the wrong unit reaches a customer

Restocking on receipt is the common cause, and a good answer names the near misses too, because each is a different control.

Grading exists but happens after restocking, as a downstream tidy-up. This looks correct on a process diagram and is the same bug with more steps: the window between the two is exactly when the unit is available to be picked.

Grading exists and is sampled rather than universal. Sampling is defensible for a low-value, low-variance line and indefensible for anything a customer can tell has been used. If you sample, the categories you sample must be a deliberate decision with the risk written down, not the consequence of a backlog.

The item is graded by its label rather than its contents. A returns station under time pressure scans the barcode on the box, sees the expected SKU, and books it in without opening it. This is how an empty box, a substituted item or a brick becomes sellable stock, and it is also the point where returns fraud is either caught or funded. Weight checks, mandatory opening for lines above a value threshold, and photographic evidence at the station are the usual controls.

The unit is graded correctly and put away in the wrong place. A put-away that offers the original sellable location as the default will get used, so the holding and refurbishment locations need to be the path of least resistance for anything not graded as new.

And finally the unit was never sellable-new to begin with: an open-box or refurbished unit sitting in the same location as new stock, distinguished only by a note somebody was supposed to read. Condition has to be part of the stock identity, so that a unit graded as open-box is a different sellable thing with its own availability and its own price, and cannot be allocated to an order for a new one.

Grading has to produce a decision, not an opinion

For grading to gate anything it needs a small, fixed set of outcomes that map to actions. The vocabulary varies by retailer, but the shape is consistent: sellable as new, sellable at a lower condition, repairable or refurbishable, returnable to the vendor under a supplier agreement, or scrap. What matters is that each grade has exactly one default destination, that the grader is recorded, and that the evidence is retained for long enough to settle a dispute with the customer or the supplier.

Two design points follow. First, the grade must be attached to the unit rather than to the return, because a multi-line return can contain units in different states and the aggregate is meaningless. Second, the grade should be capturable at the point of receipt where that is practical, because every additional touch on a returned unit costs money and time, and a process that requires the item to be handled three times will be shortcut by the people doing it. Designing the safe path to also be the fast path is what makes the control hold when volumes spike after a peak trading period.

Some categories should be excluded from sellable-new outright, regardless of what a grader concludes. Opened consumables, anything with a hygiene dimension, safety-critical goods, and items whose registration or warranty is tied to a serial that has already been issued are the usual examples. Where a legal or regulatory restriction applies it applies whatever the grade says, so the category rules belong with compliance and the system's job is to make them unbypassable rather than advisory.

Promising a unit you have not looked at yet

There is a tempting optimisation that produces the same customer outcome by a different route. Returns in transit are known, so you can add them to available-to-promise and sell the unit before it lands. Sometimes that is a genuine commercial win on a scarce line. But it creates pressure to make the incoming unit sellable the instant it arrives, because there is already an order waiting for it, and that pressure is what defeats grading. If you do it at all, the promise should be to a date that includes grading time, and the order should be allocatable only from graded stock.

The same caution applies to counting the holding location in any figure a system reads. It should be visible to operations, because a growing quarantine is an operational signal worth acting on, and invisible to every channel that sells.

Refunding early without restocking early

The reason teams collapse receipt and grading is usually a good one: they want to refund the customer quickly, and receipt is the earliest defensible trigger. That instinct is right and does not require restocking. Refund on receipt, grade afterwards, and treat any recovery — a deduction for a damaged item, a rejected return, a chargeable fault — as a separate downstream action against the return record. It will be an unwelcome conversation with a minority of customers, and it is a smaller cost than shipping used goods as new.

What that split does require is an honest measure of how long grading lags receipt, and an alert when the holding location ages past a threshold. Left unmeasured, the backlog grows until somebody decides to clear it by booking everything back in, which is how this failure recurs six months after it was fixed.

Detecting it before the customer tells you

Two signals are worth having in place. A rise in returns for a given line where the stated reason is condition-related, or a rise in returns of items that were themselves recently returned, points straight at a grading failure and can be computed from data you already hold if the unit's history survives the restock. And a second-return rate per SKU, tracked over time, is a leading indicator that stock of that line has become contaminated with units that should never have gone back out.

Receipt earns the customer their refund; only grading earns the unit its way back into sellable stock. If a returned item can be allocated before somebody has looked at it, the process has no grading step — whatever the diagram says.