Fulfillment Exceptions

Every seller has been told a fulfillment partner can "handle exceptions". In a domestic warehouse that means a wrong item and a reprint label. Move the same order across a border and the exception acquires a customs status, a currency, a carrier liability regime and a clock that runs while your buyer's marketplace metric quietly degrades. The gap between those two worlds is where most cross-border partnerships actually fail — not on price, and not on speed.

Why a border changes the nature of the problem

Inside one country, an error costs one shipment. Cross-border, an error costs the shipment, the recovery shipment, the import paperwork for the replacement, and often a week of your listing's delivery promise. Three structural features cause this. First, remediation has a second leg: a replacement unit is itself an export declaration, not a walk to a shelf. Second, the evidence chain spans two legal systems — what a US carrier accepts as proof of loss and what an EU buyer considers a delivered parcel are not the same document. Third, the value at risk is asymmetric: a USD 12 item that arrives broken can still cost a marketplace account's on-time-delivery threshold, which is measured across thousands of orders, not per order.

Our own answer to this asymmetry is deliberately boring: we hold a mis-pick and mis-ship rate below 0.01% across three networked warehouses in Shenzhen, Suzhou and Dongguan, because the cheapest exception is the one that never starts. Below that, we plan for the eight classes that do.

The eight failure classes

1. Short ship and over ship. A carton arrives at the receiving hub with fewer units than the packing list, or a mixed SKU the supplier swapped. Root cause almost always sits with the factory, so detection has to happen at inbound receipt — counted, weighed and photographed against the purchase order before anything is shelved. An undetected shortage becomes a mis-pick three weeks later and a claim against the wrong party.

2. Inventory divergence. System stock and shelf stock disagree. The causes are unscanned returns, damaged units pulled aside without a movement record, and multi-channel oversell against the same physical bin. The control is a cycle count cadence plus a hard rule that nothing leaves a location without a transaction.

3. Mis-pick and mis-ship. Right stock, wrong unit out the door. Volume SKUs in similar packaging are the usual victims. Batch-pick with a scan at each bin, and a second scan at pack, is the only defence that scales; visual checking does not survive a peak week.

4. Damage in transit. The expensive one, because responsibility is contested. A carton can pass a warehouse inspection and still arrive crushed. What decides the outcome is packaging specification recorded before hand-off — inner protection, carton test class, palletising method and photographs at seal time — not the condition of the parcel when the buyer photographs it.

5. Customs hold. Now the fastest-growing class, and the one sellers underestimate most. Since the United States suspended the de minimis exemption for China-origin parcels, an incomplete commercial invoice or a missing HTSUS classification is enough to stop a shipment that would previously have cleared by default; from 24 July 2026 even low-value postal entries require a bond and a complete data set. Holds are rarely a mystery — they are usually a data defect we could have caught at documentation, which is why our target is two hours to acknowledge and twenty-four to produce a written workaround on any exception, including this one.

6. Lane suspension and capacity denial. A carrier suspends battery-containing services, a route loses a hub, or space is denied in the August-to-January peak. The failure is not the suspension; it is having only one lane. Routing capacity across more than 50 shipping channels exists precisely so a suspension becomes a re-route rather than a cancelled order.

7. Address, contact and delivery-window defects. Undeliverable addresses, missing phone numbers (carriers in parts of South America, the Gulf and the EU require one), and buyer-absent first attempts. Mostly preventable at order import with validation rules, and the reason an order file should never be treated as trusted input.

8. Returns and reverse logistics. The class with no clean answer cross-border. Shipping a low-value unit back to China costs more than the unit, so the economics push every seller to fix disposition before the goods move rather than after a buyer opens a case. Where a return is collected, who inspects it, and whether it is resold, repaired or written off are commercial choices that change the answer order by order — sellers who have not set them in advance improvise per case, and every improvisation is a margin leak.

What an exception file must contain

Recoverable exceptions are an evidence problem disguised as a service problem. A complete file carries the order and tracking identifiers, the product's declared description and classification, the packing photograph taken at hand-off, the weight and dimension record at both origin and delivery, the carrier's own delivery or exception scan, the buyer's image of the goods, and a written statement of the remedy sought. Files assembled that way settle in days. Files assembled from memory settle in weeks, or not at all.

The escalation ladder

One named account manager owns the case from first signal to closure, in the buyer's timezone, with a two-hour acknowledgement and a same-day written position. The ladder is short by design: identify, contain, decide remedy, then trace the cause back to whichever of the eight classes it belongs to and change the process that let it through. A resolved exception that produces no process change is a subscription you pay for repeatedly.

Questions to put to any China-based partner before you hand over stock

• Which of the eight classes do you measure, and at what frequency do you report the rate?
• What is in your exception file, and who assembles it — your team or mine?
• How many shipping channels can a single order be re-routed across, and how quickly?
• When a lane is suspended mid-peak, who absorbs the cost difference?
• Where does a US or EU return physically go, and what decides whether it is restocked?

FAQ

Does a China fulfillment provider's mis-pick rate matter more than its shipping price? For most catalogues, yes, because one mis-pick generates a reshipment, a refund risk and a marketplace metric that a per-order price saving rarely covers.

Who is the importer of record when goods are held by customs? It depends on the commercial terms and on whose bond the entry was filed, and it should be fixed in writing before the first shipment — not negotiated during a hold.

How long does a cross-border claim usually take? It is driven by evidence completeness rather than by value; a full file with origin photographs and carrier scans typically moves far faster than any disputed, thinly documented case.

Can exceptions be automated away? Detection can be — scan-verified picking, dimensional capture at pack, address validation at import. Judgement cannot, which is why the human ownership model matters.

Any transit, lead-time or resolution figures in this article are operational targets and reference ranges, not guarantees; they vary by carrier, destination, season, port and customs handling. Duty, entry and returns rules summarise requirements in force at the time of writing and are general information, not advice on a specific shipment.