Pre-Shipment Inspection in China

A pre-shipment inspection report can say PASS while the buyer should still hold the shipment. PASS means the sampled lot met the decision rules and checks written into that inspection protocol. It does not mean every unit is defect-free, every requirement was tested, the lot's exact defect rate is known, or all regulatory and contract conditions are closed.

A useful PSI starts with a frozen lot file, product references, defect definitions, sampling plan, special-test methods, and release authority. Without those inputs, an inspector can produce a polished report that answers the wrong question. Read the sampling trail and limitations before treating the summary as a commercial decision.

A PSI is a lot decision under a written protocol

Pre-shipment inspection is a point-in-time check of goods presented before dispatch. It is commonly performed after production is complete and packing is sufficiently advanced for the inspector to examine finished product and export packaging. There is no universal “80% packed” rule that makes every order representative. The buyer and inspection provider should define readiness for the actual product, lot layout, packing process, and tests.

The limited question is: does evidence from the presented lot satisfy the agreed acceptance plan and product checks well enough for the authorized buyer to release, hold, reject, rework, or waive a stated condition? The inspector records observations. The buyer owns the disposition under its contract and policies.

A normal PSI may cover quantity presented, SKU/variant identity, workmanship, dimensions, basic accessible function, assortment, accessories, retail pack, labels, barcodes, master cartons, and shipping marks. It cannot recreate the development program, inspect every hidden component, prove long-term reliability, or establish every destination-market requirement. Use it where origin correction is still possible, not as a general product guarantee.

Freeze the order, lot, and reference before booking

Build the inspection pack from buyer-controlled records. Do not ask the factory to decide the final specification on inspection morning.

Order identity

Record the purchase order, contract seller, factory legal name and address, SKU/model/variant, color/size assortment, ordered quantity, shipment split, destination, and requested dispatch date. Identify subcontracted locations or off-site warehouses. If several products are combined for logistics, state whether they form one inspection lot or separate lots and why.

Presented lot

Require the factory to state production-complete quantity, packed quantity, unfinished/rework/quarantine quantity, carton and pallet ranges, storage locations, production dates or batches, and any units not presented. Define the condition for postponing the visit. A report on a convenient subset should not be labelled as the final lot without a visible limitation.

Controlled product reference

Send the dated specification, drawings and tolerances, approved sample ID and photographs, approved materials/components visible to inspection, artwork, label and barcode files, packaging diagram, assortment, and change log. The product sample workflow explains how to keep a physical reference tied to versions rather than “same as sample” memory.

Write exact expected results. “Check quality” and “function OK” are not test methods. For each measurable item, provide units, equipment, conditions, tolerance, sample quantity, and response to a failure. State whether a test is destructive and who authorizes it.

Build two decision plans, not one

The workmanship sample and the special-test sample often answer different questions. Keep them separate in the protocol and report.

Plan A: acceptance sampling for attributes. Identify the standard and edition, scheme, switching status, inspection level, lot size, sample-size code and sample size, AQL used for each defect class, acceptance/rejection numbers, and whether the report counts nonconforming units or individual nonconformities. Record how cartons and units will be randomly selected across the lot.

Plan B: product-specific checks. Define separate sample quantities and decision rules for dimensions, weight, function, assembly, barcode scans, accessories, label content, packing, drop or transit checks, powered operation, destructive checks, or other product needs. A ten-unit function test does not become a full workmanship sample merely because it appears in the same report.

A workmanship PASS cannot cancel a failed special test. Conversely, one failed measurement should be handled under its written rule rather than quietly added to a cosmetic defect count. Identify independent stop conditions in advance: a wrong market-specific voltage, missing required warning, exposed conductor, or another agreed safety/regulatory concern may hold the lot regardless of the ordinary major/minor acceptance count.

Use ISO 2859-1:2026 deliberately

ISO published ISO 2859-1:2026 in January 2026. It is the third edition, replaces the 1999 edition and its amendment/corrigendum, and defines AQL-indexed acceptance sampling by attributes for lot-by-lot inspection. The edition includes single, double, and multiple schemes, switching rules, updated application guidance, and skip-lot procedures.

The standard does not make “AQL 2.5, level II” a universal China-inspection setting. Product risk, defect consequences, customer expectations, manufacturing history, contract, and applicable requirements have to inform the plan. Use an authorized copy and qualified quality input to obtain code letters, sample sizes, and acceptance/rejection numbers. This article does not reproduce ISO's copyrighted tables or generate an ISO plan.

AQL is not the exact percentage of defects the buyer agrees to receive. Acceptance sampling makes a lot decision under probabilities and rules. NIST's indexed research paper Acceptance sampling by variables develops the producer-risk and consumer-risk sides of that decision. A sample can pass while uninspected units contain defects, and an acceptable process can sometimes produce a rejected sample.

Read “sample size” and “Ac/Re” as one complete plan. Do not choose a sample size from one standard or edition and acceptance numbers from another. Do not change the inspection level after seeing the result. Record double or multiple sampling stages exactly when used; the first sample in such a scheme does not always yield a final decision.

Do not silently mix ISO 2026 and China's current GB/T edition

China's official standards platform records GB/T 2828.1-2012 as current after a May 2025 review. It is a recommended national standard that equivalently adopted ISO 2859-1:1999. It is not the same edition as ISO 2859-1:2026.

If the purchase contract, inspection brief, factory procedure, or provider template cites GB/T 2828.1-2012, do not relabel the result as ISO 2859-1:2026. If the buyer moves to the new ISO edition, agree on the change, obtain the correct tables and guidance, train reviewers, and update the protocol. Preserve exactly what governed each historical report so results can be interpreted rather than normalized into a vague “international AQL standard.”

A factory's internal outgoing inspection can also use a different standard or plan from the buyer's PSI. Keep both records, but do not merge their results unless the populations, methods, definitions, and decision rules truly align.

Classify defects before the visit

Critical, major, and minor are practical categories only when the buyer gives product-specific definitions and examples. A generic library helps discussion; it does not determine the impact of a defect in your product or destination.

Context changes classification. A color variation on an internal bracket can be minor; the same variation across a matched retail set can be major. A spelling error on an export carton may be cosmetic; an error in a mandatory warning can be a compliance issue. Put photographs, zones, thresholds, and edge cases into the brief before the inspector needs an instant answer.

Also define the count unit. One product may carry three scratches and a loose screw. Does the plan count one nonconforming unit, four nonconformities, or each category under a separate rule? The report must follow the method used by the selected standard and protocol. Otherwise, defect totals can look precise while the accept/reject calculation is not reproducible.

Make the sample trail auditable

The report should show how a large presented lot became the inspected sample. Capture the total presented units and cartons, location map, pallet/carton ranges, number of cartons opened, selection method, sample quantity by SKU/variant, and photographs of selection and opened cartons. If goods occupy several rooms or buildings, the route should reach each relevant area.

The factory should not hand the inspector a prepared “sample table.” The inspector selects from the defined lot. Note inaccessible pallets, sealed or wet cartons, finished goods held elsewhere, quarantined production, and any factory limit on opening cartons. If the lot cannot be randomly reached, the limitation belongs near the result, not in an unread appendix.

China's government supervision inspection is not the same procedure as a buyer's commercial PSI. Even so, the current Interim Measures for the Administration of Product Quality Supervision and Spot Checks, issued by SAMR as Order No. 18 and effective since 2020, reinforce useful evidence disciplines: planned sampling, written sample records, sealing, and defined procedures for retained samples and reinspection. Write the buyer's own chain of custody and reinspection rules rather than borrowing a government label.

For each sample unit, preserve an ID or carton link when practical. That lets the reviewer see whether several defect types cluster in one unit or carton, whether failures concentrate in a batch, and whether rework scope should expand beyond the observed pieces.

Read the report in seven layers

1. Identity: factory, address, date, PO, SKU, model, buyer specification, and inspector match the intended order.
2. Readiness: ordered, produced, presented, packed, unfinished, and excluded quantities are distinct; the lot was ready under the protocol.
3. Sampling: exact standard/edition, plan, level, lot, code/sample size, AQL, Ac/Re values, carton route, switching status, and count method are present.
4. Defects: every finding has class, quantity, unit/carton link where available, description, photograph, and agreed reference. Look for clustering and multiple defects on one unit.
5. Special tests: each dimension, function, barcode, packing, destructive, or powered test shows its own sample size, method, equipment, expected value, actual result, and limitation.
6. Packaging and quantity: assortment, accessories, artwork, labels, traceability, retail packing, carton count, marks, dimensions, and gross/net weights are compared with controlled values.
7. Limitations: missing sample, unavailable equipment, unperformed test, denied access, incomplete packing, or absent document is carried into the buyer decision.

Read measurements, not “looks OK.” A dimension row should show the specification/tolerance and actual readings. A barcode row should identify the code and scan result. A function row should describe the procedure and quantity. Report photographs support the record but rarely replace those values.

Annotate the report without modifying the original. Put reviewer questions, supplier answers, commercial disposition, and later correction in a separate dated response file. A revised report should preserve its revision history and reason rather than erasing the first result.

Case file: the report passes, the shipment stays held

A fictional buyer orders 2,400 private-label rice cookers for the UK. The controlled specification calls for 220-240 V, a UK plug, English retail packaging, one measuring cup, one spoon, and the correct scannable SKU barcode. Production is complete and packing is sufficiently advanced when the inspector visits.

The agreed workmanship sample is within its acceptance numbers. A few light surface marks are counted as minor. The listed assembly, accessory, short function, measurement, and barcode checks meet their rules. The report summary says PASS.

One general photograph, however, shows a two-pin EU cord beside a unit. The inspection brief never listed plug/cord identity as a product-specific check, so the inspector did not sample, count, or classify it. The factory later says early cartons contain UK cords but some later cartons were packed from EU cord stock while it waited for a delivery.

The buyer holds the lot. The workmanship plan did not fail; the protocol omitted a market-specific component and the presented lot may be mixed. The corrective record requires:

1. segregation of cartons by cord/plug type and batch;
2. a trace showing affected quantities, carton ranges, and packing dates;
3. complete correction of the UK order under a documented work instruction;
4. updated packing and component records;
5. a focused reinspection that samples randomly across the entire corrected lot and repeats plug, rating, function, accessories, barcode, and packing checks.

The buyer also updates future briefs. PASS did not authorize the wire or shipment because the summary could only answer the written protocol. The case is a specification-control lesson, not evidence that AQL sampling malfunctioned.

Close failure with containment, correction, and reinspection

After a failed or unclear report, do not negotiate only about whether a photograph “looks serious.” Turn the result into a corrective-action chain.

1. Hold the release milestone. Identify the affected payment, shipment, booking, or production action and the person authorized to change it.
2. Contain the lot. Mark and separate affected and potentially affected goods; stop loading or mixing while scope is investigated.
3. Trace scope. Use batches, lines, dates, operators, components, cartons, or pallets to test whether the observed issue is isolated or systemic.
4. Approve the correction method. Define the work instruction, acceptance evidence, treatment of replaced parts, and who verifies completion.
5. Reinspect the corrected population. Draw samples across the reworked lot under an agreed plan. Do not revisit only the photographed pieces or allow the factory to present hand-selected corrected units.
6. Keep every version. Original report, supplier response, root cause, rework record, reinspection, waiver, and final disposition remain separate dated artifacts.

A waiver does not turn a defect into conformity. If an authorized buyer accepts a quantified commercial deviation, record the units affected, customer/market impact, price or remedy, authority, and why release remains permissible. Safety, legal, or regulatory requirements may not be waivable by the commercial team.

Protect the lot after the inspector leaves

PSI is a snapshot. Goods can be reworked, mixed, damaged, relabelled, or substituted afterwards. Record inspected carton/pallet ranges, sampled-unit disposition, resealing method, inspection completion time, and remaining factory actions. Ask for final packing lists and photographs tied to the lot, but do not pretend a photograph creates custody.

For high substitution, quantity, or loading risk, use a separate loading control: verify container condition and number, seal, carton marks, quantity loaded, and linkage to the inspected lot. A PSI and container-loading supervision answer different questions. If the goods move to an outside warehouse, identify the handoff and any opportunity for mixing.

Align the inspection date with the factory and shipping calendar. A last-day inspection with no rework window creates pressure to waive findings. The cargo cutoff is a scheduling input, not evidence that the lot is acceptable.

Know what PSI cannot prove

Regulated products may require testing and certification outside an on-site inspection. For example, 15 U.S.C. Section 2063 requires certification for covered consumer products based on the applicable testing route and accredited third-party testing for covered children's products. An on-site PSI can observe the sampled units, but it cannot replace that product-specific compliance work.

For EU-related documents, use Check a Chinese Supplier's CE Documents. A visual PSI can confirm the model, mark, label, packaging, and controlled configuration presented; it cannot determine the entire conformity route.

Use results across a continuing series of lots

Keep order, product, factory, lot, plan/edition, sample results, defect patterns, special-test failures, rework, reinspection, returns, and customer complaints in one supplier quality history. The useful signal is not the number of green report covers. It is whether defect types and process changes recur.

ISO 2859-1:2026 includes switching and skip-lot concepts for a continuing series of lots. Do not move to reduced or skip-lot inspection after one clean shipment or apply switching rules selectively. Eligibility, history, reference plan, transitions, and return conditions must follow the adopted scheme. A supplier change in factory, product, tooling, components, process, or quality performance may require the plan to be reconsidered.

Repeated barcode errors suggest artwork/version control. Increasing loose fasteners can point to a process change. Clean PSI reports followed by field failures can show that the on-site tests do not reproduce real use. Tighten the control where evidence points: specification, process, special test, sample plan, audit, laboratory work, or loading custody.

Close the inspection before release

A defensible file can end with one precise note: “Lot PO-4821, 2,400 units at the named factory, inspected 15 July under ISO 2859-1:2026 normal single plan stated in the attached licensed worksheet; workmanship within Ac/Re; special plug check failed outside that calculation; lot held; all cartons reworked; reinspection R2 randomly sampled across the corrected population and closed plug, rating, function, barcode and packing checks; release approved by buyer QC on 18 July.”

That note does not say “zero defects.” It shows the lot, method, separate failure, correction, second sample, and decision owner. Keep it with the pre-payment evidence file.

Company and site questions remain separate. Use Factory Audit vs Company Verification and the ChinaValidate company search to establish who stands behind the contract and where capability was reviewed. Inspection establishes what was observed in the presented lot. This article is operational guidance, not product-specific statistical, engineering, regulatory, legal, customs, or contract advice.