Samples, inspection and batch quality

Usually yes, you pay — and whether it comes back depends on what was agreed, not on what the sales rep said.

Three arrangements are normal, and none of them is "the standard":

• Stock sample: often free, you pay the courier.
• Custom sample: charged, and sometimes credited against the first bulk order once it reaches an agreed volume.
• Tooling, printing plates, hot-stamping dies, cutting dies, inserts: one-off start-up costs. These are normally not credited, and they are normally not included in the unit price.

Two things I check for you:

• Whether "free" is real. Compare the unit price in a "free sample" quote against a quote that charges for the sample separately. Free samples are sometimes paid for in the unit price.
• Whether the credit condition is written. What volume triggers the credit, when it is paid back, and what happens if you order less.

What I can't promise: I cannot promise you a free sample or a refunded sample fee. That is the factory's commercial decision. What I can do is get the condition in writing before you pay.

If the courier cost is the problem — it often is — ask to use your own courier account. One buyer was quoted $50 to ship a sample weighing under 1 kg, and compared it with a supplier in Poland that charged $15.

Normal — for standard products it is the most common arrangement. Fully free, product and shipping, is the exception rather than the standard.

What is not normal is finding out the freight cost after you have agreed. So ask for two numbers before you say yes: the sample price, and the freight to your address with a named courier and service level.

And watch the pattern: cheap shipping on the sample, expensive shipping on the bulk order. Ask for the bulk freight basis at the same time, in writing.

What I can't promise: I cannot give you a freight figure I have not been quoted. Freight moves, and anyone who quotes a firm number without checking is guessing.

One UK buyer on a public forum was quoted $150 for a single sample against a bulk price of $85, plus $250 in courier charges — which made comparing three factories cost more than the product. If you plan to compare suppliers, decide the sample budget before you start, not after.

This is the right question, and the honest answer is often no.

A sample can come from a trading company's shelf, from an earlier batch, or from a sample room that is not the line that will run your order. So ask in writing: which workshop, which line, which material batch, and who signed it off. Then ask for photographs of that line producing a unit — not a showroom video.

When I check a supplier, I ask for the sample's production record and compare it with the line scheduled for your order. Where I can get there, I look at both.

What I can't promise: from a desk I cannot prove where a specific sample was made. I can tell you what the factory put in writing, what the records show, and what is missing — and I will say "not verified" when that is the answer.

One more question worth asking before you pay: will your order be run in-house, or subcontracted? That answer belongs in the contract, not in a chat message.

They are three different things, and approving one does not approve the others:

• Structural / white sample: unprinted. Checks fit, size, strength. It says nothing about colour.
• Digital proof: checks text, layout, barcode position. It does not approve colour or finish.
• Pre-production sample: printed and finished the way the final product will be — colour, registration, foil, embossing, lamination, die-cutting. This is the one that should become the standard.

The standard itself is the golden sample: a physical unit, signed and dated by both sides, one kept by you and one by the factory, with the file version number written on it. If I am involved, we keep a third copy.

The mistake I see most often is approving structure from a digital file, then discovering a colour or foil problem on the bulk run. The cheapest step was used to carry the risk of the most expensive one.

What I can't promise: I cannot make bulk identical to a sample. Manufacturing has tolerances. What I can do is get the acceptable and unacceptable ranges written down as numbers while the sample is still in front of us.

More than you think, and the first sample rarely passes. One sourcing agency publishes a range of two to four rounds before a product is right; that is one company's published experience, not an industry statistic. What matters is that the rounds are planned and dated.

Get four things into the quotation:

• How many revision rounds are included, and what an extra round costs.
• The delivery date of each round, in writing.
• When each round starts, and when you will receive photos or a physical sample.
• What happens if a round is late: a new date, and the effect on the production slot and the shipping date.

Two things I do during sampling: a written status update at each agreed checkpoint, and escalation when a date slips — a written reminder first, then a request for in-process photographs, then a written statement of the impact on your launch date.

What I can't promise: I cannot guarantee a factory keeps its sampling dates, and I cannot make it faster. What I can do is make slippage visible early instead of at the end.

One caution about the evidence: the detailed public complaint I found about sampling dragging on for months was filed by a buyer inside China, not an overseas buyer. Treat it as an illustration of the mechanism, not as a typical case.

Three separate questions, and they usually have three different answers.

1. Are the plate and die fees one-time? In paper packaging, printing plates, hot-stamping dies, cutting dies and insert moulds are normally separate one-off start-up costs, not included in the unit price. On a repeat order of the same product the plate fee is normally not charged again — but the plates are normally stored by the factory, and the wording matters.

2. Who owns them? Under Chinese law this kind of job is a work contract, and paying for a tool does not by itself transfer ownership. Ownership has to be written into the contract, or it stays where it is.

3. Can they hold my tool? Between businesses, Chinese law allows a lien across legal relationships, and the Civil Code article on the contractor's lien contains an "unless the parties agree otherwise" exception. That sentence is the most valuable one in a tooling clause, and it has to be drafted by your own Chinese lawyer for your specific entities.

What I do: an asset list (what exists, what it cost, where it is stored), a keyword scan of your contract that marks only "written / not written" on each point, and a physical check of tool numbers and stamps when I visit.

What I can't promise: I do not give legal advice on ownership or liens, and I cannot recover a tool. I can tell you what your documents do and do not say — before you pay.

Two answers.

1. Inspection is worth it, with one honest limit. Inspectors regularly find what you cannot see in a photo: colour, sizing, weight and thickness off spec — measured, photographed and documented. Buyers who have been through a loss tend to say the same thing: do third-party inspection regardless of how good the relationship feels. But an inspection is a release gate on a finished shipment, not a verdict on whether a supplier is trustworthy. If you are worried before you pay, the tool you need is a supplier audit, not a final inspection.

2. AQL 2.5 does not mean "at most 2.5% defects". In the statistical definition, AQL is the producer's point on the operating-characteristic curve — the quality level the plan is built to accept from the producer's side. A lot at exactly that defect level has a high probability of being accepted. If you want protection on the buyer's side, you have to specify the other point as well: the rejectable quality level (RQL / LQ) and the consumer's risk you are willing to accept — and then check that the sample size and the accept/reject numbers follow from that.

What I can't promise: I do not sell inspection, so I have no reason to sell you more of it — and I cannot give you a single AQL number that is "safe" for your product. I also will not quote an inspection price as if it were an industry standard; the only published figure I found was one provider's own rate card.

For roll film and flexible packaging there is a further problem: AQL tables count defective units, but the failures that hurt you — seal strength, barrier (oxygen and water vapour), solvent residue, print registration — are measured, not counted, and defects in a roll are spatially correlated. Counting plans understate them.

Use sources you did not get from the supplier. That is the whole method.

• Report number: check it in the official system, and remember the data is uploaded by the labs themselves.
• Accreditation: check that the body's accreditation is current and that each test method on the report falls inside its scope. A body can be accredited and still not accredited for your test.
• Certificates (ISO 9001 and similar): check the certification body and the certificate status through an official database. ISO itself does not certify anyone — a company cannot be certified by ISO; certification is performed by external bodies.
• Lab status: check whether the lab has had an accreditation withdrawn. In January 2026, CPSC withdrew accreditation from four China-based labs — Shenzhen GTT (Lab ID 1843), Dongguan True Safety (1755), Fujian Berton (1857) and Shenzhen HUAK (1710) — after finding unreliable or falsified reports and concealed losses of accreditation, and said companies relying on those reports must obtain new testing from properly accredited laboratories. Accreditation is a status to re-check, not a fact you check once.

Two traps worth naming. Scanning the QR code on a report proves nothing: a fake lab website with a working lookup page has already been documented, where entering the fake report number returns the fake report. And "the number is not in the system" is a flag, not proof — the issuing lab is responsible for uploading it.

What I can't promise: I cannot guarantee that I will detect a well-made forged document. SGS itself says a report number alone is insufficient to verify a report, and that verification requires the complete document, with no missing pages, erasures or compression. What I do is record which sources I checked, when, and what each one returned.

Both are useful, but they answer different questions — and you should know which one you are paying for.

• Factory-sent: cheaper and faster. It shows that the material the factory chose can pass. It says nothing about whether the units that reach you are the same. In third-party compliance terms, only reports issued by accredited laboratories are considered valid, and a report a manufacturer prepares itself has no official validity.
• You-sent: you control which unit goes, from which carton, at which time. This is the version that supports a claim, and it lets you keep a chain: order number, batch number, carton number, courier record, lab receipt.

If you are sending, decide three things first: the exact test method and conditions (food simulant, time, temperature — matched to your real use); which article is tested (film, printed pouch, or the packed product); and who owns the result.

A middle path that works well: the factory does the development testing; you pay for one confirmation test on a production unit, at a lab you chose; and you repeat it when the material or the process changes.

What I can't promise: I do not quote testing prices. I could not find any authoritative published price list for migration testing per SKU, and any number I gave you would be beaten by a cheaper quote the next day. I define the test scope, then get you live quotes from two or three labs.

Look at your contract before you look at the report. At that point the goods are made, the deposit is paid, and the ship date is close — a failed report is evidence, not leverage. Whatever leverage exists was written earlier.

Four things to have in writing:

1. The defect definitions and the acceptance criteria, with standard numbers and versions.
2. Who pays for the first inspection, who pays for a re-inspection, and under what conditions.
3. The three possible outcomes, defined in advance: rework, accept with a price adjustment, or reject — and what triggers each one.
4. The balance release condition, tied to a passed inspection.

Then act in this order: send the defect list as numbers with photographs and measurements; ask for a written plan with dates; and decide whether a re-inspection is worth paying for. It usually is — paying to verify a correction is cheaper than accepting an unverified one.

What I can't promise: I do not hold your money, I do not advance anything, and I cannot force a factory to rework. I do not sell inspection either, so I have no commission on the second visit. What I can do is build the inspection specification and the payment condition so that a failed inspection actually changes something.

Write numbers, and write the conditions with them — otherwise the number means nothing.

Do not write "colour variation ±5%". In printing, a percentage like that refers to dot area, not to the colour difference your eye sees. The two do not convert linearly, so both sides can read the same clause their own way.

Write it like this instead: CIE DE2000, illuminant D65, 10° observer, SCI (or SCE — pick one), instrument model or equivalent, calibration interval. Then state the limit. Then check the limit against what your brand can live with.

Because this is where the argument usually is: ISO 12647-2's process tolerances for solid colour are ΔE*ab ≤ 5 deviation and ≤ 4 fluctuation. Brand owners usually want ΔE 2–3. So when your printer says "within standard tolerance", that may be true and still not be what you asked for. Those are two different claims.

What I can't promise: I cannot set your tolerance for you — it depends on your product, your retail channel and your customer. What I do is put the measurement conditions, the measured values and the dispute route in writing, so the argument becomes arithmetic instead of adjectives.

Practical extras: require each batch's absolute L*a*b* values, not just the ΔE; keep first-batch and new-batch stock separate while a batch is being accepted; and never compare the best unit of the old batch with the worst unit of the new one — a few hand-picked pieces cannot establish a lot's defect rate.

You don't — unless you contracted for notification and for records that make the change visible. This is the most common downgrade and the hardest to find afterwards. Five grams per square metre lighter, a thinner PE layer, a different adhesive grade, a new ink supplier: the appearance does not change, and the seal strength, barrier and residue do.

What to put in the contract, as one clause:

• A list of changes that must be declared and approved in writing before production: different material grade, different resin, adhesive or ink supplier, a new source of substrate, lower grammage, changed process parameters, subcontracted operations, a different production site.
• "Verbal approval is not valid or effective."
• The obligation extends to the factory's own suppliers.

This is not my invention. It is how automotive supply chains have worked for years: one major supplier's manual requires written approval before any such change, states expressly that verbal approval requests are not valid, and applies the same rule down the supply chain. One car maker requires change requests at least 12 weeks before introduction. Packaging has no such tradition — which is exactly why it has to be written in.

Three ways I verify rather than trust: incoming-material and batch records; the grade on the material certificate against the grade in your specification; and scheduled weight and thickness checks on the film — plus targeted testing when the numbers look wrong.

What I can't promise: I cannot guarantee that a factory does not substitute. A normal pre-shipment inspection structurally cannot catch it — in one documented Chinese case the appearance, dimensions and printing all passed while the solvent residue failed. And a supplier holding an ISO 9001 certificate proves nothing about a batch: ISO itself says a company cannot be certified by ISO.

Signing is the easy part. Three things decide whether the agreement is worth anything:

1. Where the dispute is heard. A domestic arbitration award in China can be refused enforcement for reasons that would not apply to a foreign-related or foreign award — insufficient evidence to enforce, or a finding that the law was wrongly applied. The forum you chose changes your options before the dispute even starts.
2. Whether the supplier has assets. On the evidence I have read, the most common reason an award is never enforced — in China or anywhere else — is that there is nothing to enforce against.
3. Whether the cost is proportionate. The smaller the claim, the more fees, translation and time eat the recovery.

That is why the effort belongs before the dispute, not after: a sealed golden sample, a written batch-consistency specification, a change-notification clause, inspection records, and payment terms that hold something back until the goods pass.

What I can't promise: I do not litigate, I do not collect debts, and I do not promise that an agreement will be honoured. One source I rely on here is a law firm analysis from 2011 — the structural point about assets still holds, but treat the specifics as needing a fresh legal check. And do not assume the UN Convention on Contracts for the International Sale of Goods applies to you automatically: the UK is not a contracting state, and whether it applies depends on the governing law you chose.

Keep more than you think, and keep it properly.

From the first commercial batch — not the development sample — keep a sealed, signed and dated unit for each party, plus:

• The batch number, printed or labelled on the carton or bag, matching your PO one-to-one.
• The measured values of the key characteristics for that batch: absolute L*a*b* values, grammage, thickness, seal strength, residue, barrier.
• Photographs of the sealed unit with its label, the date and an ID.
• The approval email or chat thread that established the standard.

Storage matters: dark, dry, sealed, labelled, away from a window. A retained unit that has spent a year in sunlight is not evidence of what left the factory.

Do not use one unusually good unit from the first batch and one visibly poor unit from a reorder to represent both shipments. And while a new batch is awaiting acceptance or under investigation, keep the old stock physically separate.

What I can't promise: a retained unit does not replace a written specification, and it does not prove a defect rate. It shows what you approved and what arrived — that is its job.

Start from the failures, not from a catalogue of tests. China's national spot-check notices are public, and the pattern in films and bags is consistent: barrier performance (oxygen and water vapour), solvent residue (total and benzene series), and seal strength.

• 2023: of 169 composite film and bag batches sampled, 7 failed — 4 on total solvent residue, 3 on oxygen barrier, 1 on water-vapour barrier.
• 2024: among non-composite film bags sampled, 8 failed on oxygen barrier and 2 on water-vapour barrier; in composite film bags, 1 failed on total solvent residue.
• 2025: of 33 cling-film batches, 5 failed on oxygen barrier and 3 on water-vapour barrier.

So a practical PO annex for flexible packaging covers: solvent residue (total and benzene series), oxygen and water-vapour barrier, seal strength, grammage, thickness and material structure, and colour. Each line needs the standard number, the unit, the limit and the test conditions.

Two caveats that matter. First, a Chinese recommended standard is not law until you make it contractual — it becomes binding between the parties when it is agreed into the contract. Second, the tolerance in a standard may be looser than your brand needs: the point of the annex is your limit, not the national one.

What I can't promise: I cannot tell you the minimum legal test set for your product and market without checking your specific use, and I do not quote test prices. I define the scope and get live quotes. I also have not verified the GB/T 10004-2008 text myself — the widely quoted change (total solvent residue from ≤10 to ≤5.0 mg/m², benzene series to "not detected") comes from industry reporting, so check it against the standard before you put a number in a contract.