Most articles on this subject describe a supplier who takes a deposit and disappears. That does happen, it is the easiest failure to prevent, and it is not where importers actually lose money. The expensive failures arrive on time, in the right quantity, from a company that genuinely exists — and they are caught by completely different checks from the ones on a standard due-diligence list.
The checks that feel thorough and stop almost nothing
Before the useful ones, three that get done constantly and carry less weight than they appear to.
- Being sent a photo of a business licence. The licence itself is genuinely the most useful document you can ask for — but a JPEG is not a licence check. The value comes from reading what is on it and confirming it against the public registry, which is what factory or trading company works through. An image in a chat window establishes only that an image exists.
- A video call from the factory floor. Useful for a conversation, close to worthless as verification. You are being shown a location, and the person holding the phone chose it.
- A platform verification badge. It confirms documents were submitted. It says nothing about production capability, whether the listing owner is the maker, or what the goods will look like at volume.
None of these are useless. The problem is the confidence they produce relative to what they actually rule out.
The golden sample
You approve a sample. It is excellent. Production arrives and it is visibly worse, and the supplier is genuinely puzzled that you are unhappy.
Often this is not deliberate. The sample was made by the best operator in the workshop, slowly, with hand-selected materials and no line pressure. Nothing about that process survives contact with a production run — which means the sample was never evidence about production, only about capability under ideal conditions.
What stops it: a sealed counter-sample held by both sides, and then checking production units against it while the line is still running rather than at the end. That is a during-production inspection, and it is the only stage that can still change the outcome of the order it is inspecting. The reasoning is in when to inspect a China order, and the sealing and approval process itself is in sample approval.
What does not stop it: approving a better sample. A second sample made the same way tells you the same thing the first one did.
Quality fade
The first order is right. So is the second. By the fifth, the finish is duller, the packaging is thinner, a fastener has changed, and nothing was ever announced.
This is the most under-discussed failure in the industry because it is structurally invisible. Each order in isolation passes. A pre-shipment inspection on order five reports the same defect rate it always reports, because the inspector is checking against a specification that has not changed — while the product quietly has.
What stops it: tracking results across orders rather than within one. The sampling standard already contains the logic: if two lots in any five are rejected, inspection tightens, and if a supplier cannot get back off tightened inspection the standard says to stop inspecting and address the supplier. Almost nobody applies this, because it requires keeping lot history across purchase orders. Keeping it costs nothing. The mechanics are in AQL sampling explained.
What does not stop it: a good relationship. Quality fade is usually margin pressure, not malice, and a warm relationship is exactly the condition under which nobody wants to raise it.
The certificate that was never issued
A supplier sends a test report with a familiar laboratory's logo. It looks correct because it is a copy of a real report — sometimes for a different product, sometimes for a different company, sometimes edited.
This one is unusual in that it is cheap and definitive to check, and hardly anyone does. The laboratories that issue these reports run their own verification:
- SGS document verification
- Intertek certificate validation
- TÜV Rheinland publishes certified products in its Certipedia directory
Three things are worth reading on any report before you even verify it. Whose name is the applicant — a report issued to a different company tells you nothing about your supplier. What the scope covers — a report on one model does not carry to a range. And what was tested against, because a report can be entirely genuine and still test to a standard your market does not accept.
Short shipment
The cartons are counted, the count matches the packing list, and the order is short — because the shortfall is inside the cartons, not in the carton count.
What stops it: specifying that a proportion of cartons is opened and piece-counted rather than counted and multiplied. It is one line in an inspection brief and it is left out constantly. The rest of what belongs in that brief is in the inspection checklist.
Two that are decided by one line in the purchase order
Neither of these is fraud in any prosecutable sense. Both are expensive, both are common, and both are settled in advance by a sentence nobody thought to write.
The tooling you paid for and do not own. Custom products often carry a mould or tooling charge, sometimes substantial. Buyers routinely assume that paying for a mould means owning it. In practice, unless ownership is stated, the tooling stays in that factory — and you discover this at the exact moment you want to move production, which is the moment it has the most leverage over you.
The fix is a clause, not a negotiation tactic: state that the tooling is your property, that it will be released or transferred on request, and that it is marked with your name. A factory that intends to behave normally has no reason to object. One that refuses has told you what the charge actually bought.
The order that was made somewhere else. You verified a supplier, possibly audited the line, and the goods were produced in a different workshop because capacity was tight or the order was small. Your audit describes a factory that did not make your product.
This is legitimate practice and sometimes unavoidable. The problem is only that it happened silently, because everything downstream — the defect classification, the process questions, the corrective action — was agreed with people who were not making the goods. Require disclosure of subcontracting in writing, and make a during-production check conditional on knowing where production actually is. If the answer to "which site is this order running on" is vague, that is the finding.
Which check stops which
| Failure | What actually stops it | What people do instead |
|---|---|---|
| Deposit taken, nothing ships | Paying through a platform that carries cover, or staging payments against verified events | Checking the badge, then paying off-platform anyway |
| Bank details changed mid-order | Voice confirmation on a number held beforehand | Reading the email more carefully |
| Golden sample | Sealed counter-sample plus a during-production check | Approving another sample |
| Quality fade | Tracking lot results across orders; tightening on the rule | Trusting a long relationship |
| Fake test report | Verifying the report number with the issuing laboratory | Looking at the logo |
| Short shipment | Opening cartons and piece-counting a sample | Counting cartons |
| Tooling you cannot take with you | Ownership and release stated in the purchase order | Assuming that paying for a mould means owning it |
| Order made at another site | Written disclosure of subcontracting; confirm the site before the in-production check | Auditing one factory and assuming it stays that way |
| Reseller posing as maker | Registered business scope plus operational questions | Asking whether they are a factory |
The pattern across that table is that the effective check is almost always more specific and less impressive than the one it replaces. Opening eight cartons is not a satisfying thing to put in a process document. It is the one that finds the shortfall.
The one that is not a scam at all
Worth saying plainly, because it gets mixed in with the rest: a supplier quoting a price that cannot produce the product you specified is usually not defrauding you. They are quoting the product they intend to make, which is a cheaper one, and assuming the specification will be negotiated down later — as it often is.
The defence is not suspicion. It is a specification precise enough that a cheaper interpretation is visibly non-compliant, and a defect classification that says so in writing before production starts. Vague requirements do not get honoured generously.
Where payment fits into all of this — what each instrument covers when one of these does happen — is in how to pay Chinese suppliers. The bank-details fraud has its own page: our payment security policy.
