One in five Canton Fair exhibitors is not a factory.

The registry says what the booth does not

Every Chinese company files a business scope with the state registry. It is the closest thing to a factual answer to “do you actually make this?” — a trading company cannot quietly file itself as a manufacturer, because the scope determines what it may legally invoice for.

Of the 5,252 exhibitors we profiled, 4,171 (79.4%) are registered with a manufacturing scope and 1,079 (20.5%) are registered as trading or import/export companies. Neither figure is a judgement about quality. A good trading company beats a bad factory. But a buyer who thinks they are talking to the production line, and is actually talking to a reseller, is mispricing the entire deal.

The address test works

Experienced importers pass around a rule of thumb: check whether the registered address sits in an industrial zone or an office tower. As far as we can tell nobody had ever measured it. So we did, across every exhibitor with a registered address on file.

The folk wisdom holds. Companies registered as traders are 5.4× more likely to be registered in an office tower than companies registered as manufacturers, and roughly 3.7× less likely to sit in an industrial zone. It is not a perfect test — plenty of real manufacturers register a sales entity downtown — but as a free, thirty-second check before a deposit, it earns its place.

The headcount gap

Company profiles quote an employee count. Social-insurance filings record how many people the company actually pays contributions for. For the 779 exhibitors where we have both, the median claim is 1.2× the insured number — most companies round up modestly, which is normal everywhere.

The tail is where it matters. 33.4% claim at least twice their insured headcount, and 14.6% claim five times or more. The most extreme cases in our sample:

• 4,300 staff claimed against 13 on social insurance — Taishan Group Taian Boao International Trade Co., Ltd.
• 200 staff claimed against 1 on social insurance — Wenzhou Boshi Safety Products Co., Ltd.
• 60 staff claimed against 1 on social insurance — Xinjiang Tianshan Laser Intelligent Technology Co., Ltd.

Some of this is legitimate: seasonal labour, dispatch workers, and unconsolidated subsidiaries all sit outside a single entity’s filings. But a factory claiming 4,300 staff while insuring 13 is not a rounding difference, and 76 companies in the same sample actually understate — so the filings are not systematically low.

And some traders simply say they are factories

Among traders with marketing copy we could read, 15 (2.5%) describe themselves outright as a manufacturer or factory. That percentage is a floor, not an estimate — we only counted unambiguous first-person claims, so anything phrased loosely slips through. The individual cases are the point:

What a buyer can do in five minutes

1. Read the registered business scope, not the brochure. Manufacturing scope is a legal filing; “factory” on a homepage is not.
2. Check the registered address. Industrial zone or office tower — the difference is measurable, as above.
3. Compare claimed headcount against social-insurance filings, where public.
4. Match the payment beneficiary name to the registered legal entity, exactly.
5. Ask for a live, unedited production-floor video rather than a showroom walkthrough.

Method and limits

Sample: 5,252 Canton Fair exhibitors profiled by FactoryBase as of 2026-07-26. Manufacturer/trader status comes from the registered business scope in China’s public enterprise registry, not from self-description. Address classification is keyword-based on the registered address and will misclassify some entries. Headcount comparison covers only the 779 companies where both a public social-insurance figure and a stated headcount exist, so it is not a random sample of all exhibitors. Self-description counts use deliberately strict pattern matching and therefore understate.

We verify individual data points; we do not certify companies, and nothing here says a named company is dishonest. Customs records, production capacity and certificate authenticity are not verified in this dataset. Figures update automatically as the underlying dataset is re-swept — the numbers above are live, not a snapshot pasted in at publication.

Free to cite with a link to this page. Underlying per-company evidence is public on each category page.