By Nicole · product development, Dongguan Mingmei · updated November 2026 · 9 min read

Most sourcing mistakes are made before the first sample, not after it. By the time a sample arrives you have already decided who to trust, usually on the basis of a fast reply and a low price. This is the list we would use if we were buying from someone else.

1. First, is it a factory or a trading company?

Not a moral question — a trading company can be useful. It is a question about who controls your lead time, your quality and your design changes. Three quick tests:

• Ask an engineering question. "What load cell do you use at 5 kg, and how do you compensate drift?" A trader forwards it. A factory answers with a part number, or says honestly that they would check with the engineer and come back in an hour.
• Ask where the mold is. If the tooling is at a third-party mold shop, that is normal — but they should know its address and the name of the shop. Vagueness here is the tell.
• Ask for a live video of the line. Not a promotional video. Something filmed now, walking past the assembly bench. Anyone can refuse; the way they refuse tells you a lot.

2. The twelve questions

3. Five red flags

• A price that does not move when the spec does. If a bigger platform, a better load cell and a USB version cost the same, the quote was never built from parts.
• Certificates with no model on them. A report for a different housing is not a report for your product. See the certificate guide.
• "Any colour, any packaging, any quantity." Real constraints exist. A supplier with no constraints has no factory.
• Answers that arrive faster than the question could be read. Long, instant, generic replies are a template, usually from someone who has not opened your message.
• No camera, ever. In 2026 a factory can film a bench in thirty seconds. Repeated excuses about a phone, a manager or a network are a signal.

4. The sample protocol — one afternoon, six tests

Order two or three units, not one. One unit tells you what a sample can be; three tell you what production is. Then run this, in this order:

Then keep one unit sealed as your golden sample, and write the test results into the purchase order. Without a written reference, "the production batch is different" becomes an argument you cannot win.

5. Certificates: verify, do not admire

A photo of a certificate proves a PDF exists. Check three things: the model name on the report matches what you are buying, the standard matches your destination market, and the lab can confirm the report number if you contact them directly. A valid report for the wrong model is worse than no report, because it makes you stop asking.

6. MOQ honesty

MOQ is not a number, it is a cost structure. On an existing mold, 500 pcs with your logo is reasonable. A new colour and packaging pushes it to a thousand-odd. A new housing is a tooling project and lives in the thousands. If a supplier quotes a private mold at 300 pcs, either the tooling is amortised into an inflated unit price, or the mold is not new.

7. Same questions, our answers

Fair is fair — here is how we answer the twelve:

• Own design: nine models in current production, tooled in our building in Dongguan, with CNIPA design patents registered 2023–2026. The rest of the catalogue is public mold and we say so.
• Capacity: one assembly line. Small by design; it is why a custom housing takes weeks rather than quarters.
• QC: 100% function check and calibration on every unit, packing on the same line.
• Certificates: valid CE-EMC, RoHS and food-contact reports for current models; other marks arranged per project and quoted before the order.
• Who you talk to: Nicole, who runs sampling and quoting. No sales layer between you and the person who decides.

Related: private mold vs public mold · certificates by market · how we develop a model