Most guides on choosing a machining supplier give you a list of questions to ask. That is the easy half. The difficult half is knowing whether the answers are true — and a supplier who is organised enough to win your business will also be organised enough to answer questions well.
This guide separates what you can verify from a desk, what you can only learn by placing an order, and the specific signals that tend to precede a bad experience.
Five things you can verify without visiting
1. What is actually in their business licence
Every Chinese company has a unified social credit code and a registered business scope (经营范围). The scope lists what the company is legally permitted to manufacture and trade. A company whose scope is “trade and import/export” without manufacturing is a trading company, whatever the website says.
Ask for a copy of the licence. Cross-check the company name on it against the name on the bank account you will be paying. If they differ, ask why in writing.
2. Whether they own the machines they photograph
Equipment photos are easy to borrow. Two checks that are harder to fake:
- Ask for a photo of a specific machine with a piece of paper showing today’s date and your company name next to the spindle. It is a small ask and takes them two minutes.
- Ask for the machine model and year, then check the work envelope against your part. A shop that cannot tell you the travel of its own mill is not running it.
3. Whether they have inspection equipment, not just machining equipment
Machining capability and inspection capability are different investments. A shop with five machining centres and one pair of calipers cannot certify a tolerance.
Ask specifically: do you have a coordinate measuring machine, what is its stated accuracy, and can you send a sample inspection report? A real report has a header, a drawing revision reference, measured values against nominal, and a pass/fail column. A photo of a part with a caliper on it is not a report.
4. Whether their past work matches your part
Ask for three examples of parts similar to yours — same material family, similar size, similar tolerance class. Then ask two follow-up questions about each: what was difficult about it, and what would they do differently.
A shop that has actually made the part will answer with specifics about fixturing or tool deflection. A shop that is improvising will describe the part in general terms.
5. How fast and how specifically they respond
Response speed is a proxy for how they will treat you during production. But the content matters more than the speed. A useful first reply does three things: confirms the drawing revision, flags two or three ambiguities, and gives a price with a lead time counted from a named event.
A reply that is only a price is a warning. It means either they did not read the drawing, or they are quoting optimistically to win the order and will renegotiate later.
Three things you can only learn from a trial order
No amount of document review substitutes for these. Budget for one trial order before committing to a production volume.
1. Tolerance stability, not tolerance capability
Any shop can hit ±0.05 mm once. The question is whether the tenth part in a batch is in the same place as the first. Ask for the dimensional report on the trial batch to include every part, not one sample. If they only measure one, that tells you something.
2. Whether the promised lead time survives contact with reality
Put a real deadline on the trial order and see what happens. More useful than whether they hit it is how they handle it: do they tell you two days before the deadline that it will slip, or do they go silent and then deliver late with no explanation? The first is a manageable supplier. The second is not.
3. How they respond to a problem
If the trial order arrives perfect, you have learned less than you think. Look for any small imperfection — a burr, a packaging scuff, a dimension at the edge of tolerance — and report it.
The response is the data point. A supplier who acknowledges the issue, proposes a fix, and absorbs the cost is worth keeping. A supplier who argues that the drawing was ambiguous, or that the measurement method is wrong, will do the same thing on a USD 50k order.
Designing a trial order that actually tests them
A trial order that only contains easy parts tests nothing. A useful trial has these characteristics:
| Element | Why |
|---|---|
| One part with a genuine difficulty | A thin wall, a deep pocket, a tight position tolerance, or a surface finish callout. Something that requires them to think about fixturing. |
| Three to ten pieces, not one | One piece is a prototype. Three or more reveals consistency. |
| A finish requirement | Tests their finishing partner, which is where most schedules actually break. |
| A stated inspection requirement | Ask for a dimensional report with all parts measured. See what you actually receive. |
| A realistic deadline | Not artificially tight, but real. Watch whether they flag risk early. |
This trial will cost more per part than a production run. Treat it as a fee for information, not as a purchase.
Red flags, ranked by how often they precede a real problem
| Signal | What it usually means |
|---|---|
| Price 40%+ below every other quote | Either a misunderstanding of the drawing, or a plan to renegotiate after the order is placed. Ask them to walk through the cost breakdown line by line. |
| Refuses a live video walk-through of the shop | Either no shop, or a shop they do not want you to see. Both are problems. |
| Answers questions instantly with no follow-up | Means the questions were not actually considered. Real answers to a tolerance-stack question take a day. |
| Bank account name differs from the company name | Common with trading intermediaries and a genuine fraud signal in some cases. Always ask for a written explanation. |
| Will not put the payment condition in writing | Whatever they say verbally about “we will make it right” is not a term. If it is not on the PI, it does not exist. |
| Sample made on a different process than production | The approved sample does not predict the delivered batch. Ask which machine will run production. |
What a factory visit should check (if you go)
For orders above roughly USD 20k, a visit usually pays for itself. Twenty items worth looking at, grouped by what they tell you:
Production capability — machine count and type, machine age, whether machines are running or idle, spindle hours, tooling inventory, whether fixtures are dedicated or improvised.
Quality system — inspection room presence, CMM and its calibration sticker date, whether inspection records are filed or created when a customer asks, material certificate filing, non-conformance log.
Operations — material storage (indoor, labelled, or stacked outdoors), work-in-progress tracking, whether parts have traveller cards, packaging area, how finished goods are protected.
People — ratio of programmers to operators, whether the person who quoted you is on site, how long key staff have been there.
An idle shop with clean floors and organised material storage is a better sign than a busy shop with parts piled on the floor.
Frequently asked questions
Is a trading company always a worse choice than a factory?
No. A good trading company can coordinate machining, finishing and assembly across several shops and take responsibility for the whole chain, which is genuinely useful for complex assemblies. The problem is only when a trading company presents itself as a factory, because then you cannot tell who is accountable when something goes wrong. Ask directly and expect a straight answer.
How do I verify a supplier’s business licence?
Ask them to send a copy of the licence with the unified social credit code visible. The name, legal representative and registered scope should match what they have told you. Cross-check the company name against the bank account details on the proforma invoice.
What is a reasonable trial order size?
Three to ten pieces of one part, with a genuine difficulty in it. Small enough to be affordable, large enough to reveal consistency. Expect the per-part price to be significantly higher than production pricing — that is normal and not a signal of dishonesty.
Should I ask for references?
Yes, but weight them lightly. Suppliers choose which references to give you. A more useful question is whether they can describe a project that went wrong and what they did about it — a supplier who cannot recall any problem either has not been running long or is not being straight with you.
How long should vetting take?
Two to four weeks for the desk research and the trial order combined. Compressing it to a week usually means skipping the trial, which is the only part that produces real information. If a deadline forces you to skip it, at minimum hold back the balance payment until you have inspected the first batch.
Start with a drawing
The cheapest way to evaluate a supplier is to send one drawing and study the reply. Send it with material, finish and quantity to our team — a useful quote comes back within 24 hours on working days, itemised, and with the ambiguities in your drawing named rather than guessed at.
If you are still working out the commercial side of the order, the companion guide on importing machined parts from China covers payment terms, Incoterms and customs documents. And if you want to see how an itemised quote for CNC machining service is structured, that page walks through the line items.
More guides on supplier vetting, payment terms and import logistics are on our CNC machining guides page.
