The scenario every buyer knows
You sent the drawing on Monday. The quote didn’t come back until Thursday — “two weeks’ lead time,” they promised. You followed up in week three. “Almost done.” The parts arrived in week five. Two of them were out of tolerance, and the surface finish didn’t match the sample they’d sent you.
If this sounds familiar, you’re in good company. Sourcing machined parts from China is how thousands of hardware companies stay competitive — and it’s also how hundreds of projects quietly die.
The difference between a smooth vendor relationship and a nightmare is rarely price. It’s the information nobody shares until you’ve already committed: how the shop actually runs, where lead time really goes, and which questions actually reveal the truth.
I run a 30-machine CNC shop in Shenzhen. This guide is the walkthrough I wish every buyer had before they emailed us — or anyone else. Use it to interview suppliers, audit quotes, and separate the real workshops from the middlemen with a web page.
Why sourcing in China feels like a gamble
Before the checklist, it’s worth naming why this market is so hard to read. It’s not that good shops don’t exist — it’s that the signal-to-noise ratio is brutal.
Anyone can look legitimate. A website, some photos of machines, a logo. The operation behind it may have three machines or three hundred. There is no central registry that verifies what a supplier actually owns.
Salespeople sell; engineers deliver. At many Chinese suppliers, the person answering your email has never touched a machine. They quote from a spreadsheet, promise from a script, and forward your technical questions to a workshop that’s too busy to answer them. The shop that lets you talk to an engineer is the exception, not the rule.
Time zone plus language means slow feedback loops. Every round-trip costs you a day. If the supplier’s first reply took 24 hours, multiply that by every question you’ll ever ask. Response speed in week one predicts response speed in week forty.
“Quality” means different things to different shops. A ±0.1 mm tolerance on a bracket is generous. On the mating face of an assembly, it’s a disaster. A supplier who doesn’t ask how your part will be used is quoting blind — and so are you, if you don’t tell them.
None of this is mysterious. Below are the checks that reliably separate a workshop from a facade.
Seven checks that separate a workshop from a facade
These aren’t about size or price. They’re about whether a supplier can actually deliver what they claim.
1. Ask for a machine list, not a machine count
A raw machine count tells you nothing. “Thirty machines” could be thirty identical VMCs or a balanced cell of 5-axis, 4-axis, turning, and wire EDM — and those are completely different suppliers.
Ask for the exact list: brands, models, axis counts, maximum travel, and what they run for finishing work. A shop with real capability sends it within an hour. One that stalls is either embarrassed by the list or doesn’t have one.
2. Materials with documents
Any shop can cut aluminum. Fewer can prove the material is what the invoice says. Ask: Can you supply material certificates (MTC) for 17-4PH, Ti-6Al-4V, or 630 stainless? Which alloys do you keep in stock?
A shop that stocks 6061 and 304 but orders everything else will quote you longer lead times — or quietly substitute material to hit the date. Both outcomes are worse than a longer honest quote.
3. Inspection capability, not just a caliper
A caliper checks a bracket. A CMM (coordinate measuring machine) checks a part with real GD&T requirements. Ask what inspection equipment they own and whether they issue inspection reports. If the answer is “we check everything by hand,” your tolerance stack-ups are running on their guesswork.
4. In-house finishing, or outsourced?
Anodizing, plating, heat treatment — each is usually a separate factory. The moment a part leaves the machine shop’s building, its schedule leaves the machine shop’s control.
A supplier who outsources finishing can’t honestly promise a date until they’ve confirmed the finisher’s queue. A supplier who finishes in-house — or has worked with the same finisher for years — can. This one question predicts more delays than any other.
5. How they commit to a date
Anyone can say “about two weeks.” Watch what happens when you ask: “Can you put the lead time in writing, broken down by machining, finishing, and inspection?”
A serious shop gives you the breakdown and tells you what would change it. A vague shop gives you a number and a prayer.
6. Response speed is the mirror of internal process
A quote that takes three days usually means the salesperson is waiting on the workshop, who’s waiting on the programmer, who hasn’t been told the order exists yet.
Fast quoting isn’t a personality trait — it’s evidence that estimating, material lookup, and scheduling are actually connected. When a shop quotes you in two hours, it’s not being pushy. It’s showing you its operating system.
7. Who answers your technical questions
Email the supplier a DFM question — something like: “The internal corner here is R1.5, but a standard end mill needs R2. Can you confirm that works, or should we redraw?”
A shop where an engineer replies has a different culture from a shop where a salesperson replies “no problem.” The phrase “no problem” from someone who has never cut metal is the most expensive sentence in sourcing.
The truth about lead time: where the days actually go
Here’s the mental model almost every buyer gets wrong: they assume lead time ≈ machining time.
It isn’t. For a typical small-batch CNC order, machining is a minority of the calendar:
| Stage | Typical time | Who controls it |
|---|---|---|
| Quoting | 1–3 days | Sales + engineering |
| Scheduling | 0–5 days | Shop floor — waiting for machine time |
| Material | 0–3 days | Stock vs. ordered |
| Programming | 0.5–2 days | CAM engineer |
| Machining | 1–5 days | The actual cut |
| Finishing | 2–7 days | Often outsourced |
| Inspection | 0.5–1 day | Quality control |
| Packing & shipping | 1–3 days | Logistics |
Do the math: on a “two-week” promise, the machining itself might be three days. The other eleven are queue, finishing, and paper.
Two things inflate lead time more than anything else:
Finishing is the black box. When a shop says “two weeks,” ask which part of that is anodizing — and whether the finisher is in-house. If the part has to go out to a third party, you are trusting two schedules, and only one of them belongs to the shop you hired. Prototype buyers routinely lose a week here without anyone lying to them.
Small orders get squeezed. Shops that serve large production clients will slot your 20-piece prototype around their production runs. Your parts ship “when a machine frees up.” A shop that runs a dedicated prototype cell — machines reserved for short-run work — has a schedule that doesn’t depend on someone else’s production forecast.
Four questions that expose a weak lead-time promise
You don’t need to audit their ERP. Four questions do the job:
1. “Where is the surface finishing done — in-house or outside?”
If it’s outside, ask for the finisher’s name and follow up yourself. A shop that hesitates here is a shop that will “discover” a delay in week two.
2. “If I place the order on Wednesday, when does it actually go on a machine?”
This answer separates “we start making it” from “we put it in the queue.” A few days of queue is normal. A month is a red flag — unless you were told about it upfront.
3. “What do you do when you’re going to be late?”
A good answer is a process: “we tell you by day X, with a new date and the reason.” A bad answer is “we’re usually not late.” Every shop is late sometimes. The ones who survive are the ones who warn you early.
4. “Can I see your live production schedule?”
You won’t always get access — but the reaction tells you everything. A shop that can screenshot its actual production board (even blurred) has a real schedule. A shop that can’t show you any schedule is running on a whiteboard and good intentions.
Five ways to shorten your own lead time
A good supplier will tell you this straight: part of the lead time is in your drawing, not in their shop.
1. Loosen tolerances where the part doesn’t care. Every ±0.01 mm feature means a slower setup and a slower inspection. Ask your engineer which surfaces actually mate — tighten those, let the rest breathe. On many prototype orders this alone cuts both price and lead time by 15–30%.
2. Use standard materials. A bracket in 6061 that’s in stock ships this week. The same bracket in a hard-to-source specialty alloy waits for material to land. If a material spec is a legacy habit, question it.
3. Reduce finishing steps. Every extra process — anodize this face, bead-blast that one, then a second pass for the logo — adds a hand-off and a day. Combine finishes wherever the design allows.
4. Send STEP and PDF. The PDF is for review; the STEP is for programming. Sending only a 2D PDF forces the shop to rebuild your model — a new source of both errors and delay. Sending a poorly exported native CAD file is worse.
5. Tell us the real quantity — including what comes next. “Two pieces now” with “we’ll need two hundred next year” changes the conversation completely. The shop can plan tooling and material properly instead of quoting one-offs forever.
Frequently asked questions
Is machining in China actually cheaper?
Often — but not by as much as people assume, and not always. On mature, tightly-toleranced parts, the gap has narrowed. Where China still wins clearly is speed and density: you can have a prototype revised and re-cut in days, because the whole ecosystem — material suppliers, tooling, finishing, even a replacement for a scrapped part — sits within a few kilometers in Shenzhen. That’s a supply-chain argument, not just a price argument.
What’s a realistic lead time for CNC parts from China?
For a small-batch prototype order, plan on 1–2 weeks of shop time for machining plus finishing, then shipping. Ask the supplier to split the estimate into machining, finishing, and inspection days. Anything under five working days total for a machined, finished, inspected part should make you ask how — not celebrate.
How do I verify a Chinese CNC shop is real?
Three cheap checks: ask for the exact machine list (models, axis counts, travel), request a live video call in the workshop, and ask for material certificates on your first order. Optionally use a third-party inspection company (SGS and similar all operate in Shenzhen) on the first batch. Real shops pass these checks in minutes; middlemen stall.
Should I send a 2D drawing or a 3D model?
Both. The STEP file is what the shop programs from; the PDF with dimensions and tolerances is what they check against. Sending only one of them forces guesswork somewhere — usually on the drawing callouts, which is exactly where tolerance disputes start.
What tolerances can Chinese CNC shops actually hold?
Be suspicious of blanket claims. An honest range: ±0.05 mm on standard features without extra cost, and ±0.01 mm where truly needed — with longer setup, more inspection, and a higher price to match. The right question isn’t “can you hold ±0.01” but “which of my features actually need it.”
What’s the difference between a factory and a trading company?
A trading company owns no machines — it forwards your order to a factory and adds a margin. That’s not automatically bad, but it adds a layer of communication, inspection, and accountability between you and the work. The direct way to tell: ask for a video call inside the workshop, and ask which specific machine will run your job.
Do I need to speak Chinese to work with a Chinese factory?
No — but you need a supplier who speaks engineering English, not just sales English. Every technical term matters: tolerance, surface finish, material grade, GD&T callouts. If your questions get answered with “no problem” instead of a technical explanation, the language barrier is the least of your problems.
Who we are (and what we built)
We’re Xinyongheng Precision — Eternal CNC (eternalcnc.com) — a machine shop in Shenzhen. Thirty CNC machines, including true 5-axis capability, running small-batch and prototype work on purpose, not squeezed between production lines.
We believe the problems above are systemic, not personal. So we built against them:
Quoting in two hours, not two days. Our estimating pipeline reads your drawing (STEP or PDF), identifies the material, generates a 3D view, and produces a multi-sheet quote — with an engineer in the loop for judgment calls. Fast quoting isn’t a marketing line for us; it’s the visible output of how the shop is run.
An AI agent answers the front door, 24/7. Material questions, process questions, RFQs — it answers in your language and escalates to a human the moment a question needs judgment. You’re not waiting for Shenzhen business hours to ask whether we can hold ±0.02 on that bore.
We’re not the cheapest shop in China, and we’re not the biggest. We’re the shop for engineers who are tired of “no problem” from people who have never cut metal.
The next step
Send us your drawing through the quote form on eternalcnc.com. You’ll have a real quote — with a real lead-time breakdown — within two hours.
And if we can’t make your part well, we’ll tell you that too, and point you to someone who can. Worst case, you learn something about your own part’s manufacturability. Best case, you’ve found the vendor relationship you were hoping for.