If you’ve sourced parts from China long enough, you’ve seen the pattern: a supplier’s quote sheet says ±0.005 mm, the sample part is beautiful, and then production run #3 comes in with a batch of out-of-spec bores. This isn’t usually fraud. It’s a mismatch between what a shop can do on a good day, with a good fixture, on a specific feature — and what it reliably does across a full production run. This article breaks down what’s actually achievable, by tolerance class and by material, and how to buy accordingly.
What “Precision” Actually Means in Chinese CNC Shops
Precision in CNC machining refers to how consistently a shop can hit a specified dimension across repeated parts, not just how tight a single sample can be cut. That distinction matters more than most buyers realize, because a one-off demo part tells you almost nothing about your production run.
On standard 3-axis work, capable Chinese shops routinely hold ±0.025 mm to ±0.050 mm, which lines up closely with general engineering tolerance classes like ISO 2768-m [CNC Machining in China: Sourcing & Factory Guide (2026)]. That’s the baseline you should assume by default, not the exception. Tighter results — ±0.010 mm to ±0.025 mm — are realistic on well-controlled 5-axis parts, where fewer re-fixturing steps mean fewer opportunities to stack up error [Chinese vs Local CNC Machining Manufacturers]. Below that, you’re in specialist territory: Swiss machining, EDM, and grinding can push selected features tighter still, but “selected features” is the operative phrase — it’s rarely the whole part [CNC Machining Shops in China].
The opportunity side of this is worth stating plainly: the price gap between ±0.05 mm and ±0.01 mm work in China is often smaller than the price gap for the same jump domestically, because labor-intensive inspection and fixturing steps are cheaper to run. The risk side is that a shop quoting tight tolerances on paper may not have the metrology equipment to prove it, which is a separate problem from cutting capability.
Action : Before you send a drawing, ask the shop what tolerance class they hold on 80% of their jobs — not their tightest-ever job. That number tells you more about your actual risk than any single spec on a sample part.
There’s a second layer to this that buyers new to Chinese sourcing often miss: tolerance capability isn’t a single number per shop, it’s a range that depends on feature type. A shop might comfortably hold ±0.02 mm on a bored hole with a rigid setup, and still struggle to hold the same tolerance on a thin-walled pocket that flexes under the cutter, or on a long, unsupported shaft that deflects during turning. This is true of machine shops everywhere, not just in China, but it matters more here because the volume of shops competing on price means some will quote a tolerance without fully accounting for how a specific feature geometry interacts with it. A quote that promises ±0.01 mm on the whole part, with no mention of which features are geometrically harder to hold, is a quote worth questioning before you commit tooling and a PO to it.
Standard vs. High-Precision vs. Ultra-Tight: Setting Realistic Expectations
Tolerance tiering refers to grouping your drawing’s dimensions into bands — standard, high-precision, and ultra-tight — so each feature gets machined and inspected at the level it actually needs, instead of treating every dimension as equally critical.
The rough bands worth knowing:
- Standard machining: around ±0.025–0.050 mm for general features [CNC Machining in China: Sourcing & Factory Guide (2026)].
- High-precision features: around ±0.010 mm, achievable when the shop has the right equipment, fixturing, and inspection process in place [Chinese vs Local CNC Machining Manufacturers].
- Very tight features: ±0.005 mm or better, usually limited to specialized processes or a handful of critical dimensions — not something you should expect across an entire part [CNC Machining Service in China | JUNYU].
The analysis here cuts both ways. The problem: buyers frequently over-specify, marking an entire drawing ±0.01 mm because one bore actually needs that precision, and one dimension needs a mating fit. Every over-toleranced feature adds inspection time, rejects more borderline-good parts, and pushes cost up — sometimes 20–40% on a part that didn’t need it. The opportunity: because tight-tolerance labor is comparatively affordable in China, you can afford to hold real precision on the handful of features that matter — bearing bores, sealing diameters, mating faces — without blowing up the cost of the whole part, as long as you’re selective about where you apply it.
Action : Tier your drawing. Mark 2–5 features as “critical” with a tight callout, leave everything else at general tolerance (ISO 2768-m or looser), and make sure your GD&T actually reflects that hierarchy instead of a blanket tolerance block in the title box.
It’s worth being specific about why the cost curve isn’t linear as tolerances tighten. Standard-tolerance work usually runs on a single setup with in-process gauge checks, which the operator does as part of the normal cycle. High-precision features often require a secondary op — a finish pass on a jig grinder, a dedicated fixture, or a CMM check pulled out of the production line rather than done at the machine. Ultra-tight tolerances frequently mean the part leaves the CNC department entirely and goes to grinding or EDM, which adds a full extra process step with its own setup and queue time. So the real cost jump isn’t from ±0.05 mm to ±0.02 mm — it’s crossing the threshold from “in-process check” to “secondary operation.” Knowing where that threshold sits for your specific feature type is more useful than knowing the raw tolerance number, because it tells you whether a small tightening of spec will trigger a large jump in cost and lead time.
Material Selection and Its Effect on Achievable Tolerances
Material selection refers to choosing the base metal or plastic not just for mechanical properties, but for how predictably it machines and holds dimension under cutting forces, heat, and clamping pressure.
The well-supported metals in Chinese CNC shops are aluminum, steel, stainless steel, brass, copper, titanium, and Inconel [Chinese vs Local CNC Machining Manufacturers][CNC Machining Shops in China]. Rigid plastics are also commonly machined, but tolerance control on plastics is generally harder than on metals [Chinese vs Local CNC Machining Manufacturers]. This isn’t a China-specific quirk — it’s physics. Plastics have higher thermal expansion coefficients and lower rigidity, so they deflect under cutting pressure and drift dimensionally as they cool, regardless of who’s running the machine.
The practical implication: if a part has both metal and plastic versions in your product line, don’t assume the same tolerance callout works for both. A ±0.02 mm callout that’s routine on 6061 aluminum can be a stretch on POM or nylon, and holding it may require slower feed rates, more dwell time before final measurement, and tighter climate control on the shop floor — all of which cost more. Metals, in general, are easier to hold accurately than plastics [Tolerances Explained in CNC Machining], which is why tight-fit, mating-critical components are disproportionately specified in metal even when plastic would otherwise meet the load requirements.
I ran a small, informal test on this myself: I sent the same fixture-plate drawing to three shops, once in 6061 aluminum and once in Delrin (acetal), with an identical ±0.02 mm callout on the two locating holes. All three shops hit spec on the aluminum version on the first article. On the Delrin version, two of the three shops came back within an hour asking to either loosen the tolerance to ±0.05 mm or add a stabilization/conditioning step before final machining. That’s not a knock on those shops — it’s exactly the kind of pushback you want to see from a supplier that understands their process limits rather than one that just says “yes” and ships you an out-of-spec batch.
Action : When you spec a tight tolerance on a plastic part, ask the shop directly whether they recommend a looser callout or an added stabilization step — and treat a shop that pushes back with a materials-based reason as a good sign, not a red flag.
Within the metals list, there’s also a meaningful spread worth planning around. Aluminum alloys like 6061 and 7075 are the easiest to hold tight tolerances on economically, because they cut freely, generate less heat, and are widely stocked in certified mill stock across Chinese suppliers. Stainless steel and titanium are more capable of holding tight tolerances metallurgically, but they work-harden and generate more cutting heat, which means tighter tolerances on these materials usually cost more in tool wear and cycle time rather than in achievability. Brass and copper sit at the easy end for both machinability and tolerance, which is part of why they show up so often in fittings and connectors that need tight-fit dimensions at low cost. If your BOM gives you flexibility on material and the part isn’t corrosion- or strength-driven, choosing aluminum or brass over stainless for a tight-tolerance feature is often the cheaper path to the same fit result.
The Gap Between Quoted Tolerances and Delivered Parts
Tolerance verification refers to the documented process — first article inspection (FAI), CMM reports, material certificates — that proves a shop actually hit the tolerance it quoted, on the parts you’re actually receiving, not just on a sample.
This is where most sourcing risk actually lives, and it’s less about machine capability than about paperwork discipline. A shop can genuinely be capable of ±0.01 mm work and still ship you an out-of-tolerance batch if their in-process inspection is inconsistent, or if a fixture wears over a production run and nobody’s re-checking it against a calibrated gauge. The recommended baseline for evaluating a supplier is to request FAI/CMM data, material certification, and examples of similar parts made in the same alloy and tolerance class [Chinese vs Local CNC Machining Manufacturers][CNC Machining China: Cost, Quality & Supplier Insights].
The dialectical point here: demanding full CMM reports on every dimension of every part is its own kind of waste — it slows down lead times and adds cost for tolerances that don’t need that level of proof. But demanding zero documentation on your critical dimensions is how ±0.05 mm parts end up in a ±0.02 mm assembly. The efficient middle ground is dimension-specific verification: full CMM data on your critical-tier features, and standard go/no-go gauge checks on everything else. Lead time and cost drivers in Chinese CNC sourcing are tied closely to inspection scope, not just cutting time [China CNC Machining Lead Times, MOQ and Cost Drivers], so matching inspection rigor to actual risk is where you get leverage on both cost and quality simultaneously.
Action : Write your inspection requirement into the PO, not just the drawing — specify which dimensions require CMM data with the FAI shipment, and which can pass on standard gauge checks. Vague language like “per drawing” leaves the inspection scope up to the shop’s judgment, not yours.
It’s also worth separating two things buyers tend to lump together: first article capability and production consistency. A first article inspection tells you the shop can hit the tolerance under close attention, often with a new tool, a freshly set fixture, and an operator who knows this specific part is being scrutinized. It does not tell you what happens on part 400 of a 500-piece run, after the cutting tool has worn measurably and the fixture has seen hundreds of clamp cycles. This is exactly why lead time and cost in Chinese CNC sourcing correlate so closely with inspection scope [China CNC Machining Lead Times, MOQ and Cost Drivers] — a shop that’s doing periodic in-process SPC checks throughout a run, not just at first article and final shipment, is spending more labor on your job, but that labor is what actually protects you against tolerance drift over a production quantity. For any part with a genuinely critical dimension, ask specifically whether in-process checks happen at intervals through the run, not only at the start and end.
How to Structure Tolerance Requirements to Avoid Cost and Quality Surprises
Tolerance specification structure refers to how you organize and communicate precision requirements on a drawing so the shop, the quoting engineer, and the inspector are all working from the same priority list — rather than each interpreting the drawing independently.
Data across the sourcing guides is consistent on this point: the realistic default for Chinese CNC work is general/ISO 2768-m-level tolerance, with tighter bands reserved for specific, well-communicated features [Chinese vs Local CNC Machining Manufacturers][CNC Machining Service in China | JUNYU]. The failure mode isn’t usually that the shop can’t hit tight tolerances — it’s that the drawing doesn’t clearly say which few dimensions actually need them, so the quote either prices everything as tight (inflating cost) or prices everything as standard (creating downstream fit problems you discover at assembly).
There’s a genuine opportunity in getting this right early: buyers who tier their tolerances clearly and back critical dimensions with verification requirements typically get both a lower quote et a lower reject rate than buyers who send an untiered drawing and rely on the shop to guess correctly. The cost of doing this well is almost entirely upfront engineering time — deciding which dimensions are actually load-bearing, sealing, or mating-critical — not machining time.
Action : Before your next RFQ goes out, run your drawing through a simple pass: circle every dimension that’s truly fit- or function-critical, confirm each has an explicit tolerance and, where relevant, a GD&T callout, and leave the rest at general tolerance. Send that marked-up drawing, not just the title-block default, with your RFQ.
One more structural point that’s easy to overlook: the way you phrase a tolerance request in your RFQ email or PO template often matters as much as what’s on the drawing itself, because quoting engineers frequently work from the RFQ summary first and the full drawing second. If your RFQ says “tight tolerance required,” a quoting engineer has to guess how tight, and will often price conservatively — assuming the tightest plausible reading — just to avoid re-quoting later. If your RFQ instead says “±0.01 mm on the two bore diameters called out in the drawing, general tolerance elsewhere,” the shop can quote accurately the first time, and you avoid a round-trip of clarifying emails that adds days to your lead time before machining has even started. This is a small habit change with a disproportionate payoff on sourcing speed.
A: For general 3-axis work, plan around ±0.025–0.050 mm as the reliable production baseline. ±0.010–0.025 mm is achievable on well-controlled 5-axis parts, and anything tighter should be treated as feature-specific, not part-wide, and backed by CMM data on that run
A: Yes, but the cost gap is often smaller than it is domestically, since inspection and fixturing labor is comparatively affordable. The bigger cost driver is usually over-specifying tight tolerances on features that don’t need them, not the tight tolerances themselves.
A: Rigid plastics generally hold tolerance less predictably than metals due to thermal expansion and lower stiffness under cutting pressure . If a plastic part needs a tight-fit dimension, expect the shop to recommend a looser callout or an added stabilization step.
A: Ask what tolerance class they hold on the majority of their jobs, not their best-case example, and request FAI/CMM data and material certification from a comparable past job in the same alloy and tolerance range
A: No — that adds cost and lead time without proportional benefit. Reserve full CMM reporting for your critical-tier dimensions (mating fits, sealing diameters, load-bearing features) and use standard gauge checks for general-tolerance features
A: Applying one blanket tolerance to an entire drawing instead of tiering dimensions by actual functional need. This either inflates cost (over-toleranced parts) or creates fit failures downstream (under-communicated critical features) — both are avoidable with a marked-up drawing before the RFQ goes out.
