The Question Behind the Tolerance: Choosing Inspection Methods for Chinese CNC Parts
Photo by Photo by Malcolm Choong 鍾声耀 on Unsplash on Unsplash
Most purchase orders sent to Chinese CNC suppliers carry a drawing, a tolerance block and a request for a QC report. That request is doing less work than many buyers assume. A tolerance defines an acceptable limit. It does not define the question being asked, the instrument used to answer it, the datum alignment behind it, or the uncertainty attached to the result. When the part arrives and the report looks tidy, the buyer may still have no clear view of whether the critical features were measured in a way that reflects how the part will function.
The practical remedy is not a longer report. It is a shorter, sharper specification: an inspection method schedule that maps each significant feature to a measurement approach, and maps that approach to the risk the feature carries.
Risk first, instrument second
Inspection methods are often chosen by habit. A supplier may default to a CMM because it looks rigorous, or to callipers because they are quick. Neither choice is inherently right. The starting point should be the consequence of a feature failing.
A useful split for UK buyers is fourfold. First, safety and regulatory features: those where failure could cause injury, breach a standard or trigger a recall. Second, fit and assembly features: interfaces, bores, threads, profiles and mounting points that determine whether the part goes together. Third, functional features: surfaces, clearances and geometries that affect performance but may not stop assembly. Fourth, process-control and cosmetic features: dimensions that matter for consistency, tool wear or appearance rather than immediate function.
That hierarchy should drive the inspection method. A safety-critical bore may justify a CMM with a documented datum scheme, a calibrated probe and a traceable report. A high-volume assembly hole may be better served by a hard go/no-go gauge that answers one question quickly and repeatedly. A cosmetic edge break may need a visual standard and a profile sample, not a dimensional report at all.
What each method can and cannot answer
A CMM is strongest when the question is geometric: position, profile, true position, datum relationships and complex form. It is weakest when the drawing is ambiguous, the datum scheme is not agreed, or the feature is better judged by function. A CMM report that lists numbers without stating alignment, probe qualification, temperature and software version is a partial answer, not a complete one.
Hard gauges answer attribute questions. A plug gauge tells you whether a hole is within limits; it does not tell you where in the tolerance band the hole sits. That is often enough for assembly risk, and it is usually faster and more repeatable on the shop floor than a CMM. The trade-off is that gauge design, wear and calibration must be controlled. A worn go-gauge can pass parts that should fail.
Functional checks answer the question the drawing only implies. If a bracket must locate against a mating part, a first-article fixture that simulates the interface can reveal stack-up problems that a set of individual dimensions will miss. Functional checking is not a substitute for dimensional verification, but it is often the better acceptance test for interfaces.
First-article fixtures sit between measurement and function. They are built once, validated against the drawing, and then used to confirm that the datum scheme and assembly relationships are understood. For a new Chinese supplier, a first-article fixture can be more informative than a long CMM report because it tests the supplier's interpretation of the drawing, not only its machine accuracy.
The vague QC report is a specification failure
When a UK buyer asks for a QC report without defining its content, the supplier will provide what is convenient. That may be a dimensional sheet with no feature identifiers, a CMM printout with no datum note, or a certificate that states conformity without evidence. The problem is not necessarily dishonesty. It is that the requirement was never specified.
A better approach is to attach an inspection method schedule to the purchase order. For each significant feature, state the method, the instrument type, the calibration requirement, the sample size, the acceptance rule and the report format. For CMM work, require the datum alignment and the feature identifiers to match the drawing. For gauge work, require the gauge type and calibration status. For functional checks, define the fixture and the pass/fail condition.
This schedule does not need to cover every dimension. It should cover the features where a wrong answer carries cost. The remaining features can be handled by the supplier's normal inspection, provided that the buyer has agreed what normal means.
Do not buy precision the process cannot hold
There is a commercial trap in over-specifying inspection. A buyer may demand a CMM report with micron-level uncertainty for a feature whose process capability is far wider than the tolerance. The measurement looks impressive, but it does not improve the part. It may simply generate debate about measurement uncertainty when the real issue is machine capability, tool wear or fixture rigidity.
Before raising inspection precision, ask what the process can hold. A tolerance that is tight relative to the supplier's demonstrated capability will need process controls, not just more measurement. Gauge repeatability and reproducibility should be considered for hard gauges, especially where operators judge a fit by feel. For CMM reports, the uncertainty of the measurement should be sensible against the tolerance, but it should not be treated as a substitute for capability.
A practical closing loop
For UK buyers sourcing CNC parts from China in 2026, the most useful discipline is to separate three questions: what the drawing requires, how the requirement will be verified, and what evidence will be accepted. The inspection method schedule answers the second and third. It also gives the supplier a fair chance to quote the inspection cost, rather than absorbing it invisibly or skipping it under pressure.
A report is not a measurement, and a gauge is not the truth. Both are answers to a question. The buyer's job is to make sure the question is worth asking, and that the answer is strong enough for the risk attached to the feature.
