Sampling levels used in machined parts inspection
Sampling levels translate lot size into sample size, and defect limits translate sample findings into a verdict. Together they set how much protection the buyer purchases and how much rejection risk the supplier carries. This guide explains the level system without jargon: what general and special levels do, how code letters and acceptance numbers work, and how machined-parts buyers match strictness to consequence trait by trait instead of guessing.
Key takeaways
- Use general inspection levels for the main lot verdict and reserve special levels for costly or destructive supplementary checks.
- Fix the level, sample size logic, and acceptance numbers in writing before the visit so the verdict follows a published rule.
- Understand acceptance sampling as the statistical frame that lets sample evidence support lot decisions. [1]
- Run sampling inside a disciplined acceptance event with random selection and recorded evidence. [2]
- Set strictness by consequence per trait family rather than defaulting every trait to the tightest setting.
Sampling levels are the dial buyers turn to set inspection strictness, yet most buyers turn it blindly. They copy a level from another product, accept whatever default the inspection company suggests, or demand the strictest setting on the theory that more checking always helps. Each of these habits wastes money or hides risk, sometimes both at once.
The underlying idea is straightforward. A sampling level maps the lot size to a sample size; the defect limit maps the sample findings to a verdict. Bigger samples and tighter limits both shift risk from the buyer to the supplier, and both cost more in inspection time and rejection frequency. Choosing well means matching that risk shift to the product's real consequences.
This article explains the level system in practical terms: general versus special levels, how code letters produce sample sizes, how acceptance numbers work, and how machined-parts buyers set strictness trait by trait.
General levels carry the verdict
General inspection levels are the workhorses of lot judgment. They scale the sample size with the lot size, so larger lots get larger samples, and they apply to the full range of visual and dimensional traits in the main check. Most machined-parts lots are inspected under a general level, with the middle setting as the default and tighter or looser neighbors chosen for cause.
Moving one step tighter meaningfully increases the sample size and therefore the inspection time, the cost, and the chance of catching marginal lots. Moving one step looser does the reverse. The choice is a risk purchase, and buyers should make it deliberately: what does an escaped defect cost against what does a rejected good lot cost. Neither extreme is virtuous on its own.
Machined parts inspection in China machined parts inspection in China typically runs the main sample under the agreed general level while the inspector draws cartons at random across the stacked lot. That pairing, sized sample plus random draw, is what lets a few dozen checked pieces speak for thousands.
Where the default middle setting fits
The default general level suits standard machined parts with ordinary consequences: industrial fittings, brackets, housings, and similar components where a defect costs rework rather than safety. It balances buyer protection against inspection economy for lots from a few hundred to tens of thousands of pieces.
Depart from the default for stated reasons. Safety-critical fasteners, fluid-path components, and high-value precision parts justify the tighter neighbor. Low-risk, low-value accessories with clean supplier history can justify the looser neighbor. Write the reason next to the choice so the next order inherits logic instead of habit.
Special levels serve special purposes
Special inspection levels use small samples for checks where each measurement is expensive, slow, or destroys the piece. Destructive testing, lengthy functional runs, and laboratory-bound measurements are the classic cases. The small sample keeps these checks affordable while the main verdict still rests on the general-level sample. The misuse to avoid is running the whole verdict on a special level to save inspection fees. A small sample gives weak discrimination: good lots usually pass, but poor lots pass often enough to make the exercise theater. The sampling literature treats representativeness as foundational to everything that follows. [1]
From code letter to verdict
| Step | What happens | Buyer action |
|---|---|---|
| Choose the level | General for verdict, special for supplements | Fix the level per trait family in the order |
| Find the code letter | Lot size plus level maps to a letter | Confirm the letter covers the full lot quantity |
| Read the sample size | Letter maps to a fixed piece count | Check the count suits the inspection budget |
| Apply acceptance numbers | Defect limit maps findings to pass or fail | Pre-agree limits per severity class |
| Draw and judge | Random cartons, counted findings, stated verdict | Require random draw and photographed evidence |
Acceptance numbers without mystique
The acceptance number is simply the maximum defects the sample may contain while the lot still passes, with separate numbers per severity class. Critical defects normally carry a zero tolerance: one found, lot failed. Major and minor defects carry small allowances that tighten as the chosen limit tightens.
This structure explains why classification matters so much. Moving a trait from minor to major does not just rename it; it moves its findings into a stricter allowance that fails lots sooner. Severity classes and sampling levels are two dials on the same machine, and buyers should set them together rather than letting one drift.
Keep the mathematics honest across visits. Comparing defect rates between reports only works when the level and limits stayed constant, so resist the urge to tighten silently after a scare. Deliberate changes with recorded reasons build a comparable history; reactive tweaks build noise.
Run the sampled visit with discipline
Sampling rules assume a disciplined event: complete lot, random carton selection, fixed checking order, counted and photographed findings, and a verdict stated against pre-agreed numbers. The acceptance-event tradition describes this choreography of settled roles and documented evidence, and machined-parts visits follow it directly. [2]
The statistical frame behind the choreography is acceptance sampling, whose published theory explains how sample evidence supports decisions about the larger lot under defined risks. [1] Buyers need not derive the tables, but honoring the three conditions, random draw, sized sample, fixed numbers, is what separates a verdict from an opinion.
Sources
- Wikipedia: Acceptance sampling —— Overview of acceptance sampling including AQL concepts used in product inspection.
- Wikipedia: Factory acceptance test —— Overview of factory acceptance testing performed before shipment or installation.
Frequently asked questions
What is the difference between general and special inspection levels?
General levels set sample sizes that grow with lot size and suit overall lot judgment; special levels use smaller samples for checks that are costly, slow, or destructive. Use general levels for the main verdict and special levels for supplementary measurements, never as a cheap substitute for the main sample.
How does a sample-size code letter turn into a verdict?
The code letter maps the lot size and chosen level to a sample size, and the acceptance table maps the sample size and defect limit to pass or fail numbers. Buyers need not memorize the tables; they need to fix the level and limits in writing before the visit so the inspector applies a published rule.
How should buyers choose strictness without over-inspecting?
Match strictness to consequence: safety, regulatory, and high-value traits justify tighter limits and larger samples, while minor appearance traits on hidden faces suit standard settings. Document the reasoning per trait family so future orders inherit the logic instead of relitigating it.
Can buyers compare reports that used different sampling levels?
Keep the level and limits identical and compare trait-by-trait defect rates across visits. A stable process shows steady low counts; a drifting one shows rising counts in the same family. Change limits only deliberately, because silent tightening punishes good suppliers and silent loosening hides drift.
Read the briefing before everyone else.
One email each Wednesday. Field reporting from Accra, market data, and the cultural threads shaping Ghana — read by 41,800 subscribers across 89 countries.