AQL Sampling Explained: How Inspectors Decide Pass or Fail

When an inspector checks 200 units out of a 5,000-unit order and reports "PASSED," there's real statistics behind it — not a vibe. That system is AQL sampling, and understanding it will make you better at writing POs, reading inspection reports, and arguing with factories.

What AQL Actually Means

AQL = Acceptance Quality Limit (定义于 ANSI/ASQ Z1.4 与 ISO 2859-1 标准). It is the worst tolerable process average of a production run — roughly, the defect level at which a lot has a high probability of being accepted.

The most important misunderstanding to clear up: AQL 2.5 does not mean "2.5% of units may be defective." The relationship between the AQL number and the actual accepted defect percentage is statistical, not literal. AQL 2.5 with a 200-unit sample allows up to 10 major defects — which is 5% of the sample. The AQL number is a dial on a statistical machine, not a percentage promise.

The Two Tables (Simplified)

Table 1: Lot size + inspection level → sample size.

For general consumer goods, everyone uses General Inspection Level II (GII):

Table 2: Sample size + AQL → accept/reject numbers.

For a 200-unit sample (lot of 3,201–10,000):

So: 10 or fewer major defects = PASS. 11 majors = FAIL. 14 or fewer minors = PASS. And any single critical defect anywhere in the sample fails the entire lot.

Defect Classification: Where the Real Negotiation Happens

The tables are mechanical. The power is in how you classify defects, and that's defined by your spec sheet, not by the inspector:

• Critical: unsafe for the user, or makes the product illegal in your market (sharp edges on a toy, missing required warning label, electrical fault). AQL 0.
• Major: product won't sell or will be returned — doesn't function, visible damage, wrong color, wrong size beyond tolerance. AQL 2.5.
• Minor: cosmetic imperfection a normal customer wouldn't notice. AQL 4.0.

The same defect can be major or minor depending on your product. A 1mm scratch is minor on an industrial bracket and major on a phone case sold as "premium." Write your classification list into the inspection instruction — never leave it to the inspector's judgment, and never leave it to the factory's.

Choosing Your AQL Levels

• AQL 2.5 / 4.0 (major/minor): the consumer-goods default. Start here.
• AQL 1.5 / 2.5: higher-value products, gift-grade goods, products where returns are expensive.
• AQL 1.0 / 1.5: medical, baby, safety-critical — and honestly, at this standard you should also be doing tighter process control and possibly full inspection of critical functions.
• AQL 4.0 / 6.5: promotional giveaways, price-driven goods where unit cost matters more than consistency.

Worked Example

Order: 8,000 bamboo cutting boards, FBA seller, spec sheet defines 12 major defect types and 9 minor types. AQL 2.5/4.0, GII.

• Sample: 200 boards
• Inspector finds: 6 major (2 cracked, 3 splintered edge, 1 wrong logo color), 11 minor (surface marks)
• Result: PASS (6 ≤ 10 majors, 11 ≤ 14 minors)

Two weeks later the same buyer gets Amazon complaints at a 3% return rate and asks how a "passed" inspection let this happen. Answer: AQL 2.5 accepts lots with defect rates around 2.5–5% of process average. If your business model needs sub-1% defects, you need a tighter AQL and you need to pay the factory for it — tighter AQL means the factory scraps more units, which means higher unit price. Quality is bought, not demanded.

What "FAIL" Triggers

A failed PSI is not the end — it's the start of a structured process:

1. Inspector documents defects with photos and locations in the sample
2. You (or your agent) request the factory's rework or 100% screening proposal — the factory re-inspects the full lot, removes defectives, replaces or reworks them
3. Re-inspection of the reworked lot (new man-day, new fee — usually paid by the factory as the responsible party)
4. Balance payment only after the re-inspection passes

This is why payment structure matters: see inspection types explained for tying final payment to a passed PSI.

FAQ

Why not just inspect 100% of units? Cost and reliability. 100% inspection of a 10,000-unit order takes 5–8 man-days, and human attention degrades badly after hours of repetitive checking — studies show full inspection misses 10–20% of defects. Statistical sampling beats tired humans.

Can I use a different inspection level to save money? Level I cuts the sample roughly in half (e.g., 125 → 80 units) for borderline cases; Level III doubles it for disputes or re-inspections. Special levels S-1 to S-4 exist for expensive/destructive tests.

My factory says their internal standard is AQL 2.5 — good enough? That's their incoming standard, applied by people they pay. It's a useful signal but not a substitute for independent verification. See Quality Control in China for the full framework.

Where do I get the actual tables? The standard is ANSI/ASQ Z1.4 (paid) but the tables are widely reproduced. Better: your inspection firm's booking form generates accept/reject numbers automatically from lot size + AQL — you just need to know what numbers to plug in, which is exactly what this article gave you.

给编辑的备注

• 配图建议：两张简化表格（本文的表可复用），AQL 流程图（抽样→发现缺陷→对照 Ac/Re 判定）
• 本篇是该集群最"工具型"的文章：适合做可下载的 AQL 快速参考卡（PDF）换取邮箱
• 数据核对：Ac/Re 数字基于 ANSI Z1.4 主表，发布前用任一验货公司的在线 AQL 计算器复核一遍
• 回链：结尾 CTA "We set AQL and defect lists for every order we manage — talk to us"