It is an international standard (ISO 2859-1)<\/a>, also published as ANSI\/ASQ Z1.4 in the US) that tells you exactly how many pieces to pull from a batch, and exactly how many defects are acceptable. It removes the guesswork. It gives you a number you can argue with. And it is the difference between “I hope the goods are okay” and “I statistically know the goods are within an acceptable quality range.”<\/p>\n\n\n\n Last month, a buyer from Germany asked us: “I have 2,000 power banks. How many should you inspect?” We told him 125. He said “that seems like a lot, can you do 50?” We explained why 50 is statistically meaningless for a batch of 2,000 \u2014 and why the AQL table says 125. This article is that explanation, in full.<\/p>\n\n\n\n AQL stands for Acceptable Quality Limit. It is not<\/strong> “Average Quality Level” \u2014 a common mistake. It is the maximum percentage of defects that you, the buyer, are willing to tolerate in a batch. If you set AQL at 2.5% for major defects, you are saying “I will accept a batch where up to 2.5% of the products have major defects, but no more.”<\/p>\n\n\n\n Here is the part that surprises most buyers: AQL is not a guarantee of zero defects. It is a statistical tool that helps you decide whether a batch is likely to be within your tolerance. A batch that passes AQL inspection can still have defects \u2014 just not more than your agreed limit. A batch that fails AQL inspection has more defects than you agreed to accept, and you have the right to reject it, demand rework, or negotiate a discount.<\/p>\n\n\n\n The foundation of AQL sampling in China is simple: you cannot inspect every piece in a 10,000-piece order (it would take too long and cost too much), so you inspect a statistically representative sample. If the sample passes, you accept the batch. If it fails, you reject it. The math behind this is decades old and used by every major retailer, brand, and government procurement agency in the world.<\/p>\n\n\n\n This is the core principle behind AQL sampling in China \u2014 statistical confidence, not guesswork.<\/p>\n\n\n\n AQL sampling in China uses three inputs to determine your sample size and acceptance limits:<\/p>\n\n\n\n This is the total number of pieces in your batch. 500 pieces, 2,000 pieces, 10,000 pieces \u2014 the lot size determines which row of the AQL table you use.<\/p>\n\n\n\n This determines how strict your sampling is. There are three general levels (I, II, III) and three special levels (S1-S4). For most consumer products, we use General Inspection Level II<\/strong> \u2014 it is the industry standard, balancing statistical confidence with inspection cost. Level I uses fewer samples (cheaper, less confidence), Level III uses more (more expensive, more confidence). We almost never use Level I for a first order with a new factory.<\/p>\n\n\n\n This is the maximum defect rate you will accept, set separately for each defect category. The industry standard for consumer electronics is:<\/p>\n\n\n\n These three numbers \u2014 lot size, inspection level, AQL limits \u2014 are all you need to read the table.<\/p>\n\n\n\n This is how AQL sampling in China works in practice. We take your lot size, find the corresponding row on the table, and read off the sample size code letter. Then we use that code letter to find the sample size and the acceptance\/rejection numbers for each AQL limit.<\/p>\n\n\n\n Here are three real examples from orders we inspected this year:<\/p>\n\n\n\n Example 1: 500 Bluetooth speakers<\/strong><\/p>\n\n\n\n For 500 pieces, we only pull 20. If we find 1 major defect, the batch passes (accept 1). If we find 2 major defects, it fails (reject 2). This seems lenient, but statistically, 20 pieces is enough to catch a systemic problem in a 500-piece batch.<\/p>\n\n\n\n Example 2: 2,000 power banks <\/strong><\/p>\n\n\n\n This is the order the German buyer asked about. 125 pieces out of 2,000 is 6.25% \u2014 not a huge percentage, but statistically significant. If we find 5 major defects, it passes. If we find 6, it fails. The German buyer initially wanted 50 pieces, but 50 would give us an acceptance limit of 2 major defects \u2014 much stricter than the standard, and statistically less reliable because the sample is too small to be representative.<\/p>\n\n\n\n Example 3: 10,000 LED bulbs <\/strong><\/p>\n\n\n\n Notice that for 10,000 pieces, we still only pull 200 \u2014 the same as for 5,000 or 8,000. The AQL table caps the sample size at 200 for General Level II above a certain lot size. This is because beyond 200 pieces, the statistical gain from a larger sample is minimal. You do not need to inspect 10% of a 10,000-piece order \u2014 2% is enough.<\/p>\n\n\n\n AQL sampling in China does not use one limit for all defects. It separates defects into three categories, each with its own AQL limit, because not all defects are equal:<\/p>\n\n\n\n The key insight: a batch can pass on major defects but fail on minor defects, and vice versa. We have seen batches with zero major defects but 15 minor defects in a 200-piece sample \u2014 that batch failed, because the factory’s process control was clearly sloppy, even though nothing was broken.<\/p>\n\n\n\n AQL sampling in China gives you a clear pass or fail decision \u2014 but “fail” does not automatically mean “reject the entire order.” It means you have leverage. Here is what happens when a batch fails:<\/p>\n\n\n\n The factory must fix every defective piece, and then we come back for a re-inspection. The re-inspection uses a fresh random sample (not the same pieces). If the re-inspection passes, the batch ships. If it fails again, you move to Option 2 or 3. This is the most common outcome \u2014 most factories will rework rather than lose the order.<\/p>\n\n\n\n Instead of reworking and re-sampling, the factory (or a third party) inspects every single piece in the batch. All defective pieces are removed or repaired. This is more expensive but gives you certainty \u2014 every piece that ships has been individually checked. We recommend this for high-value products or when the defect rate is high (above 5%).<\/p>\n\n\n\n If the defects are minor and you need the goods urgently, you can accept the batch with a discount. The discount should reflect the defect rate \u2014 if 3% of the batch is defective, you negotiate at least a 3% discount, plus the cost of handling returns. We always recommend getting the discount in writing before the goods ship.<\/p>\n\n\n\n If the batch has critical defects, or fails re-inspection twice, or the factory refuses to cooperate, you can reject the entire batch and demand your deposit back. This is rare \u2014 most factories will fix the problem rather than lose a customer \u2014 but it is your ultimate leverage, and AQL gives you the statistical justification to do it.<\/p>\n\n\n\n AQL sampling in China gives you the proof to back up every reject, rework, and discount decision you make.<\/p>\n\n\n\n The biggest mistake in AQL sampling in China is letting the factory pick the sample. We have said this before, but it bears repeating: the factory will always pick the best pieces This is the same principle we follow during a factory audit in China \u2014 read our guide here<\/a><\/strong>. \u2014 the ones on top, the ones inspected twice, the ones specifically placed where the inspector will look. If you let the factory “help” you pick the sample, your AQL inspection is worthless. Always use a random number generator to pick carton numbers from different positions in the stack.<\/p>\n\n\n\n The second mistake is only checking visual defects. AQL covers all defect categories \u2014 visual, functional, packaging, labeling. If you only look at the outside of the product, you will miss dead batteries, broken circuits, and non-scannable barcodes. Every piece in the sample should go through a full functional test, not just a visual check.<\/p>\n\n\n\n The third mistake is ignoring minor defects because “they don’t affect function.” Minor defects are an early warning sign. If a factory can’t be bothered to fix small scratches and misaligned buttons, it probably can’t be bothered to fix bigger problems either. A high minor defect rate indicates poor process control, and poor process control leads to major defects in the next order. We always report minor defects separately, and we always flag a high minor rate even if the batch technically passes.<\/p>\n\n\n\n The fourth mistake is using the wrong AQL limits. Some buyers set AQL at 1.0 for major defects, thinking “stricter is better.” But AQL 1.0 means you will reject more batches \u2014 including batches that are actually fine \u2014 because the acceptance limit is so tight. For a 200-piece sample at AQL 1.0, you can only accept 3 major defects (reject at 4), compared to 7 at AQL 2.5. Unless you are making medical devices or children’s toys, AQL 2.5 for major defects is the industry standard and the right balance.<\/p>\n\n\n\nWhat AQL actually means (and what it doesn’t)<\/h2>\n\n\n\n
The three numbers you need to know<\/h2>\n\n\n\n
1. Lot size <\/strong><\/h3>\n\n\n\n
2. Inspection level<\/h3>\n\n\n\n
3. AQL limits <\/strong><\/h3>\n\n\n\n
\n
How we read the AQL table (with real examples)<\/h2>\n\n\n\n
![\"warehouse](\"https:\/\/hongfengsourcing.com\/wp-content\/uploads\/2026\/09\/sampling.jpg\")
\n
\n
\n
The three defect categories and their AQL limits<\/h2>\n\n\n\n
![\"inspector](\"https:\/\/hongfengsourcing.com\/wp-content\/uploads\/2026\/09\/report.jpg\")
\n
What happens when the sample fails<\/h2>\n\n\n\n
Option 1: Reject and rework <\/strong><\/h3>\n\n\n\n
Option 2: 100% inspection <\/strong><\/h3>\n\n\n\n
Option 3: Negotiate a discount <\/strong><\/h3>\n\n\n\n
Option 4: Reject and cancel<\/strong><\/h3>\n\n\n\n
Common mistakes buyers make with AQL<\/h2>\n\n\n\n
![\"quality](\"https:\/\/hongfengsourcing.com\/wp-content\/uploads\/2026\/09\/linecheck.jpg\")
More on quality inspection<\/h2>\n\n\n\n
![\"factory](\"https:\/\/hongfengsourcing.com\/wp-content\/uploads\/2026\/09\/fullinspection.jpg\")