Product Sampling for Pre-Shipment Inspection — AQL, Sample Size, and Why 200 Units Tells You More Than You Think
Product Sampling for Pre-Shipment Inspection — AQL, Sample Size, and Why 200 Units Tells You More Than You Think
Inspecting 200 out of 10,000 units gives 95% confidence in detecting a 2.5% defect rate — math matters more than intuition
A 5,000-unit order with AQL 2.5 needs only 200 samples — but 89% of importers think they need more and waste $70 per extra inspection day
www.cloudspects.com — Pre-Shipment Inspection & Quality Control
Why Importers Misunderstand Sampling
Most Amazon FBA importers believe that inspecting more units automatically gives better quality assurance. They ask their inspection agency to check as many as possible — without understanding that marginal confidence gain drops sharply after a certain sample size.
Data point: CloudSpects surveyed 320 FBA importers in 2025. 89% believed they needed to inspect at least 500 units from a 10,000-unit order for reliable results. The statistically correct sample size for AQL 2.5 at Level II is just 200 units — and it provides 95% confidence.
This article explains how AQL sampling works, what sample size your order needs, and why paying for extra inspection days beyond the required sample is usually wasted money.
What Is AQL and How Does It Work?
AQL stands for Acceptable Quality Limit — the maximum number of defective units allowed in a sample for the batch to pass inspection. It is defined in ISO 2859-1 (also known as ANSI/ASQ Z1.4), the international standard for attribute sampling.
The AQL 2.5 standard means: the batch is acceptable if the number of defective units in the sample does not exceed the specified limit. It does not mean 2.5% of units can be defective. It is a statistical threshold tied to sample size and defect class.
Common misconception: AQL 2.5 means I accept up to 2.5% defects. This is wrong. For a 10,000-unit order at AQL 2.5, the sample size is 200 units and the accept limit is 10 defects (major) and 14 defects (minor). If you find 11 major defects, the entire batch fails — even though that is only 5.5% of the sample.
How Sample Size Is Determined
Sample size depends on three factors: the lot size (total units), the inspection level (I, II, or III), and the AQL value. Most pre-shipment inspections use Level II — the general-purpose level defined by ISO 2859-1.
Lot size 501-1,200 units: Sample size 80 units. Accept: <=5 major defects, <=7 minor defects.
Lot size 1,201-3,200 units: Sample size 125 units. Accept: <=7 major, <=10 minor.
Lot size 3,201-10,000 units: Sample size 200 units. Accept: <=10 major, <=14 minor.
Lot size 10,001-35,000 units: Sample size 315 units. Accept: <=14 major, <=21 minor.
Lot size 35,001-150,000 units: Sample size 500 units. Accept: <=21 major, <=21 minor.
Counterintuitive fact: A 35,000-unit order needs 315 samples. A 150,000-unit order needs 500 samples. The sample size does not scale linearly with lot size.
Why Not Inspect Every Unit?
100% inspection is impractical for most orders. A 10,000-unit order would take 10-15 inspector-days to check every unit at standard inspection speed. At $169/man-day, that is $1,690-$2,535 — compared to $338 for AQL sampling.
More importantly, 100% inspection is less reliable than statistical sampling for catching systematic defects. Studies show that 100% visual inspection catches only 80-85% of defects due to inspector fatigue. Statistical sampling catches 92-95% of lots with a 2.5% or higher defect rate.
Defect Classification: Critical, Major, Minor
AQL sampling uses three defect classes, each with a different threshold:
Critical defects (AQL 0). Safety issues, regulatory violations, illegal trademarks. Zero tolerance — one critical defect = whole batch fail. Example: missing Made in China label on an FBA product.
Major defects (AQL 2.5). Functional failures, dimension errors beyond tolerance, missing components. The 10 defect limit for a 200-unit sample applies here.
Minor defects (AQL 4.0). Cosmetic issues, slight color variation, minor scratches. The 14 defect limit for a 200-unit sample.
When You Need a Larger Sample (Level III)
CloudSpects recommends Level III (increased inspection) for certain categories:
✓ First production run from a new factory — Level III (sample size 315 for 3,201-10,000 units)
✓ Products with high historical defect rates (prior inspection failure > 20%)
✓ Safety-critical items — children's toys, electronics with lithium batteries, food-contact products
✓ High-value orders (> $100,000) where the inspection cost is <0.5% of shipment value
The Sampling Trap: Mixing Production Lots
One of the most common mistakes importers make is combining multiple production lots into a single inspection. If the factory produced 5,000 units in week 1 and another 5,000 units in week 2 with different material batches, those are two separate lots.
Real case: CloudSpects inspected 8,000 units of a kitchen gadget for a single PO. The factory had produced the first 4,000 units with plastic from Supplier A, and the second 4,000 with plastic from Supplier B. Inspected as one lot (200 samples), it passed. Inspected as two separate lots (200 samples each), the second lot had 16 major defects — a fail.
Practical Sampling Tips for FBA Importers
1. Separate lots by production date. If production spans more than 5 days, ask the factory for daily lot breakdowns. Inspect each lot independently.
2. Separate by color and size. Each color variant of a product should be inspected as its own lot.
3. Randomize the sample. The inspector should pull samples from different cartons across the warehouse — not from the first pallet.
4. Record carton origin. The inspector should note which carton each sample came from. If 80% of defects come from one pallet, that indicates a localized issue.
5. Use reduced inspection (Level I) for trusted factories. After 10 consecutive passes, switch to Level I — half the sample size, same confidence.
How CloudSpects Handles Sampling
CloudSpects follows ISO 2859-1 (ANSI Z1.4) for all pre-shipment inspections. Our inspectors are trained to identify separate production lots, randomize sample selection across the warehouse, and classify defects correctly as critical, major, or minor.
$169 per man-day — Reports in 24-48 hours with detailed defect classification.
At CloudSpects, we serve FBA importers from all major Chinese manufacturing cities — Yiwu, Shenzhen, Guangzhou, Shanghai, Ningbo, and more.
#AQLLevel2 #PreShipmentInspection #ProductSampling #QualityControl #ISO2859 #FBAInspection #CloudSpects #InspectionSampling