FLECHAZO
Engineering & QC Peer-Reviewed by Factory QA Lab 1 of 20 (Units that fails a 100 piece order at AQL 2.5 under general inspection level II)

AQL 2.5 Explained: What Accept and Reject Actually Mean on a Plush Order

AQL 2.5 appears on almost every plush quotation and is widely misunderstood in the same way: it is not a target defect rate, and it is not the share of defects a shipment is allowed to contain. It is a sampling plan, and the numbers move with the lot size - a 100 piece order is judged on 20 units and fails on one defect, while a 10,000 piece order is judged on 200 and fails on eleven. Here is the table, the defect classes, and how to write the clause.

M

Marcus Vance

Director of Factory Operations & Sourcing Quality Control

8 min read

Executive Summary & Core Benchmarks

  • ✓ AQL is a sampling plan, not a defect allowance. At AQL 2.5 a lot with 2.5% defective units still has roughly a 95% chance of being accepted, which is a different statement from '2.5% of the shipment may be defective'.
  • ✓ The sample size scales with the lot, not with the order value. General level II draws 20 units from a 91 to 150 piece lot and 200 units from a 3,201 to 10,000 piece lot.
  • ✓ On a 100 piece order the accept number is 1: one major defect fails the lot. On a 10,000 piece order the same AQL accepts up to 10. The small buyer is held to a stricter standard than the large one, and rarely knows it.
  • ✓ Critical defects are normally zero-tolerance and sit outside the AQL table entirely - needle fragments, sharps and contamination are not a sampling question.
  • ✓ AQL catches attribute defects you can count. It does not catch systematic problems like a colour shift across the whole run, a fill weight that is uniformly light, or a fabric that was substituted.

Every plush quotation in this industry carries a line like “AQL 2.5” or “AQL 2.5 major / 4.0 minor”, and almost nobody on either side of the contract agrees on what it means. It is one of the few numbers in a quotation that changes what actually happens at the end of a production run, so it is worth being precise about.

What AQL is not

Three misreadings, all common:

It is not a defect allowance of 2.5%. AQL 2.5 means that a lot which is 2.5% defective has a roughly 95% probability of being accepted under that plan. It is a statement about the producer’s risk, not a licence to ship 2.5% bad units. A shipment that passes at 2.5% may contain materially more defects than 2.5%, particularly in small lots where the sample is small.

It is not a quality target. A factory building to “AQL 2.5” is not trying to produce 2.5% defective goods; it is being judged against a plan that tolerates a certain number of failures in a sample.

It is not one number for all defect types. In practice an order is judged against two or three separate standards: critical (usually zero), major, and minor. A single “AQL 2.5” line on a quotation is shorthand, and the full clause should say which class it applies to.

The mechanics

The plan comes from ISO 2859-1, and its American ancestor ANSI/ASQ Z1.4 (which itself descends from MIL-STD-105E). The logic is:

  1. The lot size selects a code letter.
  2. The code letter and the inspection level select a sample size.
  3. The sample size and the AQL select an accept number and a reject number.

Unless the contract says otherwise, the inspection level is general level II, which is the default in the standard.

Lot size (pieces)CodeSample size (level II)Accept / Reject at AQL 2.5
91 – 150F201 / 2
151 – 280G322 / 3
281 – 500H503 / 4
501 – 1,200J805 / 6
1,201 – 3,200K1257 / 8
3,201 – 10,000L20010 / 11
10,001 – 35,000M31514 / 15
35,001 – 150,000N50021 / 22

The large print in that table is the part buyers should read twice.

A 100 piece order is judged on 20 units, and one defect fails it. At the minimum order quantity this site works to, a single major defect in a random sample of 20 is a rejection. That is a genuinely strict standard, and it surprises buyers who assumed a small order would be inspected more loosely.

A 10,000 piece order is judged on 200 units and fails at 11. Same AQL, one hundred times the volume, and a far wider acceptance window in absolute terms.

The reason is arithmetic rather than policy: statistical sampling needs a larger sample to discriminate a given defect rate in a larger lot. But the practical consequence is that small buyers live under a tighter regime, and a factory that knows this will inspect more carefully on small runs rather than less. Ours does, because a failed 100 piece lot is a commercial problem out of all proportion to its size.

The three defect classes

AQL is applied to attribute defects - things you can count - grouped by severity. The classification is where a plush order is won or lost, because the same physical flaw can be major or minor depending on how the clause defines it.

ClassDefinitionPlush examples
CriticalUnsafe, illegal, or would cause harmNeedle fragment, sharp edge, broken hardware, contamination, banned substance, missing safety warning
MajorReduces function, safety margin, or saleabilitySeam opening, eye or nose failing pull or torque, wrong size beyond tolerance, wrong colourway, missing component, visible stain, illegible care label
MinorCosmetic, does not affect useUntrimmed thread, minor shade variation between panels, slightly uneven stuffing, care label sewn crooked

Critical defects are normally zero tolerance and sit outside the AQL table. A needle fragment is not a sampling question and no accept number applies to it. That is why needle detection is a 100% control rather than an inspection needle detection - every piece passes the aperture, and the AQL plan covers what a detector cannot see.

The most contested line is always between major and minor. A buyer who writes “any visible defect is major” gets a strict inspection and a higher unit price, because the factory has to absorb rework. A buyer who writes nothing gets the factory’s default, which is usually generous to the factory. The clause should name the defects, not gesture at them.

Why the sample can pass and the shipment still disappoint

AQL is an attribute sampling plan with a known limitation: it is blind to systematic problems that are uniform across the lot.

If the dyed fabric is a shade off, every unit is equally off, and a sample of 20 will show it - but it may be argued as within tolerance rather than counted as a defect. If the fill weight is 4% light across the whole run, every unit is light, and no number of sampled defects will surface it, because lightness is not a defect in the attribute sense. If the fabric was substituted for a cheaper equivalent, the sample looks fine and the hand-feel is wrong.

Those failure modes are caught by specification control, not by inspection: approved fabric and a signed swatch, a stated fill weight in grams with a tolerance, and a measurement taken on the sample rather than a look at it. That is what the specification section of a sampling brief exists to lock down before production starts.

In-line, pre-shipment and loading

Three inspections are commonly specified, and they catch different things:

  • In-line, at roughly 30% of the run. Catches a systematic fault while it is still cheap to correct. This is the highest-value inspection on a first order and the one most often skipped.
  • Pre-shipment, on finished packed goods. Applies the AQL plan; this is the inspection the accept number refers to.
  • During loading / container check, on carton count, marks and condition. Not a quality inspection, but it catches carton damage and a short shipment.

A first order with only a pre-shipment inspection is a first order that finds its problems when it is too late to fix them cheaply.

How to write the clause

A clause that can actually be enforced reads roughly:

Critical defects: zero accepted. Major defects: AQL 2.5, general inspection level II, per ISO 2859-1. Minor defects: AQL 4.0. AQL applied to finished, packed goods. In-line inspection at approximately 30% of production. Defect definitions as listed in Annex A. Pre-shipment report with photographs issued before the balance payment.

Four things that clause does that “AQL 2.5” does not:

  1. It names the standard and the inspection level, so the numbers are not a matter of opinion.
  2. It separates critical from major from minor, so a critical defect cannot be averaged away.
  3. It defines the defect list, which is where disputes actually happen.
  4. It ties the report to payment, which is the only mechanism with teeth.

What this means for a 100 piece order

At the minimum order quantity, the plan is unusually strict: 20 units sampled, accept 1, reject 2. The practical implications are worth stating plainly.

Expect the factory to inspect more than the 20 units it will report on. Expect rework rather than a passed lot if two defects turn up. Expect that an order this size has little statistical power either way - 20 units out of 100 tells you a great deal about gross faults and relatively little about a low-rate cosmetic issue.

The honest conclusion for a small first order is that AQL protects you against a disaster, not against disappointment. For disappointment you need the specification locked, the swatch signed and the fill weight written down. Those are the controls that a certificate does not provide, because a certificate speaks to a test standard, not to whether the goods match the approved sample.

The short version

AQL 2.5 is a sampling plan with an accept number that changes with lot size. At 100 pieces it is 1 defect in 20 sampled units; at 10,000 it is 10 in 200. It counts attribute defects and is blind to uniform drift, so it complements a locked specification rather than replacing one.

If a quotation says “AQL 2.5” and nothing else, ask which defect classes it applies to and what the defect definitions are. The answer usually tells you more about the supplier than the number does. Every style in the catalogue is inspected to that plan, and the factory profile states the inspection regime; the contact form is where a specific inspection clause gets written into a quotation.

M

Marcus Vance

Director of Factory Operations & Sourcing Quality Control

Writes the test scope for each export order against EN 71, ASTM F963 and GB 6675, and coordinates third-party testing with accredited laboratories before shipment. Focus areas: mechanical stress testing, electromagnetic metal detection, and destination-market import compliance.

Recommended Technical Studies

View all research →