AQL Inspection 2.5 Explained: How Pre-Shipment QC Actually Works in China

AQL — Acceptable Quality Limit — is the statistical standard underneath nearly every pre-shipment quality inspection done on Chinese-made goods. "AQL 2.5" is the most-used variant. Most importers know they want it; few know what it actually means.

This guide is the working-knowledge version: how AQL is defined, how sample sizes are picked, what counts as a major versus minor defect, what an inspection report looks like, and what AQL doesn't protect you against. The maths is built on ISO 2859-1 (the sampling standard) and MIL-STD-105E (its US predecessor) — the actual technical references used by every legitimate inspection firm.

Table of contents

1. What AQL is, what it isn't
2. The three defect classes
3. Reading the ISO 2859-1 sample-size table
4. The full AQL 2.5 specification importers use
5. How a pre-shipment AQL inspection actually runs
6. What the inspection report contains
7. The seven defect categories you'll see in a typical report
8. AQL pass — but you still get defects: managing expectations
9. Common AQL inspection mistakes
10. AQL alternatives: 100% inspection, S-3, S-4
11. What an AQL inspection costs
12. Worked example: a 5,000-unit Bluetooth speaker shipment

1. What AQL is, what it isn't

AQL is a sampling standard for accepting or rejecting a lot of finished goods. It answers a single question: given a representative sample of N units pulled from a lot of size M, how many defects are too many?

It's not:

• A defect rate. Confusingly, AQL 2.5 doesn't mean 2.5% of the lot is defective. It's a statistical confidence threshold built around a long-run average.
• A guarantee. A passing AQL inspection means the lot is statistically likely to have a defect rate at or below the AQL — not certainly. There's a probability the lot has more defects than the AQL but still passes (called "consumer's risk").
• A specification of what defects matter. The AQL number applies to whatever defect class you've defined as Major. You have to define the classification yourself or accept the inspector's defaults.

The actual definition: AQL = the maximum defect rate that, when offered as the average quality across a continuing series of lots, has a high probability of acceptance under the given sampling plan.

In plainer English: if a factory produces lot after lot at a 2.5% major-defect rate, the AQL 2.5 sampling plan will accept ~95% of those lots. If the factory's true defect rate is much higher (say 5%), the same plan will catch and reject most of those lots. The plan is calibrated to make this discrimination work.

2. The three defect classes

Every AQL specification distinguishes three classes of defect, each with its own AQL number:

Critical defects. Defects that render the product unsafe, illegal, or completely unusable. Standard tolerance: AQL 0.0 (zero allowed). Examples: live wire exposed in a power product, broken glass in a kitchen tool, missing safety label on a children's product, severed strap on a baby carrier.

Major defects. Defects that materially reduce the usability or saleability of the product, or that a customer would reject. Standard tolerance: AQL 2.5. Examples: incorrect colour of branded surface, broken zip on a bag, missing component, scratched LCD on a smartwatch, cracked ceramic.

Minor defects. Cosmetic flaws that don't affect function and that a typical customer wouldn't return. Standard tolerance: AQL 4.0. Examples: small surface scratch in a non-visible area, slight ink smudge on packaging, minor inconsistency in stitching, minor variation in colour shade.

When importers say "AQL 2.5" they almost always mean the standard 0/2.5/4.0 specification — Critical 0, Major 2.5, Minor 4.0. The 2.5 number specifically refers to the major class. If you want stricter or looser tolerances, you specify each class separately ("0/1.0/2.5" is tighter, "0/4.0/6.5" is looser).

3. Reading the ISO 2859-1 sample-size table

The size of the sample drawn for inspection depends on two inputs: the lot size (total units in the shipment) and the General Inspection Level (GIL). Most consumer-goods inspections use General Inspection Level II as standard.

Reduced extracts from the ISO 2859-1 sample-size table for GIL II:

For our typical example — a 5,000-unit lot — sample size is 200 units. The inspector pulls 200 randomly from across the lot (multiple pallets, multiple cartons) and inspects each one against the spec.

The accept/reject thresholds for AQL 2.5 at sample size 200 are then read from a second table:

So for a 5,000-unit Bluetooth-speaker shipment with AQL 0/2.5/4.0:

• If the inspector finds 0 critical defects, and
• 10 or fewer major defects in the 200-unit sample, and
• 14 or fewer minor defects in the 200-unit sample,

the lot passes. Eleven majors, or 15 minors, or even one critical defect, triggers rejection.

The thresholds are not a defect rate — they're acceptance numbers calibrated to the sampling plan to give the desired statistical behaviour.

4. The full AQL 2.5 specification importers use

When you write AQL into your order or inspection request, the full specification reads something like:

Pre-shipment inspection per ISO 2859-1, General Inspection Level II, single sampling plan, AQL 0/2.5/4.0 (Critical / Major / Minor). Defect classification per attached schedule.

That's the floor. Add:

• Defect classification schedule — a list of specific defects and which class each belongs to. Without this, the inspector defaults to industry-standard classifications which may not match your expectations.
• Functional test requirements — for electronics, what tests must each sampled unit pass (power on, key feature works, charging works, etc.).
• Packaging tests — drop test, vibration test, carton-strength check.
• Sample retention — what happens to defective samples (returned to factory, retained for record, destroyed).
• Pass/fail communication — who is notified within how many hours of inspection completion, what triggers a re-inspection.

A complete AQL specification is typically a 2–4 page appendix to the order or supply agreement.

5. How a pre-shipment AQL inspection actually runs

A typical inspection day for a 5,000-unit lot:

Hour 0–1: Arrival and quantity verification. Inspector arrives at factory, verifies order details, confirms total carton count matches the packing list, and confirms the lot is at least 80% packed (a partial-production inspection isn't representative).

Hour 1–3: Random sample pull. Inspector pulls cartons across the warehouse — different pallets, different production batches if any. Opens cartons, pulls 200 individual units randomly.

Hour 3–8: Defect inspection. Each unit is inspected against the spec — visual check for cosmetic defects, dimensional check on key dimensions, functional test for any electronic/mechanical features, label check, accessory completeness check, packaging check.

Hour 8–9: Tally and report. Inspector classifies all defects found (Critical / Major / Minor / not-a-defect), tallies counts, applies AQL acceptance criteria, prepares preliminary findings.

Same day: Report submission. Inspection firm sends the formal report (PDF, photos, defect list, accept/reject result) to you within 24 hours of inspection.

For larger lots (10,000+), inspection runs 1.5–2 days with 2 inspectors. For complex products (consumer electronics with multi-step functional tests) inspection time per unit doubles, so an inspection of 200 smartwatches might take a full day, while 200 ceramic mugs takes a few hours.

6. What the inspection report contains

A real AQL inspection report from any reputable firm (SGS, BV, AsiaInspection, or competent independents) includes:

1. Order summary — buyer, supplier, factory address, product name, PO number, lot size, sample size.
2. Inspection result — single line: PASS / FAIL / CONDITIONAL.
3. Defect tally — count of Critical, Major, Minor defects, with the AQL acceptance numbers for comparison.
4. Defect list — every defect found, with description, classification, photo reference, and unit ID.
5. Photo evidence — typically 50–100 photos: overall product, packaging, labels, every defect with close-up.
6. Functional test results — pass/fail per test for each functionally tested unit.
7. Dimensional check — measurements vs spec for key dimensions, with tolerance.
8. Packaging assessment — carton condition, label correctness, weight/dimensions vs spec.
9. Sample drop-test result — typical for fragile items.
10. Inspector's notes — anything unusual observed at the factory.

A 30-page report on a 5,000-unit lot is standard. If your inspection report is two pages, you're being undersold.

7. The seven defect categories you'll see in a typical report

Across thousands of inspections, defects cluster into a few categories:

Functional defects. Product doesn't work as specified. Dead-on-arrival electronics, leaking bottles, jamming mechanisms.

Cosmetic defects. Scratches, dents, paint inconsistencies, smudges, surface marks.

Dimensional defects. Out-of-tolerance dimensions, missing components, wrong sizes.

Material defects. Wrong material substituted, lower-grade material, material thickness below spec.

Assembly defects. Loose parts, misaligned components, missing screws, wrong assembly orientation.

Packaging defects. Wrong packaging, damaged cartons, wrong label, missing inserts, bar-code issues.

Documentation defects. Wrong country-of-origin marking, missing safety label, wrong certification marks, missing warranty card.

A specification with explicit defect-classification rules covering all seven categories is what separates a professional inspection from a generic one.

8. AQL pass — but you still get defects: managing expectations

This is the most-misunderstood point about AQL. An AQL 2.5 pass does not mean zero defects in the shipment. It means the lot is statistically consistent with a defect rate at or below 2.5% major-defect frequency.

In a 5,000-unit lot at exactly 2.5% major-defect rate, you'd expect ~125 major-defective units across the entire shipment. The inspection samples 200 — so the inspector might find 5 majors in their sample, calculating to ~125 across the lot. That passes the AQL but still ships you 125 defective units.

What this means for your downstream operations:

• Plan for 2–4% effective return-and-replace rate even on AQL-passing shipments.
• For higher-value items, consider tightening to AQL 1.0 or 1.5 on majors.
• For low-margin commodity items, AQL 4.0 may be acceptable — at the cost of higher returns.

If "near-zero defects" is what you actually need, AQL is the wrong tool — you need 100% inspection (every unit checked) or significantly tighter sampling. Both are more expensive.

9. Common AQL inspection mistakes

No defect classification list. "AQL 2.5" alone, without specifying what counts as Major, leaves classification to the inspector's judgement. Different inspectors classify the same defect differently. Always provide an explicit list.

Inspecting a partially-completed lot. If only 60% of the order is finished, the sample is not representative. Insist on 80%+ completion before inspection. (Reputable inspection firms enforce this; unscrupulous ones will inspect anyway, which gives meaningless results.)

Using factory-employed inspectors. Some factories will offer "internal QC" as a substitute for third-party inspection. The conflict of interest is obvious; the report value is near zero. Always use a genuinely independent third-party inspector or your own QC team.

Skipping functional tests on electronics. A visual-only inspection on Bluetooth speakers won't catch the 3% that don't pair, the 1% with poor sound quality, or the 2% with battery defects. Functional testing is essential and substantially increases inspection cost — and is worth it.

Mismatched expectations on cosmetic defects. Cosmetic standards vary by market. A scratch acceptable in B2B industrial use is unacceptable in retail. Specify your acceptance level explicitly with reference photos if needed.

Not following up on rejection. A failed AQL is a leverage point — the supplier wants the goods to ship. Use it to negotiate rework, full re-inspection, or a price discount on accepted goods. Don't just absorb the cost.

10. AQL alternatives: 100% inspection, S-3, S-4

AQL Level II is the workhorse but isn't always right.

100% inspection — every unit checked. Used for safety-critical items (medical devices, electrical safety items, children's products with high recall risk). Cost is much higher (5–10× AQL inspection per unit) and time is correspondingly longer.

Special Inspection Levels S-3 and S-4 — smaller sample sizes than General Level II, used for low-cost expendable items where inspection time per unit must be limited. S-4 typically halves the sample. Statistical power is correspondingly weaker.

Reduced sampling (Level I) — for products where defect rates are historically very low and you want to save inspection cost. Tightened sampling (Level III) does the opposite — used for new suppliers or recently-failed lots.

For most importers buying mid-volume mid-value goods, GIL II at AQL 0/2.5/4.0 is the right default. Move to 100% only when truly needed; consider tightened sampling for first orders with new suppliers.

11. What an AQL inspection costs

In China, third-party AQL inspection costs roughly $250–$400 per man-day. A typical inspection of a 5,000-unit lot is 1–1.5 man-days = $300–$600.

For comparison: returning 100 defective units from the US to China costs $1,500–$3,000 in freight alone, never mind discount/refund liability. The math on inspection is overwhelmingly favourable above ~$5,000 cargo value.

We typically include AQL pre-shipment inspection at no margin in our quality control service — it's the single highest-leverage part of the sourcing stack.

12. Worked example: a 5,000-unit Bluetooth speaker shipment

Specification:

• 5,000 units of branded Bluetooth speakers
• AQL 0/2.5/4.0 per ISO 2859-1, GIL II
• Defect classification list (3-page schedule attached)
• Functional test on 100% of inspected units (power on, Bluetooth pairing, audio clarity, battery hold for 1 hour)
• Drop test from 1m on 5 randomly-pulled units

Inspection executed at factory, sample size 200.

Findings:

• Critical: 0 (acceptance: 0; pass)
• Major: 8 (acceptance: 10; pass) — examples: 3× dead-on-arrival units, 2× cracked housing, 1× wrong logo orientation, 2× failed Bluetooth pairing
• Minor: 12 (acceptance: 14; pass) — examples: 6× small surface scratches, 4× minor packaging damage, 2× imperfect logo print

Result: PASS.

Estimated downstream impact: with 8 majors / 200 = 4% sample rate (above the 2.5% AQL but within the statistical accept band), expect ~200 major-defective units across the 5,000-unit shipment. Plan for ~4% return rate.

If we'd specified AQL 1.0 on majors (acceptance threshold 5 instead of 10), this lot would have FAILED. The supplier would either rework or accept a discount. For a mid-value Bluetooth speaker brand, the economics often favour the tighter AQL — fewer downstream returns more than offsets the cost of occasional rework.

The bottom line

AQL is the right pre-shipment inspection standard for most consumer-goods imports from China. AQL 2.5 (technically AQL 0/2.5/4.0) is the right default level. ISO 2859-1 General Inspection Level II is the right sampling table.

But the spec is only as useful as the supporting documentation: defect classification list, functional test requirements, drop tests, packaging checks, photo standards, and a real third-party inspector executing it.

If you've never run an AQL inspection, run one on your next shipment — even a flawed first one teaches you more about your supply chain than any amount of WeChat assurance from the factory.

If you'd like our team to define the AQL specification, classify defects for your specific product, and execute pre-shipment inspections, get a quote — QC is where most of our clients first realise what they were missing.

Related guides: Supplier Verification: 30-Point Checklist · Alibaba Trade Assurance Complete Guide · How to source from China in 2026