AQL Inspection Guide: How Quality Control Sampling Really Works
16 min read · Published September 8, 2026 · By the CN Ally sourcing team
Acceptable Quality Limits explained without the jargon: sample sizes, AQL 1.0/2.5/4.0 levels, defect classification, and how a real pre-shipment inspection runs.
What AQL is — and what it is not
AQL stands for Acceptable Quality Limit (sometimes called Acceptable Quality Level). It is the statistical backbone of product inspection worldwide, defined in the ISO 2859-1 / ANSI Z1.4 sampling standard. The core idea: instead of checking all 10,000 units in your order, an inspector pulls a statistically determined random sample — say 200 units — checks them against your criteria, counts the defects, and the lot passes or fails based on whether defects stay under the agreed limit.
The number everyone quotes, like "AQL 2.5," is a defect percentage threshold, not a guarantee. AQL 2.5 for major defects means the sampling plan is designed so that a lot with 2.5% defective units has a high probability of passing — it does not mean 2.5% of your shipment will be defective, and it definitely does not mean zero defects. Understanding this distinction matters when you set expectations with both your factory and your customers.
AQL is also not a product specification. It tells you how many defects are tolerable in a sample; your product spec, checklist, and reference sample tell the inspector what counts as a defect in the first place. The two work together: a perfect sampling plan applied to a vague spec still produces arguments. Write the spec first, choose the AQL second.
- AQL = Acceptable Quality Limit, from the ISO 2859-1 sampling standard.
- "AQL 2.5" is a statistical threshold, not a promise of 2.5% defects.
- AQL needs a clear product spec to work — sampling plans don't define quality, they measure it.
Reading an AQL table: sample sizes
Every AQL inspection starts with two inputs: your lot size (total units) and the inspection level. The standard general inspection level is Level II, which balances cost and confidence. You look up your lot size in the sample-size code table to get a code letter, then use that letter to find your sample size. For example, a lot of 3,201-10,000 units at Level II gives code letter L, which means a 200-unit sample. A lot of 91-150 units gives code letter F: 20 units.
Special inspection levels (S-1 through S-4) use much smaller samples and are reserved for checks that are destructive, slow, or expensive — like lab testing or drop tests. You would not use S-levels for a visual pre-shipment inspection; the sample would be too small to mean anything. Conversely, Level III gives larger samples for high-risk products where you want tighter confidence.
The practical takeaway: sample size grows with lot size, but sub-linearly. Inspecting 200 units out of 10,000 gives you roughly the same statistical confidence as 200 out of 5,000 — which is why splitting one 10,000-unit order into two 5,000-unit inspections mostly just doubles your inspection bill without doubling your assurance. Plan inspection around production batches, not arbitrary splits.
- Standard level: General Inspection Level II. Example: 3,201-10,000 units -> 200-unit sample.
- S-1 to S-4 levels are for destructive/expensive tests only.
- Sample size grows sub-linearly with lot size — don't split orders just to 'inspect more'.
Choosing your AQL levels: 1.0, 2.5, 4.0
AQL levels are set separately for each defect class, and the industry-standard starting point for consumer goods is: 0 for critical defects, 2.5 for major defects, 4.0 for minor defects. Critical defects are safety hazards or regulatory failures — a children's toy with a detachable small part, a charger that fails insulation tests. The tolerance is zero: one critical defect fails the lot. Major defects are functional failures or serious appearance issues the customer will notice and return — a dead-on-arrival electronic, a large visible scratch. Minor defects are cosmetic imperfections most customers accept — a tiny thread end, a barely visible molding mark.
You can tighten or loosen from the standard. Medical-adjacent or baby products often use 1.0 or even 0.65 for majors; simple promotional giveaways might accept 4.0 for majors. Tighter AQLs mean more failed inspections and higher factory costs (which come back to you in price), so match the level to the product's risk and your brand's tolerance — not to an abstract ideal of perfection.
Whatever you choose, put it in the contract before production, not in the inspection booking. "Pre-shipment inspection to AQL 2.5 major / 4.0 minor, zero criticals" is a sentence that prevents an enormous category of disputes. If the factory knows the standard it will be judged against while it is still making the goods, quality improves before the inspector ever arrives.
- Consumer-goods standard: 0 critical / 2.5 major / 4.0 minor.
- Tighten for safety-critical products (1.0/0.65 majors); loosen only for low-risk giveaways.
- Write the AQL levels into the contract before production starts.
Defect classification: the part everyone skips
Most inspection disputes are not about the AQL math — they are about whether a given flaw is major or minor. That classification has to be agreed in advance, in a defect list with photo examples. Build it from your reference sample: photograph acceptable and unacceptable versions of common issues (stitching, print alignment, color, finish, function) and attach them to the QC checklist. An inspector with photos makes consistent calls; an inspector with adjectives ("good finish") makes judgment calls the factory will dispute.
A practical defect list for, say, a canvas tote bag might read: critical — none applicable beyond regulatory (lead in print ink); major — broken stitching longer than 2 cm, misprinted logo, wrong dimensions beyond tolerance, non-functional zipper; minor — loose thread under 1 cm, slight print misalignment under 2 mm, minor fabric slubs. Notice the measurements: wherever possible, replace subjective words with numbers.
Also define your measurement tolerances in the spec — dimensions ±2%, weight ±5%, color within an agreed Pantone or physical swatch. Inspectors check against these, and factories respect numbers they agreed to far more than impressions they did not.
- Agree a photo-illustrated defect list before production — adjectives cause disputes, photos don't.
- Replace subjective criteria with measurements wherever possible.
- Define dimensional, weight, and color tolerances in the written spec.
The inspection process, step by step
A standard pre-shipment inspection (also called Final Random Inspection, FRI) runs like this. First, the inspector confirms the setup: is at least 80% of the order finished and packed (100% is ideal — below 80%, reschedule), are cartons accessible, is the reference sample and checklist on hand? Then quantity verification: counting cartons, checking carton markings, and confirming the packed quantity matches the order.
Next comes the sampling itself, and this is where professionalism shows. The inspector must pull cartons randomly from throughout the stack — not the top layer the factory conveniently left accessible — and pull units randomly from within cartons. The sample is then checked unit by unit against the checklist: workmanship, dimensions, function tests, packaging and labeling, and barcode scans where applicable. Common on-site tests include drop tests for packaging, tape adhesion, zipper cycle tests, and basic electrical function checks.
Finally, the inspector photographs everything — the good, the bad, and the setup — and issues a report, usually within 24 hours, with a clear PASS, FAIL, or PENDING verdict plus defect photos and counts. You then decide: accept, reject, negotiate rework, or request a re-inspection. The report is also your leverage for the 70% balance payment — never release the balance on a failed inspection without an agreed rework and re-inspection plan.
- Book FRI when 100% of goods are packed (minimum 80%) — earlier bookings waste the trip.
- Random sampling means random: cartons from throughout the stack, units from within cartons.
- Never release the 70% balance on a failed report without rework + re-inspection.
When to inspect: DUPRO vs. final inspection
Pre-shipment inspection is the last line of defense, but for custom or high-value orders it should not be the only one. A during-production inspection (DUPRO) happens when 20-40% of the order is complete. The inspector checks workmanship, materials, and early output against the spec while there is still time to correct course. Finding a systematic defect at 30% completion costs you a production adjustment; finding it at 100% costs you a rejected container.
Consider DUPRO mandatory for: first orders with a new factory, custom-developed products, orders above $20,000, and any product with a history of quality issues. For simple reorders of proven products from a trusted factory, final inspection alone is usually enough. Container loading supervision is a third option — an inspector watches the container being loaded and sealed, verifying quantities and carton condition — worth it for high-value FCL shipments where loading damage or short counts are the main risk.
Budget roughly $200-350 per inspector-day in China for a reputable third-party QC firm, with most single inspections completed in one day. Against a $15,000 order, that is about 2% — the cheapest risk reduction in importing. Book inspections 3-5 working days ahead; factories need notice to have goods ready, and good inspectors' calendars fill up.
- DUPRO at 20-40% completion catches systematic defects while they're still fixable.
- Use DUPRO for new factories, custom products, and orders over ~$20,000.
- Expect $200-350 per inspector-day; book 3-5 working days ahead.
What happens when a lot fails
A FAIL verdict is the start of a negotiation, not the end of the world. First, read the report carefully: are the failures concentrated (one defect type, one carton range — suggesting a fixable batch issue) or scattered (suggesting systemic problems)? Concentrated failures often resolve with targeted rework; scattered failures may require re-production.
Your options, in escalating order: (1) accept with a discount — sometimes rational for minor-heavy failures on low-risk products, but get the discount in writing and adjust your retail expectations; (2) require rework and re-inspection — the factory fixes or replaces defective units, then a second inspection (usually at the factory's cost if the contract says so) confirms; (3) reject and re-produce — for systemic failures, with a new delivery timeline; (4) cancel and recover the deposit — the nuclear option, governed by your contract terms.
This is why the contract matters more than the inspection. An inspection without contractual consequences is just expensive information. Your agreement should state: inspection standard (AQL levels), who pays for re-inspection after a fail (standard: the factory, if the failure is their fault), maximum allowable delay for rework, and your right to cancel with deposit refund if rework fails twice. Agree all of this when the factory wants your order — not when you are holding a failed report.
- Read the fail report for patterns: concentrated defects = rework; scattered = systemic.
- Options escalate: discount, rework + re-inspection, re-production, cancellation.
- Contractual consequences (who pays for re-inspection, cancel rights) must be agreed upfront.