AQL 2.5 for Knit Bags: What 2.5 Actually Means, the Sample Size It Sets, and Where It Misleads Buyers

What AQL 2.5 Means in One Sentence

AQL 2.5 is the worst average process quality that a sampling scheme will still accept most of the time, expressed as 2.5 percent nonconforming. It is a property of the sampling plan, not a promise about the lot, and the distinction is the single most useful thing a buyer can carry into an inspection conversation.

That sentence has a consequence buyers rarely price in. A plan indexed at AQL 2.5 is designed to accept the majority of lots at or better than 2.5 percent nonconforming; it is not designed to reject every lot worse than that. A lot at 4 percent will sometimes pass, and a lot at 1 percent will occasionally fail, because sampling plans carry producer risk and consumer risk on both sides.

For a knitted bag programme, the practical reading is that AQL 2.5 is a routine commercial threshold, not a guarantee of uniformity. Where a buyer needs near-zero defects on a safety or child-related item, a tighter limit and a different plan are required, and the place to say so is the order acknowledgement rather than the inspection report.

AQL 2.5 Is Not a Sample Size

This is the error that costs the most time in a dispute. AQL 2.5 fixes the accept and reject numbers once the sample size is known, but the sample size itself comes from two other decisions: the lot size and the general inspection level. Both are chosen by the buyer, and if the buyer does not choose them, someone else does at the dock.

General inspection level II is the default in most bag programmes and is the right starting point. Level I halves the sample and is sometimes proposed because it is faster; level III roughly doubles it and is appropriate where a buyer has seen quality drift or is inspecting a first lot from a new programme. None of the three changes the AQL; all three change the sample.

Read the table as a chain rather than a lookup. Decide the lot, decide the level, read the sample, then read the accept number. A buyer who writes only AQL 2.5 has handed over the first two links in that chain and can be presented with a legitimate inspection on a sample of 20 when 200 was expected.

Critical, Major and Minor: Where Defects Are Classified

A single AQL figure applied to everything is a blunt instrument, and most bag programmes use three. Critical defects, covering anything unsafe or legally non-compliant such as a sharp exposed needle fragment or a choking hazard on a child's item, are normally held at zero tolerance regardless of the AQL. Major defects, covering anything that makes a bag unsaleable, typically sit at 2.5.

Minor defects, covering workmanship issues a customer would notice but probably not return over, commonly sit at 4.0. The split matters because a lot with one critical defect fails outright while a lot with eight minor defects on a sample of 200 may still pass at 4.0, and both outcomes are correct under the plan as written.

The classification list is where the real work is, and it should be written down before production. For a knitted bag that typically means dropped stitches and holes as major, handle misalignment beyond a stated tolerance as major, loose thread ends as minor, and shade variation beyond the approved reference as major on the darkest colourway first.

Checklist: Ten Things to Fix Before the Inspector Arrives

1. Define the lot: one lot means one style, one colourway and one production run, not everything packed in one container.
2. State the general inspection level in writing; level II is the normal default and level I quietly halves the sample.
3. Write the defect classification list with critical, major and minor examples specific to a knitted panel.
4. Set the AQL per class: zero for critical, 2.5 for major, and 4.0 for minor is the common bag programme split.
5. Agree the dimensional tolerance in millimetres, because an inspector cannot classify a measurement against no limit.
6. Confirm the approved reference sample is the one on the table, with its gauge and yarn count recorded.
7. Decide whether inspection is at goods-in, during packing or both, and name who pays for a re-inspection.
8. State what happens on rejection: segregate, rework, re-inspect, and who carries the re-inspection cost.
9. Require photographs of every defect found, referenced to the piece number, not a summary defect count.
10. Name the sampling standard and edition on the report, so the accept numbers can be checked against the plan.

Ten items is a short list and it resolves most of what turns into an argument. The reason these arguments happen is not disagreement about quality; it is that two people each assumed a different inspection level, a different lot definition or a different defect class, and the report is correct under both readings.

The Acceptance and Rejection Numbers Buyers Misread

An accept number of 10 on a sample of 200 means the lot passes when ten or fewer nonconforming pieces are found. It does not mean that up to 10 pieces in 200 may be defective by design, and it certainly does not mean the buyer has agreed to accept 5 percent bad goods. The number is a statistical threshold for releasing a process, not a licence for defects.

The rejection number sits one above it, at 11 in this example, and the gap between the two is deliberate. There is no outcome between accept and reject in a single sampling plan: ten passes, eleven fails, and a sample of 200 that finds exactly eleven nonconforming pieces is a rejected lot no matter how good the other 189 are.

Where a result lands close to the line, a double sampling plan offers a second sample before a decision, which is useful when a borderline result would otherwise trigger a costly re-inspection. The trade-off is administrative complexity, so most bag programmes use single sampling and settle the borderline case by agreement rather than by plan design.

Switching Rules: Normal, Tightened and Reduced

ISO 2859-1 is a system, not a single table. It includes switching rules that move a programme between normal, tightened and reduced inspection based on the recent run of results, and the point of the system is to concentrate effort where quality has drifted and relax it where quality has been stable.

Tightened inspection normally follows a defined run of rejected or marginal lots and raises the bar without changing the sample size, by lowering the accept number. Reduced inspection is available after a sustained run of good results and cuts the sample, which lowers inspection cost but also lowers sensitivity. A buyer should say whether reduced inspection is permitted at all, because it is often applied by default.

On a programme of repeated seasonal orders these rules are genuinely useful, because they convert inspection history into a decision rule instead of a judgement call. On a single order they do nothing, and stating a switching rule for a one-off lot adds complexity without benefit.

Where AQL 2.5 Misleads on Knitted Panels

A knitted panel carries failure modes a woven one does not, and a pure attribute count does not see them. Pilling propensity, loop distortion after load, and handle elongation under a repeated 5 kg cycle are performance behaviours, not visible defects at goods-in, and a bag can pass AQL 2.5 on the day and pill or stretch within a month.

Shade is the second trap. A knitted body and a knitted handle may be from the same dye lot and still read differently because the two structures reflect light differently, and an inspector working to a numeric shade tolerance on a flat swatch will pass what a customer sees as two colours. Where shade matters, inspect the assembled bag under the lighting it will be sold under.

The third trap is dimensional. A knit relaxes after knitting and again after packing, so a bag measured at goods-in can be within tolerance and change by a further percentage once it has hung on a retail fixture for a week. Setting the tolerance against a conditioned and rested specimen, under the standard atmosphere in ISO 139, is the fix and it costs one line in the order.

Lot Definition: The Decision That Changes Everything

A lot is a defined quantity of product made under essentially the same conditions, and its size drives the sample. The mistake buyers make is letting the lot be defined by logistics, so that three colourways packed into one container become one lot of nine thousand pieces when they are really three lots of three thousand.

That distinction changes the sample and the risk. One lot of 9,000 gives a sample of 200; three lots of 3,000 give three samples of 200 for a total of 600 pieces inspected. The second reading is more work and substantially more sensitive, which is why a buyer with a shade concern should insist on lot-by-lot definition rather than per-shipment.

The same logic applies across machines and across days. On a floor running 7 production lines at around 200,000 pcs per month, a single style may be knitted on more than one machine, and defining the lot per machine catches a gauge difference between two machines that a single combined lot would average away into an accept.

Writing AQL 2.5 into a Purchase Order

The clause needs five fields and nothing else: the standard and edition, the AQL per defect class, the general inspection level, the lot definition, and what happens on rejection. Five fields, one line each, and the inspection becomes a check against a written plan rather than a negotiation on the day.

Two of those five are usually missing. The general inspection level is the most commonly omitted, and the rejection remedy is the second. Without the remedy, a rejected lot enters an open-ended conversation about rework and re-inspection cost at exactly the moment the buyer has least leverage and the least time.

Frequently Asked Questions

People Also Ask