What Does AQL 2.5 Mean for a Bag Order: A Practical Explanation

Every bag quotation carries three letters and a number somewhere near the bottom, usually written as "inspection: AQL 2.5". Very few of those quotes say what the buyer is actually buying. AQL 2.5 is not a guarantee, not a pass rate and not an inspection report. It is one component of a sampling plan, and on its own it is nearly meaningless: without the lot size and the inspection level it has no sample size attached, and without a defect classification it has nothing to count.

What Does AQL Actually Stand For?

AQL stands for acceptance quality limit. ISO 2859-1 defines it as the worst tolerable process average when a continuing series of lots is submitted for acceptance sampling. Two phrases inside that definition carry all the weight. "Process average" means the figure describes a stream of production, not a single carton. "Continuing series" means the plan assumes the same product will be offered again, so performance is judged over time rather than once.

The wording changed deliberately. Older documents used acceptable quality level, which read as though the buyer had agreed to accept goods of that quality. The shift to "limit" fixed that reading: the figure is a boundary the vendor submits against, and a plan built on it should accept most lots at that quality and progressively fewer lots as quality gets worse. Expressed as a percentage, AQL 2.5 means an average defect rate of 2.5% sits on that boundary.

How Does AQL 2.5 Turn Into a Number of Units?

Three inputs build the plan, in this order. The lot size and the inspection level go to the code letter table and return a letter. The letter returns a sample size. Only then does the AQL enter, and it selects a column in the second table that gives an accept number and a reject number. Sample size first, accept and reject second. Buyers routinely invert this, assuming a tighter AQL buys more units examined. At General Inspection Level II it usually does not.

Take the standard first order. A lot of 500 pieces at General Inspection Level II returns code letter H, and H means a sample of 50 units. Under single sampling at AQL 2.5, that plan accepts the lot when the sample holds 3 defects or fewer and rejects it at 4. The same 50 units are then read against the other classes: zero acceptance for critical defects, and AQL 4.0 for minor ones. One sample, three verdicts, three columns.

Why Is the Sample 50 Units and Not 13 or 200?

Because the standard fixes it. General Inspection Level II maps a lot of 281 to 500 pieces to code letter H, and H carries a sample of 50. Switching the inspection level is what moves the number: for that same 500-piece lot, Level I draws 20 units, Special Level S-4 draws 13, and Level III draws 80. Levels above II cost more inspector time and produce more rejections of marginal lots; Level I and the special levels examine so little that a bad lot can pass unnoticed.

What the plan does beyond accept or reject is described by its operating characteristic curve: the probability of acceptance at each possible defect rate. A plan of 50 units with accept 3 will take a lot at the AQL most of the time, will take a lot roughly twice as bad some of the time, and still passes a genuinely poor lot occasionally. That residual risk is the price of sampling rather than sorting. It can only be reduced by examining more units, not by moving the AQL column.

What Counts as a Defect Under AQL 2.5?

Nothing does, until somebody writes it down. This is the single largest gap in most inspection arrangements. ISO 2859-1 supplies the sampling scheme and nothing else: no defect definitions, no list of faults, no tolerances. The classification is the buyer's document, and it has to be written per product because the same fault changes class with duty. A lifted corner on the binding of a bag that carries 8 kg is cosmetic; the same lift on a strap root is structural.

A usable classification names the fault, states the acceptable limit, gives the measuring method, and assigns the class. "Print illegible" is not measurable. "Print loses legibility under normal reading light at 400 mm" is. Where an inspector has to guess, the guess will favour whoever wrote the report, and that is generally not the party paying for the goods.

Does AQL 2.5 Mean 2.5% of Units Are Allowed to Be Bad?

No, and the difference matters when a shipment arrives. First, the figure is a process average across many lots, not a ceiling for one lot. A single lot sitting at exactly 2.5% can still fail, because the verdict is taken on 50 units and whatever those 50 units happen to contain. Second, the figure says nothing about the 450 units that were never opened. A passed lot is an estimate that the lot is probably acceptable, nothing stronger.

There is a third, quieter point. Sampling assumes defects are spread through the lot. When defects cluster — one machine, one shift, one roll of fabric, one carton packed last — a random sample misses them, and this is why random selection has to be genuine. Draw ex-works from the full packing list rather than from the two cartons nearest the door. A plan described as random and executed as convenient is worse than no plan, because it carries the appearance of control.

Where Does AQL 2.5 Sit Between 1.5 and 4.0?

Those three numbers are columns in the same table, read on the same sample. At code letter H with 50 units: AQL 1.5 accepts 2 and rejects 3; AQL 2.5 accepts 3 and rejects 4; AQL 4.0 accepts 5 and rejects 6. The whole difference between the tightest and loosest of the three is three units of tolerance on the same 50. That is why arguing over the column while ignoring the sample size is wasted effort — the sample is doing most of the work.

General merchandise bags run comfortably at zero for critical defects, 2.5 for major defects and 4.0 for minor ones. Tighter levels have their place: where the unit cost is high, where the downstream sort cost dwarfs the inspection cost, or where the defect cannot be found by the customer until the product is in use. The cost of a tight column is real and predictable — more lots fail on borderline counts, each failure costs a re-work cycle plus a second inspection, and the programme loses access to reduced inspection.

What Happens When the Sample Contains Exactly the Accept Number?

The lot passes. The accept number is inclusive: three defects found in a 50-unit sample at AQL 2.5 is a pass, four is a failure. Everything commercially interesting happens on that one-unit boundary, which is why the count should be recorded per defect rather than in total, with a photograph of each instance and its location. Totals hide the fact that three defects might be three instances of one systematic fault.

Two procedural details belong in the instruction. The first is when counting stops: most agencies stop on reaching the reject number, some continue to produce a fuller report for a fee, and the difference changes how much you learn from a failure. The second is whether the plan is single or double sampling. Single sampling is the default and should be stated as such; double sampling takes a first smaller sample and only draws the second when the count lands between two thresholds, which suits programmes with a high failure cost per unit.

How Do the Switching Rules Change the Meaning of AQL 2.5?

ISO 2859-1 is not one plan but a system with four states, and it moves between them on the recent history of lots. AQL 2.5 describes behaviour in the normal state only. Once a programme is pushed into tightened inspection, the same AQL 2.5 label runs with a stricter accept and reject pair on an unchanged sample, and lots that would have passed last month now fail.

Ask which state the programme is in before agreeing a number, and have it printed on every report. A vendor quoting AQL 2.5 while running under tightened inspection is offering more than appears; a buyer comparing two suppliers by column alone, with one in each state, is comparing nothing.

What Does AQL 2.5 Cost in Calendar Terms?

Inspection sits at the end of the bulk window, not inside it. Bulk production runs 35-50 days from written sample approval, and the inspection is booked against packing progress — normally when the order is at least 80% packed and fully finished. Drawing 50 units from a 500-piece lot, opening them, measuring them and writing the report occupies one inspector for most of a working day. The report follows within a day of the visit.

The expensive case is the failure. Re-work and resubmission adds the repair cycle plus a complete second inspection to the schedule, and because the re-inspection draws a fresh sample it carries the same statistical risk as the first. This is why the outcome of a failed lot belongs in the purchase order rather than in an argument afterwards: who pays for the re-work, who pays for the second visit, and whether a partial release against a sorting operation is available when the retail date is fixed.

How Should AQL 2.5 Be Written Into an Order?

Six lines, attached to the purchase order and repeated on the tech pack. Each one removes a decision that would otherwise be taken by whoever is in the room on inspection day, usually to their own advantage. Write them once, keep them identical across every purchase order in the programme, and the inspection stops being negotiable.

A specification written this way survives a change of inspector, a change of account manager and a change of shipping date. The general-plan checks behind the number can be worked through with the AQL sampling table before a purchase order is issued, and the question of whether 2.5 is the right column for the duty is argued out in AQL 1.5 vs 2.5 vs 4.0. The number itself is the smallest part of the work.

Frequently Asked Questions

People Also Ask