Handbag OEM Quality Control: AQL 2.5 Checks for OEM Runs

What does AQL 2.5 promise on a handbag run? It promises a sampling plan, not a perfect bag. Under ISO 2859-1 at general inspection level II, a 501-1,200 piece lot means 80 units drawn, 5 defects accepted and 6 rejected; a 3,201-10,000 piece lot means 200 units drawn, 10 accepted and 11 rejected. The value of naming AQL 2.5 is that it converts quality from an opinion into a number with an accept and a reject attached, and a number can be enforced at the inspection table. What it does not do is guarantee that every bag in the carton is good; it guarantees that a lot is statistically unlikely to be worse than the threshold.

The inspection standard in eight numbers

What AQL 2.5 Actually Promises

Acceptable Quality Limit sounds like a promise about products and is in fact a rule about lots. ISO 2859-1 defines how many units are drawn from a lot of a stated size, and how many defects of each class trigger a pass, a fail or a re-sort. The 2.5 in AQL 2.5 is the percentage of defective units the plan tolerates at the boundary, and the plan is designed so that a lot at roughly that level is usually accepted and a lot well above it usually is not.

That is a smaller promise than most buyers believe, and a far more useful one. It means the output of an inspection is a decision about a lot rather than a verdict on a product, and the two available decisions are accept or reject. Everything else in quality control is an argument about which of the two it is.

• The plan is fixed, not chosen at the table. Lot size determines the sample size; the sample size determines the accept and reject numbers.
• The defects are counted, not weighed. A critical defect, a major defect and a minor defect each count against their own threshold.
• The result is a document. A written report recording the lot size, the sample size and the defect counts is the only record that survives the conversation.

How Many Units Are Drawn From Your Lot

These four rows cover almost every handbag programme, because quantities cluster at the 500 pcs per style and colourway minimum and then at 2,000 or 5,000 for a season. Note that the sample size grows far more slowly than the lot: ten times the pieces means four times the units drawn, which is the entire economic logic of acceptance sampling.

Which Defects Count, and How Are They Classed

A defect is only a defect if the specification said so first, which is why the classification has to be written into the purchase order before the run starts. Three classes are standard, and each is counted separately against its own threshold.

Critical defects are the ones with a zero tolerance, and in this category they are usually about safety and chemistry rather than about construction: a broken needle fragment, a sharp edge on a metal part, or a restricted substance. Those are also the ones that intersect with the compliance dossier, because a needle-fragment recall and a nickel-release finding are both more expensive than any cosmetic re-sort.

What Gets Measured on the Bag Itself

The construction checks are the ones an inspector reaches for first, and they are all measurable. Written into the specification with a number attached, each becomes an instruction rather than a preference.

• Stitch density at 7-9 stitches per inch with a 10 mm seam allowance and Tex 40 thread. Below 7 the seam opens; above 9 the needle perforates the material and weakens it.
• Dimensions within plus or minus 3 mm on width, gusset and strap drop, measured against the approved sample rather than against the drawing.
• Handle and strap static load at 15 kg held for 30 minutes, with the attachment inspected afterwards for any movement at the stitch line.
• Seam strength tested to ISO 13935-2, typically around 180 N per 50 mm for a load-bearing seam.
• Zipper reciprocation for 500 open-and-close cycles, which is the check that catches a slider that works in the sample room and jams in the carton.

The second item is where most disputes start, and it is worth being precise: measurement is made against the approved physical sample, which is why approving and retaining that sample is a quality control act rather than an administrative one. A lot that matches the drawing but not the approved sample is a lot that will be returned by the customer.

What Gets Measured on the Hardware

Hardware is the part of a handbag that fails in the field rather than in the carton, and it is the part most often specified without a number. Three checks cover it, and each has a published test method behind it.

• Plating thickness at 8-12 microns, nickel-free, verified by a coating thickness gauge rather than by eye.
• Salt spray to 24 hours before first corrosion under ASTM B117, or its ISO equivalent ISO 9227, with 48 hours for a topcoated finish destined for a coastal climate.
• Nickel release within REACH Annex XVII, capped at 0.5 micrograms per square centimetre per week for items in prolonged skin contact, and reviewed under Proposition 65 for California.
• Pull and slider function at 500 cycles, the same reciprocation test applied to the zipper rather than to the seam.

The third of those is a legal requirement rather than a preference, and it is the check that most often arrives late. Asking for the test report at brief stage costs nothing; asking for it after a retailer has raised a question costs the listing.

What Gets Measured on the Colour

Colour disputes are the most common and the least necessary of all handbag arguments, because they are entirely preventable. Three decisions at brief stage remove almost all of them.

1. Approve a physical swatch, not a screen image, and retain it as the reference the later lots are matched against.
2. Name the standard, a Pantone TCX reference, and the tolerance, Delta E 2.0 or better measured in a light box at a defined illuminant.
3. Approve the lining and the thread as well as the body, because three materials in three shades is a mismatch even when each is individually acceptable.

Crocking is the other half of it: colour fastness to rubbing, tested to AATCC 8 or ISO 105-X12, with a grade 4 result expected dry and 3-4 wet for a coated material. A dark lining that crocks onto a light garment is a return that no sampling plan will catch.

When Is the Inspection Actually Booked

The inspection date is booked when the deposit is paid, not when the run finishes, and the difference is the whole point. A pre-shipment inspection needs the goods to be fully packed, which means the cartons have to be ready before the booking can be made, and a late booking either pushes the shipment or skips the inspection.

A Hamburg importer adopted the inline check on its first run of a new shape in 2025 and caught a strap attachment sequence that was one operation short. Fixing it at day 20 cost a re-sort of 1,400 pieces; fixing it at pre-shipment would have cost the whole lot.

What Happens When a Lot Fails

A failed lot is not the end of the process, it is the part where the process earns its keep. Four outcomes are available, and the right one depends on the defect class that caused the failure.

1. Re-sort. The lot is inspected piece by piece and the defective units are removed. Appropriate for a minor-defect failure.
2. Rework. The defective units are corrected, for example a strap resewn or a hardware part replaced, and then re-inspected.
3. Reject and remake. Appropriate where the defect is structural and present throughout, which is rare but not unusual on a new shape.
4. Accept with an allowance. A negotiated outcome where the buyer accepts the lot at a reduced price; it should be documented, because an undocumented allowance becomes the precedent.

Whatever the outcome, the switching rules follow: two lots rejected out of five consecutive lots moves a programme from normal to tightened inspection under ISO 2859-1, which draws a larger sample and applies a stricter threshold until five consecutive lots are accepted. That is the mechanism that makes a standard a system rather than an event.

How the Report Is Used Afterwards

The report is the only part of an inspection that survives the conversation, and it has three uses beyond the immediate accept-or-reject decision. All three are available at no extra cost if the report is requested by name.

• It records the defect profile of the shape, which is what makes the second season's specification sharper than the first.
• It satisfies an onboarding requirement; retail and marketplace buyers ask for the document, and it is usually the one missing from the folder.
• It becomes evidence, if a claim later arises about what was shipped. An undated photograph is not evidence and a signed report is.

Requesting it should be a line in the purchase order, alongside the ISO 9001 quality system reference and the audit references for the validated partner factory, BSCI (amfori) and Sedex / SMETA. Those three documents together answer the question a retail buyer will actually ask.

What AQL 2.5 Cannot Do for You

The standard has limits and they are worth stating plainly, because the disappointment that follows an accepted lot usually comes from expecting the wrong thing. Three limits matter.

• It does not inspect every unit. A 200-unit sample from 5,000 pieces will pass a lot containing some defective bags; that is the design of the plan, not a failure of it.
• It does not test chemistry. A substance restriction is verified by a laboratory report, not by a visual inspection at the carton.
• It does not fix a vague specification. If the purchase order says good quality, AQL 2.5 has nothing to count against and the inspection becomes a matter of opinion.

The third limit is the one worth acting on. Every number in this article only becomes enforceable once it is written into the specification, and the minimums and quantities and the sampling and bulk timeline are the two documents a purchase order is usually missing. The OEM programme reference carries the full document set.

What Defect Rate Actually Passes a 2.5 Plan

AQL 2.5 is not a line that separates good lots from bad ones; it is the quality level the plan is designed to accept most of the time. Every draw carries its own probability of acceptance, and the shape of that curve is what a buyer should hold in mind when reading a pass. The rows below are computed from the Poisson form of the single-sampling plan for a fifty-unit draw accepting three majors, and they show why an accepted lot can still disappoint. Choosing between the 2.5 and 4.0 plans is really choosing where on this curve a programme wants to sit.

The Checks That Run Before a Bag Is Unpacked

The carton is the first artefact an inspector handles and the least specified item in most briefs. Carton-level checks take twenty minutes, need nothing beyond a tape and a count, and catch two failure modes that no unit-level attribute check can see: a wrong assortment and a packing error that somebody else would otherwise discover later, at a charge.

• Count reconciliation. The physical carton number sequence is matched to the packing list before any carton is opened; a mismatch is recorded as a finding in its own right.
• Weight and handling. Gross weight per carton held at 15-18 kilograms so one person can lift it, with a five-ply corrugated specification stated rather than assumed.
• Assortment. Mixed size or mixed colourway cartons are flagged; on a three-colourway order this is the most common packing error and the most expensive one to find downstream.
• Moisture control. Twenty to fifty grams of desiccant per carton for leather and coated materials on a humid routing, and cartons stacked no higher than six in the container.
• Marks. Shipping marks read against the documents on at least two panels, because a carton found with the wrong mark at the destination is a relabelling job rather than a note.

The Needle Sweep: One Programme a Sampling Plan Cannot Replace

Metal contamination is a critical defect with an accept number of zero and no statistical argument available: eighty drawn units cannot characterise a risk that occurs perhaps once in ten thousand. It is therefore run as a hundred per cent operation with its own records, and it belongs alongside the material safety and restricted substance checks rather than inside the visual inspection.

• Detection sensitivity. Every finished unit passes a conveyor aperture set to detect ferrous spheres of 1.0-1.2 millimetres and non-ferrous of 1.5-2.0 millimetres at the tunnel centre.
• Validation three times a shift. Certified test coupons in ferrous 1.0, non-ferrous 1.5 and stainless 2.0 millimetres are run at start-up and every two hours; a failed validation voids everything since the previous one.
• Re-scan protocol. A triggering unit is rescanned twice from different orientations and cleared only if it passes; a persistent trigger means the unit is opened rather than argued about.
• Needle control at source. Each needle is issued against a signature and retrieved whole before a replacement is issued, with reconciliation logged per machine per day.
• Retention. Sweep logs and coupon checks are kept per lot, which is what turns a needle policy from a claim into evidence.

Frequently Asked Questions

People Also Ask