Most quality conversations happen at the sample stage, when everyone is optimistic. The order is actually won or lost in the last 72 hours before the container is sealed, when a pre-shipment inspection either releases the lot or holds it.
Buyers who understand AQL 2.5 — and its limits — stop discovering defects at the port. Here is the short version: AQL 2.5 is not a defect target, it is a sampling plan. It tells you how many random units must be inspected and how many defects are tolerated before a lot is rejected. What it does not tell you is which defects matter. That decision is yours.
What AQL 2.5 means in numbers
AQL is applied per defect class, on a random sample drawn from the finished lot. Take a 5,000-piece backpack order. Under ISO 2859-1 general inspection level II, the sample size is 200 pieces. For major defects at AQL 2.5, the acceptance number is 10 and the rejection number is 11 — so 10 units fail acceptance, 11 fail the lot.
Minor defects are usually run at AQL 4.0, where acceptance extends further. Critical defects are set at AQL 0: one unit fails the lot. That asymmetry is deliberate. A loose thread is not a product recall. A snapped buckle on a child’s bag is.
Defect classification is where the argument happens
Define three tiers before production starts, with examples written into the tech pack.
Critical — exposed sharp metal, a zipper that splits under load, mould or odour, an incorrect safety label. Zero tolerance.
Major — a failure the customer will return: broken stitching at a strap anchor, a misaligned zipper, hardware that will not latch, a logo placed more than 5 mm off spec.
Minor — cosmetic issues that do not affect use or brand: light loose thread, colour deviation inside tolerance, a faint scuff.
The gap between “major” and “minor” is where most inspection disputes live, and it is set by your tech pack, not by the inspector’s mood.
The 72-hour sequence
Hours 0–4: the sample is drawn at random from finished, packed cartons — never from units the factory selects.
Hours 4–24: measurement, function and cosmetic checks, with photographic evidence for every defect and a written report showing pass or fail per defect class.
Hours 24–48: rework or sort, depending on the defect tier.
Hours 48–72: re-inspection of the reworked portion, then release or hold.
Run it as a fixed protocol and the container date becomes predictable instead of hopeful.
Rework or release?
Critical defects: 100% rework and full re-inspection. There is nothing to negotiate.
Major defects above the acceptance number: sort the whole lot, rework the failures, and re-inspect the reworked units before release.
A pass with a handful of minors: ship, but log them so the next production run closes the gap permanently.
A supplier that argues to ship a failed lot is also telling you what the next order will look like.
Inspection is a system, not a formality
The value of pre-shipment inspection is not the report. It is the discipline that produces it: a defined sample, a defined defect tier, a defined hour-by-hour sequence and a defined release rule, applied identically to every run.
SimiBag runs pre-shipment inspection as a fixed gate on every production run, with photographic records and AQL 2.5 sampling as standard — so quality control lives on the line, not in the paperwork. See how we build it at https://simibag.com.