What AQL actually means
AQL — Acceptable Quality Level — is the maximum percentage of defective units that, on average, a buyer is willing to accept in a shipment. It is not a defect target. It is a statistical line in the sand: cross it, and the lot is rejected.
The standard reference is ISO 2859-1 (formerly MIL-STD-105E). It tells you two things: how many units to inspect from a lot of a given size, and how many defects trigger rejection.
Why one AQL number isn't enough
Most quote requests ask for "AQL 2.5". That is incomplete. Real AQL plans distinguish between three defect classes:
- Critical: defects that make the product unsafe or unusable. Magnus default AQL 0 — zero tolerance.
- Major: defects that reduce usability or shorten product life. Default AQL 1.0–2.5.
- Minor: cosmetic or non-functional issues. Default AQL 2.5–4.0.
If you specify only one number, you are bundling all three classes under it — which means a single shipment of cosmetically perfect parts with a fatal safety defect can still pass. We have seen this happen. We have a folder of recall stories proving it.
How we calibrate the inspection level
ISO 2859-1 lets you pick a general inspection level (I, II, III) or a special level (S-1 through S-4). The level controls sample size for a given lot.
Our rule of thumb for new suppliers:
- First three shipments: General Level III (largest samples, tightest control).
- Shipments 4–10, if zero majors: reduce to Level II.
- Shipments 10+, if first-pass rate above 98%: Level II with reduced inspection on minors.
- Any major defect found: immediate return to Level III for next three shipments.
This ratcheting approach means we spend more on inspection when risk is high and less when the supplier has earned trust. Net cost stays flat over a 12-month relationship.
If your supplier agreement says "AQL 2.5" without separating critical, major, and minor — fix the contract before you fix the parts. The single-number AQL is the most expensive shortcut in B2B sourcing.
Sample size in real numbers
What does this look like for a 10,000-piece shipment?
| Inspection Level | Sample size | Accept (AQL 2.5) | Reject |
|---|---|---|---|
| S-2 (special) | 20 | 1 | 2 |
| II (general) | 200 | 10 | 11 |
| III (general) | 315 | 14 | 15 |
For a Level III plan with AQL 2.5: 315 units inspected, 14 defects pass, 15 defects reject. The math is unforgiving. One extra defect flips a passing lot into a failing one.
The honest limits of AQL
AQL is a statistical filter, not a guarantee. A lot that passes at 2.5 still has, on average, 2.5% defects shipped. For 10,000 units, that's 250 defective pieces in your warehouse.
If you need defect rates below 1%, AQL sampling is the wrong tool. You need 100% inspection, automated optical inspection, or process control that ensures the defect rate is structurally lower than your tolerance.
Magnus typically combines both: AQL sampling at the supplier's gate, plus 100% functional testing for safety-critical parts at our consolidation warehouse.