A lot of 1,200 9 mm screw-thread vials left the dock last March with a certificate that said AQL 0.65, and 14 of them came back from a customer’s IQ/OQ run with septa that had been punched by the filling needle and failed to reseal. The certificate was accurate. The sampling plan behind it was not. The lot was split into three sublots at the filling head, and the inspector drew a single sample from the combined lot using a plan that assumed homogeneity. That is the gap this article is about: the aijiren AQL sampling plan and outgoing QC record, what it should contain, and where the numbers usually go wrong.
I have written and audited these plans for vials, caps, septa, syringe filters, and pipette tips. The plans that hold up under a customer audit share four things: a defect classification that matches the failure mode, a sample size tied to the real lot size, an AQL level chosen per defect class rather than one number for everything, and an inspection record a third party can reconstruct without calling you. Below is how I build them, with the numbers I actually use.
Most supplier certificates I see quote a single AQL, usually 1.0 or 0.65, against the whole product. That collapses three very different failure modes into one acceptance number. A dimensional drift on a vial OD is a nuisance. A cored septum is a data integrity problem. A missing cap in a case is a line-stoppage problem. They should not share a plan.
I classify defects on aijiren lines into three bands, and I set the AQL per band, not per product:
| Defect class | Examples on a vial/cap/septum line | AQL used | Inspection level |
|---|---|---|---|
| Critical | Wrong septum material (silicone vs. PTFE-faced), wrong thread pitch, cap that will not engage GPI 9 mm, visible particulate in the vial bore | 0.10 | General II, tightened to S-4 when a prior lot failed |
| Major | Septum thickness out of tolerance, cap torque below reseal threshold, OD outside the 11.6 ± 0.2 mm window, liner bond separation | 0.65 | General II |
| Minor | Label skew, cosmetic scuff on the cap skirt, case print legibility, mold flash under 0.1 mm | 2.5 | General I |
The AQL levels come from ANSI/ASQ Z1.4 (the successor to MIL-STD-105E), which is the same table ISO 2859-1 uses for its normal inspection plans. If a customer asks for ISO 2859-1 explicitly, the acceptance numbers are identical for single sampling at General Inspection Level II. The difference is in the switching rules (normal, tightened, reduced), and those rules are what most certificates leave out.
Sample size follows lot size and inspection level, not habit
The single most common error I find is a fixed sample size. An inspector draws 32 pieces from a 5,000-piece lot and 32 from a 50,000-piece lot and calls both “AQL 0.65.” They are not the same plan. Under General Inspection Level II, the code letter changes with lot size, and the sample size changes with it.
Here is the table I hand to the QA lead on a new line, using the Z1.4 single-sampling normal inspection numbers:
| Lot size (pieces) | Code letter (Level II) | Sample size | Ac/Re at AQL 0.65 | Ac/Re at AQL 0.10 |
|---|---|---|---|---|
| 151 to 280 | G | 32 | 0 / 1 | 0 / 1 |
| 281 to 500 | H | 50 | 1 / 2 | 0 / 1 |
| 501 to 1,200 | J | 80 | 1 / 2 | 0 / 1 |
| 1,201 to 3,200 | K | 125 | 2 / 3 | 0 / 1 |
| 3,201 to 10,000 | L | 200 | 3 / 4 | 0 / 1 |
| 10,001 to 35,000 | M | 315 | 5 / 6 | 1 / 2 |
| 35,001 to 150,000 | N | 500 | 7 / 8 | 1 / 2 |
Two things to notice. First, at AQL 0.10, the acceptance number stays at zero up to lot size 10,000. That means a single critical defect rejects the lot, which is the right posture for a septum material mix-up but is expensive if you cannot trace the defect to a sublotal cause. Second, going from Level II to S-4 (special inspection) cuts the sample roughly in half but widens the confidence interval. I only drop to S-4 when a lot has passed three consecutive normal inspections and the switching rule allows reduced inspection. Never as a cost-saving default.
If you are buying through Vial Supplier or a similar distributor, ask which inspection level the plan uses. A certificate that quotes “AQL 0.65” without a level and a sample size is not a plan. It is a sentence.
Defect classification has to be written against the failure mode, not the part
I have seen classification lists that read like a dictionary of every possible flaw. That is not useful on the floor. The inspector needs a decision in under 10 seconds per piece. So I write the classification against the failure mode the part actually produces in use, and I tie each class to a measurable threshold.
What a critical defect looks like on a septum
For a 9 mm septum with a PTFE-faced silicone body, the critical defects I write into the plan are:
- PTFE facing absent, delaminated, or on the wrong side (measured by a 10-second visual under 10x and a fingernail lift test at the edge).
- Septum thickness outside 1.0 ± 0.1 mm at the center, measured with a dial indicator, 5 pieces per sample, all five within tolerance.
- Visible puncture or needle track in a “new” septum (rejected outright, no rework).
- Bore diameter of the vial neck outside 6.0 ± 0.1 mm, which will not seal against a standard 9 mm cap liner.
Notice that each one is measurable and each one maps to a failure the customer would see. That is the test. If a defect class cannot be measured or observed in under a minute, it belongs in a separate audit, not in the outgoing sampling plan.
Major defects carry the reseal threshold
The major class is where the reseal behavior lives. A cap that seals on first crimp but leaks after one puncture is a major defect even though it passes the initial leak test. I write the acceptance as: cap torque 8 to 12 in-oz at the crimp station, and reseal after 3 punctures with a 0.5 mm needle at 1 mL/min, holding 1.0 bar for 30 seconds with no bubble. Three punctures is the number I use because it matches the typical autosampler sequence on a 100-vial tray where a few positions get reinjected. If a customer runs a method that reinjects more, the plan tightens to 5 punctures and the AQL drops to 0.40.
This is also where the Vials With Caps catalog becomes useful as a reference: the cap and septum are sold as a matched set, and the torque window is a property of the set, not the cap alone. Sampling them separately is a mistake I have seen on two audits.
What a realistic outgoing inspection record contains
This is the part that separates a plan from paperwork. A record a third party can reconstruct has to answer: what lot, what sample, what defect, what decision, and who signed. Here is the field set I require on every outgoing inspection record, and it is shorter than most people expect.
- Lot or batch ID and the manufacturing date range it covers.
- Lot size in pieces and the number of sublots if the lot was combined.
- Inspection level (General II, S-4, etc.) and sample size drawn.
- AQL per defect class (critical, major, minor) and the Ac/Re numbers used.
- Defect tally by class with a short description of each defect found (not just a count).
- Measurement data for the dimensional checks: at minimum the min, max, and mean of OD, septum thickness, and cap torque across the sample.
- Functional result for the reseal and leak tests: pass/fail per piece, with the test parameters (pressure, needle gauge, puncture count).
- Disposition: accept, reject, or accept under deviation with the deviation number.
- Inspector ID and date, plus the QA reviewer signature.
What it should not contain: a single “PASS” stamp with no sample size, a copy of the previous lot’s data, or a defect count that does not reconcile with the sample size. I have rejected three supplier certificates in the last two years for exactly that third item. If the sample was 200 and the tally says 3 major defects with an Ac of 3, that is an accept. If the tally says 3 and the Ac is 2, the lot should have been rejected and the certificate is wrong.
The record should also state the switching status: normal, tightened, or reduced inspection, and the reason. A lot accepted under reduced inspection carries less confidence, and the customer has a right to know. Under Z1.4, reduced inspection is only allowed after 10 consecutive lots on normal inspection with no rejects and a satisfactory quality history. If your supplier cannot tell you which mode the lot was inspected under, the certificate is incomplete.
Two field cases where the plan caught what the certificate missed
Case 1: sublot mixing on a 35,000-piece screw-thread lot
Symptom: a pharma customer reported 6 of 100 vials in a tray failing the autosampler needle guide, with the needle catching on the septum edge. The certificate showed AQL 0.65, sample 500, zero defects.
Measurement: we pulled the retained sample and measured septum concentricity with a vision system. The mean offset was 0.08 mm, but the distribution was bimodal: one mode at 0.03 mm and one at 0.14 mm. The 0.14 mm mode came from a single cavity on a 16-cavity mold that had been run for the last 4 hours of the lot.
Fix: the sampling plan was drawing from the combined lot, so the defective cavity’s output was diluted across 35,000 pieces. We changed the plan to draw a stratified sample by cavity: a minimum of 5 pieces per active cavity, with the cavity ID recorded on the record. The AQL stayed at 0.65, but the sample size went from 500 combined to 500 stratified, and the next lot caught the same cavity at 2 defects in 5 pieces, which rejected the sublot before shipment. The mold was pulled for a pin replacement.
The lesson: a sample size of 500 sounds large until it is spread across 16 cavities and 4 hours of production. Stratify by the process variable that can drift, and the plan becomes sensitive again.
Case 2: cap torque drift on a crimp line
Symptom: a customer running EPA Method 8260 on a SW-846 volatile organics panel saw 4 of 40 vials lose more than 15% of the internal standard after 24 hours in the autosampler tray. The vials were sealed, the septa were new, the caps were the specified 9 mm magnetic.
Measurement: we measured cap torque on the retained sample with a torque tester at the crimp station. The mean was 7.2 in-oz with a range of 5.8 to 9.1. Our specification is 8 to 12 in-oz. The crimp head had drifted on the low side, and the reseal after puncture was failing at 2 punctures on the low-torque pieces.
Fix: the outgoing plan had been checking cap torque on 5 pieces per sample, which is too few to catch a drift that only affects the bottom of the distribution. We moved cap torque to a 20-piece measurement per sample, with the min and max recorded, and set the acceptance at min ≥ 8.0 in-oz and max ≤ 12.0 in-oz. The crimp head was recalibrated weekly, and the next 6 lots held 8.4 to 11.2 in-oz. The customer’s ISTD recovery went back above 90%.
If you are comparing this to a general incoming inspection program, the difference is that this check happens before the lot leaves the building, not after it arrives. The Lab Chromatography resource library has the method context for why a 15% ISTD loss matters on a volatile panel, but the fix is a torque specification, not a method change.
Where the sampling plan interacts with the standards you already cite
Two standards do most of the work here, and they are easy to confuse.
ANSI/ASQ Z1.4 and ISO 2859-1 set the acceptance sampling tables. They tell you how many pieces to draw and how many defects to accept. They do not tell you what a defect is. That is your job, and it is where the plan lives or dies.
USP <788> and <789> set particulate and visible inspection requirements for the finished product, and for a consumable the relevant guidance is usually the customer’s method validation, not the compendial chapter. If a customer cites USP <788> on a vial, they are asking about particulate in the vial bore, and that check belongs in the critical defect class with a 100% visual under controlled lighting, not in a sampling plan. Sampling cannot reliably catch a single particulate in a 1,200-piece lot at AQL 0.10, so the plan has to say “100% visual” for that one defect and sample everything else.
The external references that help calibrate the functional side are the instrument vendor application notes. Agilent’s GC columns and accessories documentation gives the septa and liner specifications that a vial has to match, and Thermo Fisher’s chromatography sample handling pages cover the autosampler tray conditions that drive the reseal requirement. Restek’s technical resources are useful for the septa bleed and puncture-count data that justify the 3-puncture acceptance number. None of these tell you your AQL. They tell you what the part has to do, which is the input to the defect classification.
A working decision checklist for a new lot
This is the list I run before a lot is released. It takes about 40 minutes for a 10,000-piece lot and catches most of what a certificate misses.
- Confirm the lot is homogeneous. If it was filled across more than one shift or more than one cavity set, treat each as a sublot and sample each separately.
- Set the inspection level from the lot size using the Z1.4 table, not from habit. Record the code letter.
- Draw the sample stratified by cavity and by shift. Minimum 5 pieces per cavity for a multi-cavity mold.
- Run the critical checks first. Septum material (visual + fingernail lift), septum thickness (5 pieces, dial indicator), bore diameter (5 pieces, pin gauge).
- Run the functional checks. Cap torque on 20 pieces (min 8.0, max 12.0 in-oz), reseal after 3 punctures at 1.0 bar for 30 seconds, leak test on 10 pieces at 1.5 bar for 60 seconds.
- Tally defects by class and compare to the Ac/Re numbers. Do not accept a lot on a “looks fine” judgment.
- Record the switching status and the reason. If reduced inspection is in force, note the 10-lot history that justifies it.
- Sign and date the record, with the inspector ID and the QA reviewer. File it with the retained sample, not in a separate folder.
One more operational note: retain a sample of at least 30 pieces per lot for 12 months. When a customer reports a failure six weeks later, the retained sample is the only way to reconstruct whether the defect was in the lot or introduced downstream. I have settled three disputes in the last year with a retained sample and a torque measurement, and lost none where the retained sample was missing.
The Aijiren outgoing QC program uses this structure as the baseline, with the AQL per class, the sample sizes above, and a record that ships with the certificate. If you are auditing a supplier, ask for one complete record from a recent lot and check three things: the sample size against the lot size table, the defect tally against the Ac/Re numbers, and the switching status. If any of the three is missing, the plan is not being run as written.
The decision rule is simple. Set the AQL per defect class, not per product. Derive the sample size from the lot size and inspection level, and stratify it by the process variable that can drift. Write the record so a stranger can reconstruct the accept/reject decision from the fields alone. If the certificate cannot survive those three checks, the lot has not been inspected, it has been stamped.
References & further reading
Related resources from our lab consumables network:
- Aijiren at aijirentech.com
- Vials With Caps at vialswithcaps.com
- Vial Supplier at vialsupplier.com
- Lab Chromatography at labchromatography.com
External references and standards: