Set AQL by defect class, not by a single house number

A 20,000-piece lot of 20 mL clear vials went out on a Monday with a signed inspection sheet that read “AQL 1.0, accept.” By Thursday the customer had quarantined it. Their incoming check found 14 vials in 200 with a crimp groove diameter running 0.15 mm under nominal, and their capper was producing about one loose seal per hundred. Our outgoing record had passed the lot because the sample was 80 pieces, the defect class was wrong, and nobody had measured the groove at all. That is the failure mode this article is about: an outgoing QC sampling plan that looks rigorous on paper and does not control the thing the customer actually measures.

An outgoing headspace vial AQL sampling plan is a decision rule, not a formality. It fixes three numbers before the lot is built: how many vials you pull, how many defects you will tolerate, and which characteristics count as defects at all. Get those three right and the inspection record becomes a defensible document. Get them wrong and you are shipping your customer’s reject rate back to them with your logo on the box.

Most vial plants I have walked through use one AQL for everything, usually 1.0 or 1.5, applied to a grab sample. That is the wrong structure. Headspace vials fail in ways that are not equivalent, and a single number flattens the difference between a cosmetic scratch and a seal that will not hold 1.5 bar.

I classify defects into three bands and assign a separate AQL to each. This is the same logic ISO 2859-1 uses for attribute sampling, applied to the characteristics that matter on a glass vial with a crimp finish.

Critical defects: AQL 0, zero tolerance

A critical defect is one that can cause a leak, a pressure failure, or an autosampler crash. On a headspace vial the critical list is short and specific:

• Chips or cracks in the neck seat or the crown, any size. A 0.2 mm chip on the sealing land is enough to lose the seal.
• Missing or malformed crimp groove (the 11 mm finish groove on a 20 mL vial, per the GPI 11 mm neck standard).
• Wrong neck finish entirely, for example an 18 mm vial packed into an 11 mm carton.
• Bore obstruction: glass spicules or a blocked septum seat that prevents needle entry.
• Septum not seated, or a septum that falls out when the cap is inverted.

For critical defects I do not run an AQL at all. The sample is inspected 100% for the critical list on the capping line, and the outgoing sample is a verification of that control, not the control itself. If a critical defect appears in the outgoing sample, the lot is rejected and the line is stopped. There is no accept number. AQL 0 with c=0 is a real plan; it just means every unit in the sample must pass.

Major defects: AQL 0.65

Major defects degrade function without guaranteeing failure. They are the ones that show up as an elevated reject rate at the customer’s incoming inspection rather than a single catastrophic seal loss. On a 20 mL headspace vial:

• Crimp groove diameter out of tolerance. Nominal for an 11 mm finish is around 11.0 mm at the groove; I hold ±0.10 mm and treat anything outside as major.
• Neck height or crown-to-groove dimension out of tolerance by more than 0.15 mm.
• Body diameter or concentricity drift that will not sit square in a 22.5 mm autosampler nest.
• Septum durometer out of range, or a PTFE facing with a visible pinhole or wrinkle.
• Cap torque out of band on pre-capped product (see the torque section below).

AQL 0.65 with a sample of 125 gives an accept number of 1 and a reject number of 2 under ISO 2859-1 general inspection level II. That is tight enough to catch a process that has drifted, and loose enough that a good lot does not get held for a single handling scratch.

Minor defects: AQL 2.5

Minor defects are cosmetic or handling-related. Faint scuffs on the body, a label out of register, a cap color slightly off, a bubble in the glass that does not sit on the sealing land. AQL 2.5 on a sample of 125 accepts up to 7. These do not stop a shipment. They do get logged, because a rising minor count is often the first sign that a forming tool is wearing.

Sample size comes from lot size and inspection level, not from habit

The most common error I see is a fixed sample size. “We pull 80.” That number is arbitrary. Under ISO 2859-1, general inspection level II, a lot of 20,000 maps to sample code letter M, which is a sample of 315 for normal single sampling. A lot of 3,200 maps to code letter L, sample 200. A lot of 500 maps to code letter H, sample 80. If you are pulling 80 from every lot regardless of size, you are over-inspecting small lots and under-inspecting large ones.

Here is the working table I use for headspace vials, general inspection level II, normal severity, single sampling:

Two operational notes on this table. First, the sample must be drawn randomly across the lot, not from the top layer of the top carton. I have seen a lot pass on a top-layer sample and fail on a bottom-layer resample because the pallet had taken on moisture during a dock wait. Draw from at least five cartons spread across the pallet, and from the middle and bottom of each carton. Second, if you are running reduced inspection because the last ten lots passed, you are allowed to under ISO 2859-1, but I do not recommend it for a new customer or a new tool. Reduced inspection on a vial is how a slow drift goes undetected for three months.

What a realistic outgoing inspection record actually contains

This is where most plants fall short. The record is usually a single line: lot number, quantity, “passed,” initials, date. That record will not survive a customer audit and it will not help you when a complaint comes in, because it does not let you reconstruct what was measured or how.

A usable record has six fields, and I have argued with quality managers about every one of them:

• Lot identity and traceability. Glass melt or forming batch, cap lot, septum lot, and the date the lot was sealed. If your septum supplier changes a compound, this is the field that lets you find every affected shipment.
• Sampling plan reference. The standard (ISO 2859-1), the inspection level, the severity, and the sample size actually drawn. Not “AQL 1.0.” The full plan.
• Characteristics measured, with the method. Groove diameter measured with what, at what resolution. Neck height with what gauge. Septum durometer on what scale. “Visual” is not a method unless the acceptance criterion is written down.
• Raw counts by defect class. Number of critical, major, and minor defects found, and the accept/reject numbers for each. A count of zero is a result, not an absence of data.
• Measurement data for the variable characteristics. Mean and range for groove diameter and neck height across the sample. This is the field that catches drift before it becomes a defect. A lot can pass on attributes and still be walking toward the edge.
• Disposition and signature. Accept, reject, or accept under deviation with the reason, plus the name of the person who made the call. A record without a named decision-maker is a record nobody owns.

If you want a model for the format, the attribute sampling logic is spelled out in ISO 2859-1 and the measurement discipline borrows from the same thinking behind U.S. EPA SW-846 data-quality documentation, which insists that a result carry its method and its uncertainty. Your vial record does not need to be that heavy. It does need to name the method.

Torque and crimp: the two numbers the record must carry

For pre-capped or pre-crimped headspace vials, two mechanical numbers belong on every outgoing record because they are the ones the customer’s autosampler will feel first.

Crimp torque. On a manual crimper, the setting that produces a reliable seal on an 11 mm aluminum crimp is typically in the range that yields a cap skirt compression of about 0.3 to 0.5 mm, which most shops express as a crimper setting rather than a torque value. On an automated capper, I specify a target torque and a band, commonly 1.2 to 1.6 N·m for an 11 mm aluminum seal, verified on a torque tester at the start and middle of each shift. Under-torque gives a seal that holds at room temperature and leaks at 80 °C in the headspace oven. Over-torque deforms the septum and produces coring on first puncture.

Crimp groove diameter. Measured with a go/no-go gauge or an optical comparator at the groove, not across the crown. This is the number that predicted the quarantine in the opening case. A lot running 0.10 mm under nominal will pass a visual and fail a capper.

If you buy pre-assembled vials, both numbers are your supplier’s responsibility and both belong on the certificate of analysis you receive. If you assemble in-house, both belong on your outgoing record. Either way, the customer’s incoming inspection will check them, so your outgoing plan should check them first.

Two field cases where the sampling plan was the problem

Case 1: the lot that passed on attributes and failed on variables

A 12,000-piece lot of 10 mL headspace vials passed outgoing inspection with zero major defects on a sample of 200. The customer rejected it at incoming with a 6% leak rate on a capping trial. The measurement that identified the cause was a groove diameter run on 50 retained vials: mean 10.94 mm, range 10.88 to 10.99. Nominal was 11.00, tolerance ±0.10, so every single vial was inside tolerance and the mean was sitting 0.06 mm low, close to the lower limit. The customer’s capper had been set up against a nominal-mean lot and could not close the gap on a low-mean lot without an adjustment they did not make.

The fix was not a tighter tolerance. It was adding mean and range for groove diameter to the outgoing record and flagging any lot whose mean sits more than 0.03 mm off nominal for a capper-setup note on the carton. The attributes plan was working. The variable data was missing.

Case 2: the septum lot that changed compound

A customer reported ghost peaks at retention times matching a common plasticizer, appearing only on vials from two shipments. Outgoing inspection had passed both lots on all attributes. The record showed the same cap and septum part numbers as the prior six months. What the record did not show was the septum compound lot. The supplier had moved from one silicone formulation to another with a different post-cure, and the extractable profile shifted. The measurement that found it was a 30-minute blank run at 120 °C on ten vials from each retained shipment, comparing total peak area above baseline. The new compound lot ran roughly three times the blank area of the old one.

The fix was adding septum compound lot to the traceability field and a periodic blank-run check to the outgoing plan for any lot built from a new compound lot. That check is not an AQL sample. It is a one-time qualification per compound lot, and it belongs in the record because it is the only thing that would have caught the change.

The outgoing plan does not replace in-process control, and vice versa

One structural point before the checklist. Outgoing inspection is a verification step, not a sorting step. If your outgoing sample is finding defects at a rate near the accept number, the process is out of control and the sampling plan is doing the job of a control chart badly. The plan I describe here is designed to confirm that a controlled process stayed controlled through packing and shipping. It is not designed to catch a forming tool that has been drifting for a week.

That distinction matters because it changes what you do with a reject. A reject at outgoing should trigger a look at the packing and handling steps first, then the process. A reject at incoming at the customer should trigger a look at your outgoing plan. If the customer finds what you did not, the gap is in your plan or your sample draw, not in their inspection.

An outgoing inspection checklist you can run tomorrow

This is the sequence I would put on the wall of a vial packing area. It assumes the lot is built and waiting for release.

• Confirm the lot is homogeneous. One melt, one forming run, one cap lot, one septum compound lot. If the lot mixes two of any of these, split it into two lots for sampling purposes.
• Draw the sample per the table above. Minimum five cartons, spread across the pallet, top and bottom of each. Record the carton numbers drawn.
• Inspect 100% of the sample for the critical list. Chips, cracks, groove presence, neck finish, bore, septum seating. Zero accept number.
• Measure groove diameter and neck height on the full sample. Record mean and range. Flag any lot with a mean more than 0.03 mm off nominal.
• Check crimp torque on ten vials if pre-capped. Target band 1.2 to 1.6 N·m for 11 mm aluminum.
• Run a blank headspace injection on three vials at the customer’s method temperature if the lot is from a new septum compound lot. Compare total peak area to the prior qualified lot.
• Count major and minor defects against the accept numbers. Record the counts, not just the disposition.
• Sign and date the record with the decision-maker’s name. Accept, reject, or accept under deviation with reason.

For the dimensional characteristics, the gauges and comparators you need are standard metrology items, and the reference dimensions for the 11 mm finish are the same ones the instrument makers publish in their consumables catalogs, for example the Agilent GC columns and accessories documentation or the Thermo Fisher chromatography and sample handling reference pages. For septum extractable behavior, the Restek technical resources library has blank-run guidance that maps well onto an outgoing qualification check.

Where the plan connects to the rest of your consumables

The sampling logic here is not unique to headspace vials. The same defect-class structure applies to the GC vial line, where a crimp or a screw-thread defect is the major class, and to the HPLC autosampler vial line, where septa and cap torque dominate. It also applies to the reagent bottle line, where the critical class is the pour ring and the closure thread rather than the crimp groove. If you run one sampling standard across all four lines, you save audit time and you stop arguing about sample sizes in four separate meetings. The full product range and the dimensional references for the headspace line are on the headspace vial pages, and the finish dimensions there are the ones your outgoing gauges should be set against.

Troubleshooting: when the outgoing plan and the customer disagree

When a customer rejects a lot you passed, work through this in order. The answer is usually in the first two steps.

• Did you measure the characteristic they rejected on? If it was not on your record, the plan has a gap. Add it before the next lot ships.
• Was your sample drawn from the same part of the pallet they sampled? Bottom-layer moisture and top-layer handling damage are real. Widen the draw.
• Was the defect class the same? A groove diameter 0.10 mm low is a major on your record and a critical on theirs if their capper cannot compensate. Align the class with the customer’s method.
• Did the lot mix two compound lots or two forming runs? If so, the sample was not representative and the reject is legitimate regardless of the count.
• Was the accept number applied to the wrong class? A single major defect against an accept number of 1 is an accept. Two is a reject. Check the arithmetic before you argue.

The decision rule is this: an outgoing headspace vial plan is defined by three numbers fixed before the lot is built, a sample size from the ISO 2859-1 table for the lot size, an AQL of 0 for critical defects and 0.65 for major, and a record that carries mean and range for groove diameter and neck height alongside the defect counts. If any of those three is missing, the plan is not controlling the lot. It is documenting a hope.

References & further reading

Related resources from our lab consumables network:

• Headspace Vial at headspacevialsupplier.com
• Gc Vial at gcvialfactory.com
• Reagent Bottle at reagentbottleforsale.com
• Hplc Autosampler Vial at hplcautosamplervials.com

External references and standards:

• Agilent chromatography consumables and columns
• Thermo Fisher chromatography and sample handling
• U.S. EPA SW-846 test methods for waste and materials
• Restek technical resources for chromatography