A 96-well plate of blanks run on a Monday morning came back with siloxane peaks at m/z 73, 207, and 281 in every well, and the source traced to a single lot of 10 mm PTFE/silicone septa whose post-cure had drifted 12 °C low. The vendor’s certificate of analysis said “conforming.” That gap, between a document that says pass and a lot that fails, is what a vial septa supplier audit checklist has to close. Not the factory tour. Not the ISO certificate on the wall. The quality-system items that show up in the data months before they show up in a complaint.
I have run this audit from both sides of the table, as a buyer scoring suppliers and as a supplier getting scored. The vendors that fail quietly are almost never the ones with bad molding equipment. They are the ones whose process control lives in a supervisor’s head, whose change notifications arrive after the shipment, and whose CoA is a template with the numbers typed in at the end. This checklist is built to find those vendors before their septa find your chromatograph.
If you already know the product and just need the scoring framework, the sections below give you the evidence to request, the questions that separate a controlled process from a lucky one, and a weighted score you can defend to procurement. For the product side itself, the reference library at vialsepta.com covers the material and dimensional fundamentals; this piece stays on the supplier.
A price list tells you what a vendor wants to charge. A control plan tells you whether they can hold the charge. Ask for the control plan for the specific septum SKU you are buying, not a generic quality manual. It should name every characteristic that gets measured, the sample size, the frequency, the gauge, and the reaction plan when a point goes out of trend.
What I look for first is whether cure temperature and cure time appear as controlled parameters with a recorded setpoint and tolerance. For a silicone-backed septum, the post-cure is where residual volatiles drop to a level that will not ghost your GC. A vendor who controls cure by “the operator watches the oven” has no control at all. A vendor who logs ±5 °C at 200 °C for a defined hold and keeps the chart has something you can audit.
Second item: incoming material control on the compound. The septum is only as good as the gum and the cure system that went into it. Ask for the incoming CoA and the receiving inspection record for the last three lots of compound. If the vendor cannot tell you the lot number of the silicone that went into your septa, traceability stops at their door, and any failure becomes uninvestigable.
Evidence to request in advance, not on audit day
- Control plan for the SKU, current revision, with characteristic, method, sample size, frequency, and reaction plan.
- Process flow diagram with inspection points marked, from compound receipt to final pack.
- Last 12 months of SPC charts for the two or three characteristics that matter: thickness, durometer, and cure. Not a summary. The raw charts.
- Change notification log for the past two years. Every compound, mold, or process change, with the date the customer was told.
- CoA template plus three actual CoAs from recent lots, so you can check the template against reality.
- Calibration records for the gauges that measure your dimensions, with the calibration interval and the last due date.
- Nonconformance and CAPA log for the past 12 months, redacted as needed, with closure dates.
The change notification log is the single most predictive document in that list. A vendor who has changed compound suppliers twice in two years without telling anyone is a vendor whose lot-to-lot variation will show up in your retention times. Ask for it, then ask what triggered each change.
A CoA that lists only dimensions is a shipping document, not a quality record
Most septum CoAs I receive report thickness, diameter, and maybe durometer. That is a dimensional inspection, and it says nothing about whether the part will core, leak, or ghost. A CoA worth the paper should carry the characteristics that map to function.
For a silicone/PTFE septum, I want to see durometer (Shore A) with a tolerance band, typically 45 to 55 for a standard liner, and I want the actual reading, not just “pass.” I want thickness measured at multiple points across the part, because a 0.010 in nominal that drifts to 0.013 in changes the compression set and the seal. I want cure or volatile content reported, even if it is an internal method. And for anything going into a headspace or trace GC application, I want a background or bleed test result with the method stated.
The trap is the CoA that reports a value with no method and no tolerance. “Durometer: 50” means nothing if there is no band and no standard cited. Ask which ASTM method, what sample size, and where on the part it was measured. A vendor who can answer in one sentence without checking has a real system. One who says “I’ll get back to you” has a template.
| CoA characteristic | What good looks like | Red flag |
|---|---|---|
| Thickness | Nominal + tolerance, measured at 3+ points, gauge ID cited | Single value, no tolerance, no gauge |
| Durometer | Shore A, band of ±3, ASTM D2240 cited | Bare number, no method |
| Cure / volatiles | Internal method named, result vs. limit | Field omitted entirely |
| Bleed / background | GC method, conditions, result vs. limit | “Low bleed” as a word |
| Lot traceability | Compound lot + mold date + operator shift | Only a finished-goods lot number |
| Packaging count | Verified count, bag lot, date | Count assumed from weight |
That last row is not trivial. I have seen a shipment short by 4 percent because the vendor counted by weight and the parts ran heavy. The count on the bag is a quality record too.
Traceability has to reach the compound lot, or the audit is theater
When a lot of septa fails, the first question is “what changed.” You can only answer it if the finished part traces back through the mold run to the specific compound lot. Ask the vendor to walk one lot number from the bag on your bench backward to the compound receipt date. If the chain breaks anywhere, the vendor cannot do root cause analysis on their own product, which means every failure becomes your problem to absorb.
I want to see the mold identification, the press or cavity, the cure date, and the compound lot on the traveler. For multi-cavity molds, cavity-level traceability is ideal but rare. At minimum, the vendor should know which press ran the lot, because press-to-press variation in cure is real and shows up as bleed variation across a shipment.
This is where a lot of audits get soft. The auditor asks “do you have traceability?” and the vendor says yes and points at a finished-goods lot number. Push one level deeper. The question is not whether a number exists. It is whether the number connects to the material and the process.
Score the vendor on the items that predict lot-to-lot drift
An audit without a score is a conversation. You need a number you can put next to another vendor’s number and defend to a sourcing committee that does not run a chromatograph. The weights below are the ones I use, tuned over enough failures to know which items actually move the needle.
| Audit item | Weight | What earns full marks |
|---|---|---|
| Process control (cure, thickness, durometer) | 25 | SPC charts with limits, reaction plan, 12 months of data |
| Incoming material control | 15 | Compound CoA on file, receiving inspection, lot trace to part |
| Change notification discipline | 15 | Written PCN procedure, 30-day advance notice, log with no silent changes |
| CoA content and method rigor | 15 | Functional characteristics reported with methods and tolerances |
| Traceability depth | 10 | Bag to compound lot in one walk, cavity or press identified |
| CAPA effectiveness | 10 | Closed loop, verification of fix, recurrence rate reported |
| Calibration and gauge control | 5 | Gauges in cal, intervals respected, records current |
| Packaging and count control | 5 | Verified count, cleanroom or controlled pack, bag lot |
Set a pass line. I use 80 out of 100 for a primary supplier and 65 for a backup, with no single item scoring zero on process control or change notification. A vendor can be weak on packaging and still be usable. A vendor with no cure control is not usable at any price, because you will spend the savings on investigations.
How to score without arguing about opinion
Score each item against evidence, not against the vendor’s confidence. If the SPC chart exists and shows a stable process, full marks. If it exists but shows a process that was adjusted every week, half marks, because the process is being chased, not controlled. If there is no chart, zero, regardless of how good the parts looked last time.
Do the scoring on the evidence you received in advance. The audit visit confirms what the documents claim. It is not where you first learn the answer. That sequencing alone filters out half the vendors who would otherwise pass on personality.
Two field cases where the audit item predicted the failure
Case 1: the change notification that never came
A supplier switched silicone compound vendors to save cost and did not issue a PCN. We caught it three months later when a headspace method started showing a 1.8 percent RSD increase on the internal standard across a sequence, and the blank runs carried a low siloxane background that had not been there before. The septa looked identical. The durometer was in band. The CoA was clean.
The measurement that identified it was a bleed test on a retained sample from the previous lot against the current lot, run on the same GC under the same conditions. The current lot bled roughly three times the previous background at 280 °C. The fix was a compound re-qualification requirement written into the supply agreement and a mandatory 30-day PCN for any compound change. The audit lesson: we had asked “do you have a change notification procedure?” and accepted yes. We should have asked for the log and read it.
Case 2: the mold that ran hot on one side
A multi-cavity mold developed a heater zone fault that ran one side of the tool about 8 °C hotter than the other. Parts from the hot side came out slightly harder and cured further, which read as fine on a spot check. On the bench, septa from that side showed higher puncture resistance and a tendency to core on the third or fourth injection with a 22-gauge needle, while the cool side ran clean past ten. The complaint pattern was scattered because the parts were mixed in the bag.
We caught it by pulling 30 septa at random, measuring durometer individually, and plotting the distribution. It was bimodal, not normal. The fix on the vendor side was a thermocouple repair and a cavity-side identification on the traveler. The fix on our side was adding a durometer distribution check to incoming inspection for that SKU. The audit item that would have caught it earlier was cavity or press-level traceability, which the vendor did not have at the time.
Ask the questions that expose a paper system
Some questions separate a controlled process from a documented one faster than any chart. Ask them in the pre-audit questionnaire and watch how the answers are produced. An answer that comes from memory is a system in someone’s head. An answer that comes from a record is a system you can audit.
- “Show me the last time a cure parameter went out of limit. What did you do, and what does the record say?”
- “Which gauge measures thickness on this SKU, and when was it last calibrated?”
- “Walk me from this bag’s lot number to the compound lot. How long does that take you?”
- “What is your AQL for a cosmetic defect versus a dimensional defect on this part?” (If the answer is one number for everything, the sampling plan is not risk-based.)
- “How many CAPAs did you open last year, and how many recurred within 90 days?”
- “What happens to a lot that fails in-process inspection? Where does it go?”
The recurrence question is the one that separates real CAPA from paperwork. A vendor who reports a recurrence rate is measuring their own fix. A vendor who cannot answer has never checked whether the fix held. That matters to you because the failure you receive next quarter is probably a repeat of one they already “solved.”
The AQL question ties to the defect classification you should already have on your side. If you have not set a risk-based AQL for septa, do that before you audit, so you can compare the vendor’s sampling plan to yours. The HPLC vial and cap side of your consumables program benefits from the same discipline.
Match the audit depth to the application, not to the spend
Not every septum purchase justifies a full audit. A general-purpose screw-top septum for a prep method can survive a lighter review. A septum going into a validated trace method, a headspace run at 260 °C, or a regulated release test needs the full treatment. Scale the audit to the consequence of failure, not to the invoice value.
Here is the decision rule I use. If a septum failure would invalidate a batch release, trigger a deviation, or send you into an OOS investigation, it is a critical SKU and gets the full scored audit with on-site verification. If a failure would cost you a rerun and a shrug, a document review and a bleed check on the first lot is enough. The autosampler vial and its septum share that criticality, so audit them together.
For regulated work, the container and closure sit inside the method, and the septa sit inside the container. The EPA SW-846 framework and the USP container-closure expectations both assume the closure performs as specified. That assumption is only as good as the vendor’s process control, which is why the audit belongs in your method validation file, not just your purchasing folder. Vendors who publish detailed technical data, like the resources at Restek, tend to have the underlying records to back it up; ask for them.
What to do with the score once you have it
Two vendors, two scores, one decision. A vendor at 85 with a clean change log beats a vendor at 70 with a lower price every time I have run the numbers, because the cost of a bad lot is not the price of the septa. It is the rerun, the investigation, and the delayed release. Put the score in the sourcing file and revisit it annually, or after any PCN. A score is a snapshot, not a permanent grade.
Keep a retained sample from each lot and a copy of the CoA together. When the background creeps up in six months, you will have the physical reference to compare against, and the audit trail to prove what changed. That pairing is what turns an audit from an event into a control.
The decision rule
Audit the process control and change notification, not the showroom. If a vendor cannot produce twelve months of SPC data on cure and thickness, a compound lot traceable to your bag, and a change log with no silent substitutions, score them below the pass line regardless of how the parts measure today, because today’s parts are not the ones that will fail you. Set the pass line at 80 for a primary supplier, weight process control and change notification at 40 points combined, and require evidence in advance so the site visit only confirms what the documents already told you. The septa that pass a casual audit and fail on the bench are the expensive ones; the checklist exists to find them before the chromatogram does.
References & further reading
Related resources from our lab consumables network:
- Vial Septa at vialsepta.com
- Hplc Vial at hplcgc.com
- Autosampler Vial at autosamplerhplc.com
- Vial at vials-filters.com
External references and standards:
