Last spring a 40-lot incoming inspection on 2 mL crimp vials turned up a dimensional drift that had nothing to do with the glass: the aluminum seal skirt ran 0.18 mm under nominal, which pushed the crimp force needed to pass a dye-leak test from 2.4 kN to 3.1 kN, and two of every ten vials in the affected lots leaked at the septum-to-neck interface under 1.5 bar. The vendor’s certificate of analysis said “conforms.” That is the whole reason a vial supplier audit checklist has to be built around evidence you can pull before the audit, not around what you see on the day you show up. If your checklist can’t predict that a lot will fail a dye test six months out, it’s a tour itinerary, not an audit.
I’ve sat on both sides of this table. I’ve been the vendor getting audited and the buyer running the audit, and the pattern is consistent: the suppliers that fail you in year three are the ones whose quality system had a soft spot you walked past because the paperwork looked complete. Here is how I score a vial vendor now, and what I ask for before anyone books a flight.
The single most useful change I made to our audit process was moving 70% of the assessment into a document request that goes out four weeks ahead of any site visit. Vendors that can’t produce the package in that window are telling you something about their document control. Vendors that produce it in three days with clean cross-references are usually the ones whose system is live, not reconstructed.
Ask for these, and ask for them as records, not summaries:
- Control charts, not just CoA values. The last 12 months of dimensional data on three critical features: neck OD, crimp groove depth, and septum thickness. I want to see the chart, the control limits, and the sample size per subgroup. A vendor who sends a table of pass/fail is hiding the distribution.
- Change control log for the last 24 months. Resin supplier changes, mold refurbishment, annealing oven recalibration, glass tubing source. Any of these can shift a critical dimension without a single out-of-spec result on the CoA.
- Gauge R&R study for the dimensional measurements. If their measurement system contributes more than 10% of the tolerance to the variation, their data can’t detect a real drift. I’ve seen a vendor’s “in-spec” neck OD data turn out to have a 22% gauge contribution, which meant the chart was noise.
- Calibration certificates for every gauge that touches a critical dimension, traceable to a national standard, with the calibration interval stated. Anything past interval is a finding.
- Resin and glass certificates of analysis from the raw material suppliers, matched by lot to the finished goods lots you’ve received. This is where hydrolytic resistance class gets verified at the source. For Type I borosilicate the relevant property is the USP <660> / ISO 720 hydrolytic resistance class, and I want to see the grain test result on the incoming tubing certificate, not a blanket statement on the finished vial.
- Complaint log and CAPA register for the last 24 months, with closure dates and effectiveness checks. Open CAPAs older than 90 days are a yellow flag; CAPAs closed with no verification step are a red one.
- Sub-tier supplier list for septa, caps, and any secondary operation (siliconization, marking, packaging). You are auditing the weakest link in that chain whether you visit it or not.
If a vendor pushes back on the control charts, offer to sign an NDA and take the data on-site. The refusal itself is data.
The quality-system items that actually predict a vial failure
Most audit checklists are organized around ISO 9001 clauses. That’s fine for a management-system certificate and useless for predicting whether next quarter’s vial lots will pass your autosampler. I organize the on-site assessment around four failure modes I’ve actually chased down, and I score the system on whether it would catch each one.
Dimensional drift on the crimp and neck features
The question is not “do you inspect.” It’s “what is your sampling plan, and does it catch a 0.1 mm shift in neck OD?” A vendor running AQL sampling at 1.0% on a lot of 50,000 vials inspects roughly 200 pieces. A 0.1 mm shift that affects 3% of the lot will pass that plan more often than not. What I want to see instead is in-process SPC with a subgroup every 30 minutes and a reaction plan that triggers on a trend rule, not just a point outside the limit. Ask them to show you the last time the chart triggered a reaction and what they did. If they can’t produce one in 12 months, either the process is perfect or the chart isn’t being watched.
Septum compression set and puncture performance
This is where the headspace vial and crimp-side vendors separate themselves. The property that matters is compression set after 24 hours at the service temperature, and the property that gets missed is puncture retention after multiple injections. Ask for the vendor’s data on puncture count before coring for their standard 10 mm septum, and the test method behind it. A good silicone/PTFE septum holds 5 to 8 punctures with a 22-gauge needle before the reseal force drops below 3 N. A bad one cores on the second. The audit question: is that test run on every lot, or was it a one-time qualification? If it’s one-time, the septum you get in year two may not be the one that was qualified.
Annealing and the residual stress nobody measures
Glass that isn’t properly annealed passes every dimensional check and then cracks at the crimp six weeks later in your customer’s hands. The predictor is the annealing oven’s temperature profile record. Ask to see the continuous temperature log for the last 30 days, the setpoint, and the alarm history. A properly controlled lehr holds the annealing point within ±5 °C across the belt width. I’ve found ovens running a 15 °C gradient across the belt that still produced vials passing hydrolytic resistance, because the stress was mechanical, not chemical. That’s a failure your incoming inspection will never see until it becomes a field return.
Cleanliness and the ghost peak that isn’t the column
For GC vial suppliers, the audit item is the wash and dry cycle: what solvent, what water resistivity (I want ≥18.2 MΩ·cm for the final rinse), what drying temperature, and how the cleaned vials are handled before packaging. Ask for the last environmental monitoring result in the packaging area and the bioburden data if the vials are sold as clean. The failure mode here is a silanizing agent or mold release residue that shows up as a ghost peak at retention time 4.2 minutes on a DB-5 column and gets blamed on the column for a month. The audit question: can they show you a lot-specific GC-MS scan of a solvent extract from a cleaned vial? If not, they’re selling you cleanliness you can’t verify.
Score the vendor on a weighted matrix, not a gut feel
The reason audits produce inconsistent decisions is that most checklists are unweighted. A vendor with a beautiful calibration program and no change control scores the same as one with the reverse. I use a weighted matrix where the weights reflect how much each item predicts a real failure in my workflow. Here is the version I use for vial vendors. Adjust the weights for your application, but keep the structure.
| Audit area | Weight | What earns full marks | Score 1-5 |
|---|---|---|---|
| In-process SPC on critical dimensions | 20% | Subgroup ≤30 min, trend rules active, documented reaction in last 12 months | |
| Change control and lot traceability | 20% | Full raw-material-to-finished-goods trace in under 4 hours, change log with impact assessment | |
| Septum performance testing | 15% | Lot-specific puncture and compression-set data, method stated | |
| Annealing / thermal process control | 15% | Continuous profile logs, ±5 °C belt uniformity, alarm history clean | |
| Measurement system (gauge R&R) | 10% | Gauge contribution <10% of tolerance on all critical features | |
| Cleanliness and residue control | 10% | Lot-specific extract scans, 18.2 MΩ·cm final rinse, controlled packaging area | |
| CAPA effectiveness | 10% | All CAPAs closed with verification, no open items >90 days |
Multiply each score by its weight, sum, and you get a number out of 5. My decision rule: 4.0 or above is approved for critical applications, 3.0 to 3.9 is approved with a corrective action plan and a 90-day re-audit, below 3.0 is not approved regardless of price. The number matters less than the discipline of scoring every vendor the same way and keeping the scores comparable across years. A vendor that dropped from 4.2 to 3.4 between audits is a conversation; a vendor that has always been 3.2 is a habit.
Two field cases where the checklist caught what the CoA missed
The crimp seal that passed the dye test and failed the autosampler
A Thermo autosampler user reported intermittent “vial not detected” errors on about 1 in 400 vials, always on the same tray position. The vendor’s CoA was clean and their dye-leak test passed. We pulled the dimensional SPC data and found the neck OD chart had been running at the upper control limit for three weeks with no reaction, because the vendor’s trend rule only fired on 7 consecutive points on one side of the mean, and the drift was oscillating. The measurement: neck OD was 0.09 mm over nominal on the affected lots, enough to change the gripper’s grip height by a fraction that the autosampler’s optical sensor read as a miss. The fix was a vendor-side reaction plan that triggers on a 3-point trend, plus a buyer-side incoming check on neck OD with a go/no-go gauge at the tray. The audit finding that predicted it was the missing reaction plan, not the dimension itself.
The septum that cored on the third injection
A contract lab running 96-well headspace work started seeing a silicone fragment in the injection port after about 200 injections. The septum vendor’s qualification data showed 8 punctures before coring. We asked for lot-specific puncture data and got a blank look. The audit item that caught it was the “one-time qualification” question: the vendor had qualified the septum on a 2019 resin lot and never re-qualified after a resin supplier change in 2022. The new resin had a lower tear strength, and puncture retention dropped to 3. The fix was a re-qualification requirement tied to any resin change, written into the supplier agreement, plus a buyer-side puncture test on every incoming lot using a 22-gauge needle and a 5-puncture protocol. The vendor’s Restek-style reference data was fine; the problem was that the reference data wasn’t lot-specific.
What to verify on the day, and what to leave to the documents
I keep the on-site portion deliberately short and focused on the things that only make sense in person. The physical walk-through and the contract terms are their own exercises and I won’t rehash them here. What I want from the day itself is a live demonstration of three things: the SPC chart being read and reacted to in real time, the traceability system pulling a finished-goods lot back to its raw material lot in under four hours, and the gauge R&R being run on a feature the auditor picks, not one the vendor prepared. Those three demonstrations tell you more than a binder of certificates. If the traceability exercise takes a day, the system is paper.
For the packaging side, ask to see how vials are handled between the wash cycle and the bag. A reagent bottle line running in the same facility is a useful cross-check on whether the cleanroom discipline is real or staged for the vial line only. If the bottle line is casual about particulate control, the vial line probably is too.
Recertify on a schedule tied to risk, not to the calendar
The audit isn’t a one-time event. I recertify every vendor on a cycle that depends on their score and their change history. A vendor at 4.5 with no changes in 24 months gets a desk review annually and a full audit every three years. A vendor at 3.5, or one that has changed resin, mold, or glass source in the last 12 months, gets a full audit annually and a targeted re-audit 90 days after any change. The trigger for an out-of-cycle audit is any of these: a lot rejection rate above 0.5% on incoming inspection, a field complaint traced to a vial, a raw material supplier change, or a CAPA that closes without an effectiveness check. That last one is the one most buyers miss, and it’s the one that predicts the next failure.
The decision rule is simple. Score every vial vendor on the same weighted matrix, require the evidence package before you visit, and treat a change in the vendor’s process as a reason to re-audit regardless of the calendar. A vendor who can’t show you a control chart with an active reaction plan and lot-specific septum data is a vendor you’re managing by hope. The matrix turns that hope into a number you can defend to procurement, and the number is the only thing that survives the next price negotiation.
References & further reading
Related resources from our lab consumables network:
- Vial at vials-filters.com
- Headspace Vial at headspacevialsupplier.com
- Gc Vial at gcvialfactory.com
- Reagent Bottle at reagentbottleforsale.com
External references and standards: