A lot of 2000 septa arrived with a certificate that said “puncture tested, 100 cycles.” Six weeks later a client running EPA 8260 on a Thermo TRACE 1310 saw siloxane peaks at m/z 73 and 207 appear above the 10 µg/L calibration point by run 40, and the 9 mm blue screw caps had already started coring at the 60-puncture mark. The vendor’s own QC sheet, when we finally asked for the raw data instead of the summary, showed testing on a different durometer batch. That is the kind of gap a chromatography products supplier audit checklist is supposed to catch before the PO goes out, not after the method fails.
I have run these audits from both sides of the table. The checklist that actually predicts failure is short, boring, and almost never included in the glossy vendor questionnaire. It is not about ISO certificates on the wall. It is about whether the vendor can produce, on request and within 48 hours, the specific evidence that maps to the failure modes you will hit on your instrument. Below is the version I use, with the scoring model attached.
Most supplier questionnaires ask whether the vendor is ISO 9001 certified. That tells you they have a document control system. It does not tell you whether their incoming resin inspection catches a durometer shift, whether their mold cavity numbers are traceable to a lot, or whether they keep retain samples from every batch. Those three things do.
Incoming material traceability down to the resin lot
For septa, ferrules, and vial caps, the failure mode almost always starts at the raw material. A silicone septum that reads 45 Shore A instead of 50 cores differently, and a PTFE-faced silicone septum with a thin face will punch through at 40 cycles instead of 100. Ask the vendor to show the incoming inspection record for the resin or elastomer lot tied to a specific finished-goods lot number. If they cannot link finished lot to raw lot in under an hour, that is a red flag. I have seen vendors produce a “material cert” that was actually a supplier’s generic datasheet, not a lot-specific CoA. Those are not the same document.
The same applies to borosilicate glass for vials. A 2 mL autosampler vial that reads 11.6 mm OD instead of 12.0 mm will jam in a tray or fail to seat in the crimper. Ask for the dimensional inspection record on the glass lot, not the drawing tolerance. The drawing says ±0.2 mm. The inspection record tells you whether they measured it.
Mold and cavity traceability
Injection-molded parts (caps, septa, vial bodies) come off multi-cavity tools. Cavity-to-cavity variation is real. A 24-cavity cap mold can produce a 0.05 mm spread in skirt ID across cavities, and if the vendor does not track which cavity a lot came from, you cannot investigate a fit problem when it appears. Ask for the cavity map and the setup sheet for the lot. If they look at you blankly, they are a trader, not a manufacturer, and you should price accordingly.
Retain samples and stability data
Every serious consumables manufacturer keeps retain samples from each lot for at least 24 months. Ask how many they keep, where they are stored, and under what conditions. Silicone septa and PTFE-lined caps have a shelf life, and a lot that sat in a hot warehouse for eight months before shipping will behave differently than one that shipped fresh. If the vendor cannot produce a retain sample for a lot you bought 14 months ago, you have no recourse when a method drifts.
Change control notification
This is the single most underweighted item on any chromatography products supplier audit checklist. Ask in writing: will you notify us before you change resin supplier, mold, or manufacturing site? Get the notification window in days. A vendor that swaps a septum compound to save 4 cents a piece without telling you has just invalidated your method validation. I require a 90-day advance written notice and a re-qualification sample before any change ships. Vendors who resist this are telling you something.
The evidence to request before the first PO, not after
Ask for these documents up front and read them. Not the summary page. The raw data behind it. A vendor who can produce all six within a week is a vendor you can work with.
- Lot-specific CoA with measured values, not pass/fail. For septa: durometer (Shore A), thickness in mm, and puncture count from an actual test. For vials: OD, ID, height, and neck finish dims with the measurement method named.
- Extractable/leachable data on the septum compound. At minimum, a GC-MS scan of a blank injection after 24-hour solvent contact at 40 °C. You are looking for siloxane, sulfur, and phthalate peaks.
- Puncture test protocol with needle gauge, septum temperature, and the pass criterion. A “100 punctures” claim is meaningless without the needle size and whether the test ran at ambient or 250 °C.
- Dimensional capability data (Cpk) on the critical dimensions. A Cpk below 1.33 on a 12 mm vial OD means you will see fit problems on a fraction of the lot.
- Temperature rating evidence for septa and ferrules. A “350 °C rated” septum claim should come with a thermogravimetric trace or a bleed test at that temperature, not a marketing line.
- Reference customer list with contact permission for at least two accounts running similar methods. Call them. Ask what happens when a lot is out of spec.
For the vial side specifically, I also ask for the crimp or decap torque data. A 11 mm aluminum crimp cap should seal at 1.5 to 2.0 in-lb on a standard crimper; if the vendor’s cap skirt is 0.1 mm short, you will chase leaks you cannot see. The vials with caps catalog is a useful reference for what the dimensional callouts should look like on a proper spec sheet.
How to score a vendor without arguing about it
Weighted scoring beats a gut call, and it gives you a defensible number when procurement pushes back on price. I use a 100-point model split across four buckets. The weights are not arbitrary; they track how often each item has actually caused a failure in my experience.
| Bucket | Weight | What it measures | Pass threshold |
|---|---|---|---|
| Material and lot traceability | 30 | Raw-to-finished lot linkage, retain samples, cavity maps | ≥ 24/30 |
| Test evidence quality | 30 | Lot-specific CoA with measured values, extractables data, Cpk on critical dims | ≥ 24/30 |
| Change control and responsiveness | 25 | 90-day notification, 48-hour document turnaround, corrective action history | ≥ 20/25 |
| Commercial and logistics | 15 | Lead time, MOQ, landed cost, packaging integrity | ≥ 10/15 |
A vendor scoring below 68 is a no. Between 68 and 80, conditional approval with a first-article inspection on the first three lots. Above 80, standard approval. The commercial bucket is deliberately the smallest. A vendor that wins on price and loses on change control costs you more in failed runs than the discount ever saves.
Score the buckets independently, then look at the spread. A vendor at 85 overall but 12/30 on traceability is a vendor that will fail you on an out-of-spec lot. I would rather approve a 78 with balanced scores than an 85 with a hole in it.
Field case one: the septum that passed the CoA and failed the method
Symptom: a GC-MS client running EPA 8270 saw baseline rise and siloxane peaks at m/z 73, 147, 207, and 281 appear after 30 injections from a fresh box of 11 mm septa. The CoA said “bleed: pass.” The fix came from asking for the raw bleed chromatogram, which showed the test ran at 280 °C for 30 minutes, not the 320 °C the client’s inlet ran at. The septum compound was rated to 300 °C. The vendor had tested to a lower temperature and reported a pass.
The measurement that identified it: a thermogravimetric run on the septum showing 1.8% mass loss between 300 and 320 °C. The fix was a switch to a higher-temperature compound and a revised spec that required a bleed test at the client’s actual inlet temperature, with a pass criterion of no peak above 0.5% of the internal standard area at 30 minutes. That criterion now sits in the audit checklist as a scored item.
Field case two: the 12 mm vial that was not 12 mm
A different client, an HPLC autosampler running a 1260 Infinity II, started getting intermittent “vial not detected” errors on one tray position. They replaced the tray, the gripper, and eventually the autosampler. The problem was the vials. A spot check with calipers on 30 vials from the lot showed OD ranging from 11.72 to 12.04 mm, against a spec of 12.00 ± 0.15. The vendor’s inspection record showed they were measuring with a go/no-go gauge, not a micrometer, so the out-of-spec parts passed.
The fix was a supplier change and a new audit requirement: dimensional inspection with a calibrated micrometer, Cpk reported on the lot, and a ±0.10 mm tolerance on OD instead of ±0.15. That last number matters because the autosampler gripper on that instrument has about 0.2 mm of clearance, and anything over 12.10 mm binds. The vial supplier comparison pages are worth reading before you set your own tolerance, because the callouts vary more than you would expect between vendors.
What to ask on the phone that never makes it onto the form
The written questionnaire gets you the documents. The phone call gets you the truth. Ask these four questions and listen for hesitation.
- When was your last major nonconformance, what was it, and what did you change? A vendor who says “we have never had one” is either new or lying. A vendor who describes a resin change and the corrective action is telling you their system works.
- Who signs the CoA, and what is their authority to reject a lot? If the answer is “the sales manager,” walk away.
- What is your current lead time on a custom septum, and what is the longest it has been in the last 12 months? A 4-week quote with a 14-week historical max tells you the quote is fiction.
- Can you ship a first-article sample from the exact lot I will receive? If not, the sample you test is not the product you buy.
I also ask who their own suppliers are for resin and glass, and whether they have a second source qualified. Single-source vendors on a critical resin are a supply risk you are absorbing. For finished chromatography products, the chromatography products range on this site is organized so you can cross-check dimensional and material callouts across categories before you write the spec.
The audit items that map to specific instrument failure modes
Generic quality-system questions miss the instrument-specific failure modes. A checklist that predicts problems is built backwards from the failure. Here is the mapping I use.
| Instrument failure mode | Audit evidence to request | Pass criterion |
|---|---|---|
| Septum coring / plug in inlet | Puncture test at operating temp, needle gauge named | ≥ 100 cycles, no core, at 250 °C |
| Siloxane bleed / baseline rise | Bleed chromatogram at client inlet temp | No peak > 0.5% ISTD at 30 min |
| Vial fit / autosampler error | Dimensional inspection with micrometer, Cpk | OD 12.00 ± 0.10 mm, Cpk ≥ 1.33 |
| Crimp cap leak | Crimp torque test, skirt height data | Seal at 1.5–2.0 in-lb, no leak at 1.5 bar |
| Ferrule leak / ghost peak | Ferrule compression test, graphite vs. Vespel data | No leak at 100 psi He, 3 re-torques |
| Column phase drift | Lot-to-lot retention time and selectivity data | k’ within ±3%, α within ±2% |
Every row on that table is a question you can ask a vendor and a document they can either produce or not. That is the whole point. The audit is not a relationship exercise. It is a document request with a pass criterion attached.
Scoring the vendor’s corrective action history
One item deserves its own section because it separates real manufacturers from resellers: the corrective action record. Ask for a redacted list of the last 12 months of nonconformances and the CAPA (corrective and preventive action) closure for each. You are looking for three things. First, whether they track nonconformances at all. Second, whether the root cause is specific (“resin lot durometer out of spec, incoming inspection tightened to 100% durometer check”) or vague (“operator error, retrained”). Third, whether the fix was verified with data.
A vendor with zero nonconformances in 12 months is not a good vendor. They are a vendor with no detection. A vendor with five nonconformances and five documented, data-verified fixes is a vendor whose system catches problems before they reach you. I score CAPA quality heavily because it is the best single predictor of whether the next bad lot gets caught at their dock or at yours.
The decision rule
Approve a chromatography products supplier only when they can produce, within 48 hours and on request, a lot-specific CoA with measured values, a raw-to-finished lot traceability record, retain samples for any lot shipped in the last 24 months, and a written 90-day change-notification commitment. Score the four buckets, require a minimum of 24/30 on traceability and 24/30 on test evidence regardless of total, and run a first-article inspection on the first three lots from any new vendor. If any one of those four documents is missing, the vendor is a trader with a certificate, and the certificate will not save your method when the lot drifts.
For the underlying material and method references behind these criteria, the U.S. EPA SW-846 test methods set the container and holding-time expectations that most chromatography consumables audits trace back to, and the technical resources at Restek and Agilent give you the dimensional and bleed callouts to put in your spec. Cross-check any vendor’s claims against the manufacturer data sheets from Thermo Fisher and the manufacturing standards published by Aijiren before you sign the approval. The audit is only as good as the numbers you hold the vendor to.
References & further reading
Related resources from our lab consumables network:
- Chromatography Products at chromatographyselling.com
- Aijiren at aijirentech.com
- Vials With Caps at vialswithcaps.com
- Vial Supplier at vialsupplier.com
External references and standards: