Start with the measurement system, not the certificate

A 40-lot shipment of 2 mL amber vials arrived with certificates of analysis showing septum thickness at 3.05 mm against a 3.00 ± 0.05 mm spec. Twelve lots passed on the incoming gauge. Twenty-eight lots measured 3.09 to 3.12 mm on our own drop indicator, and the vendor’s QC bench later admitted they were using a caliper with 0.02 mm resolution and a single operator. That is the moment a chromatography supplier supplier audit checklist stops being a filing exercise and becomes a purchasing tool. The supplier had a certificate. The certificate was real. The measurement behind it was not defensible. This article is about which quality-system items predict that failure before you sign, what evidence to request in advance, and how to score vendors so the decision is reproducible across buyers.

Every chromatography supplier supplier audit checklist I have used starts with the same trap: the COA looks clean, so the buyer stops reading. The COA is an output. The measurement system that produced it is the thing you are auditing. Ask three questions before you look at a single specification.

• What gauge, what resolution, what operator? A 0.10 mm tolerance on a septum or a 0.05 mm tolerance on a crimp neck demands a measurement instrument with at least 4:1 resolution against the tolerance. A 0.02 mm caliper on a 0.05 mm tolerance is 2.5:1 and fails the rule of ten. Ask for the gauge make, model, calibration interval, and the last three calibration certificates.
• Is the characteristic measured or derived? Septum thickness can be measured with a drop indicator or derived from a molded-part drawing. Bore diameter on a PTFE-lined silicone septum is usually measured with pin gauges. Glass ID on a 300 µL insert is measured optically or with a bore gauge. If the COA does not say which, the number is unverified.
• What is the Cpk on each critical characteristic, and over how many lots? A Cpk of 1.33 on a single lot tells you almost nothing. Ask for a 12-month capability study on the three characteristics that matter for your method: neck OD, septum thickness, and bore concentricity. If they cannot produce it, they are inspecting, not controlling.

The reason this matters more than the specification itself is that a chromatography supplier’s quality system either catches drift or it does not. A vendor running 100% incoming glass inspection with a 0.01 mm optical comparator will catch a 0.04 mm neck OD shift inside a shift. A vendor sampling 5 pieces per lot from a 10,000-piece run will not. The audit question is not “do you inspect?” It is “what is your sampling plan, what is your AQL, and what happens to the lot when it fails?”

Request the evidence pack before the site visit

You will get more from a document request than from a day of walking a plant. The evidence pack below is what I ask for 30 days before any audit, and I score the response as part of the audit itself. A vendor that cannot assemble this in 30 days has a document-control problem, and document control predicts lot-to-lot consistency better than almost anything else.

Two of these items do most of the work. The PCN procedure is the single best predictor of a surprise. If a vendor changes silicone resin supplier without telling you, your GC-MS baseline will drift and you will spend three weeks blaming the column. The complaint log is the second. A vendor with 40 complaints in 24 months and 40 closed root causes is safer than a vendor with 6 complaints and no closure records, because the second one is not logging them.

Score the vendor on a weighted matrix, not a gut call

The audit needs a number at the end, and the number needs to be defensible when two buyers disagree. I use a 100-point matrix with fixed weights. The weights are opinionated because the failure modes are asymmetric: a dimensional miss costs you reruns, a material change costs you a method revalidation.

Score each category 0 to 5, multiply by the weight, divide by 5. Anything below 70 is a conditional approval with a corrective action plan and a 90-day re-audit. Below 55 is a no. Above 85 with a clean 24-month complaint trend is a candidate for a sole-source position on a critical SKU, which is a decision you make once and revisit annually.

Two field cases where the audit predicted the failure

Case one: the septum that passed the COA and failed the needle

A contract lab running 500 GC injections a day on autosamplers started seeing coring at 40 to 50 punctures on a lot of 11 mm septa rated for 100. The symptom was a plug of silicone visible in the injection port liner after every 200 injections. The measurement that identified it was a durometer check: the lot measured 48 Shore A against a 45 ± 3 spec, within tolerance, but the compression set after 24 hours at 200 °C was 38% against a 25% maximum. The audit had flagged the vendor’s compression-set test as derived, not measured, and the vendor had been calculating it from a formula rather than running the ASTM D395 method. The fix was a supplier change and a new audit line item: compression set must be measured per ASTM D395 Method B, with the raw data attached. The vial closures vendor you qualify should be able to produce that data on request, not on a six-week lead time.

Case two: the neck OD drift that broke the crimper

A pharma QC group was getting 2 to 3% seal failures on a 20 mm crimp finish, detected as peak-area RSD above 3% on an internal standard. The measurement that identified it was a go/no-go gauge on 30 vials per lot: neck OD measured 20.12 to 20.18 mm against a 20.00 ± 0.10 mm spec. The lot was in spec. The problem was the crimper was set for a nominal 20.05 mm and the torque was fixed at 1.2 N·m, so the higher-OD vials deformed the cap skirt instead of compressing the septum. The audit finding was that the vendor’s Cpk on neck OD was 0.9, meaning the process was capable of producing out-of-spec parts on any given day. The fix was a vendor change to one holding Cpk ≥ 1.67 and a crimper torque recalibration to 1.0 to 1.1 N·m with a 20 mm crimp setting verified against a go/no-go gauge every shift. If you buy glass through an OEM vial channel, ask for the capability study on the neck finish, not just the dimensional drawing.

What the quality system says about the material, not just the part

Dimensional audits catch drift. Material audits catch revalidation. The two are separate questions and most checklists conflate them. For septa, the material questions are silicone formulation, post-cure schedule, and extractables profile. For glass, they are hydrolytic class (Type I borosilicate for most HPLC work), coefficient of expansion, and the annealing schedule. For closures, they are cap resin (polypropylene homopolymer vs copolymer) and liner bonding method.

The audit item that catches the most material problems is the post-cure schedule. Silicone septa that are not fully post-cured will outgas at injection port temperatures and produce siloxane peaks in GC-MS at m/z 73, 147, 207, and 281. A vendor running a 4-hour post-cure at 200 °C will have a clean background. A vendor running a 2-hour post-cure to hit a shipment date will not, and the COA will not show it. Ask for the post-cure temperature, dwell time, and the last three background scans at your injection volume. If they cannot produce a background scan, they have not characterized the part for your application. The Restek technical library has published background data you can use as a benchmark, and the Agilent GC consumables documentation lists bleed specifications for comparable parts. Use both as reference points when you evaluate a vendor’s data.

Supply continuity is a quality-system item, not a commercial one

I treat supply continuity as a quality question because a vendor that cannot ship forces you into substitute lots, and substitute lots are where dimensional and material variation enters your method. The audit questions are specific: how many weeks of safety stock on the top 20 SKUs, how many qualified sources for the glass tubing and the silicone resin, and what happened during the last shortage.

Ask for the last shortage narrative in writing. A vendor that says “we had no impact” is either lying or not tracking. A vendor that says “we qualified a second glass source in 2022, ran a 3-lot bridging study, and issued a PCN 120 days before the switch” has a continuity system. The bridging study is the evidence that matters: 3 lots, 30 vials per lot, dimensional and extractables comparison against the incumbent. If the vendor cannot produce a bridging study, the second source is not qualified, it is just a name on a list.

The other continuity item is the minimum order quantity and the mold or tooling ownership. If the vendor owns the mold for your custom vial, you have a single source by default. Ask who owns the tooling and whether you can transfer it. A vendor that owns the tooling and refuses transfer is telling you the switching cost is the business model. That is a commercial risk that belongs in the quality score because it determines whether you can act on a failed audit.

Run the audit as a scored document review with a short verification visit

The site visit is a verification step, not the audit. By the time you walk in, the evidence pack should already be scored and the visit should be confirming three things: the gauge on the bench matches the gauge on the certificate, the sampling plan on the floor matches the written procedure, and the complaint log in the QA office matches the one you were sent. If any of those three diverge, the audit score drops by 15 points regardless of everything else, because it means the documents are curated for auditors.

Bring a calibrated gauge with you. Measure 10 pieces from a production lot against the drawing. I have seen a 0.06 mm neck OD discrepancy between the vendor’s bench gauge and a certified drop indicator on the same lot. That is not a rounding error, it is a measurement system failure, and it is invisible in the evidence pack. The same logic applies to torque: bring a calibrated torque wrench and check the crimper setting against the SOP. A crimper running at 1.4 N·m against a 1.1 N·m SOP is a seal-failure generator, and the operator will often tell you it has “always been set that way.”

Decision rule

Qualify a chromatography supplier when the evidence pack shows Cpk ≥ 1.33 on the three characteristics that touch your method, a written PCN process with 90-day notice, a closed-loop complaint log with root causes, and a bridging study for every second source they claim. Score the matrix, set the threshold at 70 for conditional approval and 85 for sole source, and re-audit annually or on any material change. If a vendor cannot produce the capability data and the PCN record before the visit, the visit will not change the answer. The audit checklist is a filter, and the filter works best when it runs on documents first.

For buyers assembling the full vendor list across glass, closures, and consumables, the Chromatography Supplier directory and the method-validation resources at Analytical Testing are useful starting points, and the Thermo Fisher chromatography and sample handling documentation gives you a benchmark for what a complete consumables specification should contain. The EPA SW-846 method set is worth a read for the container and handling requirements it imposes on environmental work, because those requirements become audit criteria when your lab runs those methods.

References & further reading

Related resources from our lab consumables network:

• Chromatography Supplier at chromatographysupplier.com
• Oem Vial at oemsamplevials.com
• Vial Closures at vialclosures.com
• Analytical Testing at labanalyticaltesting.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