TL;DR: A COA that lists only nominal values without process capability data (Cpk) is a compliance document, not a quality signal — treat it accordingly.
TL;DR: In our incoming inspection protocol, any lot where dimensional deviation exceeds ±0.3mm on a critical registration feature triggers a Category B hold, blocking production entry until a minimum 200-piece re-sample is cleared.
When the COA Doesn’t Match What Arrives on the Dock #
A brand we work with in personal care had approved a supplier based on three consecutive clean COAs. Registration tolerance: ±0.25mm. Colour delta E: ≤1.5. Burst strength: ≥800 kPa. Every document passed. When the first production lot of 120,000 folding cartons arrived, our incoming inspection team flagged register drift on the front panel eye mark — averaging 0.41mm off centre across 15 sampled units. The supplier’s COA, issued the same week, showed 0.18mm. No contradiction was noted on their end.
That gap between document and reality is not unusual. It happens because many suppliers measure from a single press sheet pulled at job start, before the press thermalises and substrate tension stabilises. By the time a 10,000-sheet run is complete, cumulative mechanical drift can push actual registration well outside the approved window. If your inline camera system is calibrated to flag at 0.3mm, you will catch it. If you are relying only on supplier COAs, you will not.
The root cause in that case was that the supplier’s press lacked closed-loop register control. Their measurement was accurate for sheet one. Sheets 4,000 through 10,000 were a different story. We retained that supplier but required them to submit CPK data on registration derived from a minimum 30 mid-run samples per job — not pull-sheet measurements. That one requirement change eliminated the recurrence across the next six shipments.
The Parameters That Actually Predict Incoming Lot Conformance #
When we qualify a new supplier for folding carton, rigid box, or flexible packaging components, we look at five parameters in their COA and incoming verification data before anything else.
Process capability index (Cpk) on critical dimensions. A Cpk of 1.33 or above on registration, caliper, and burst strength means the process is running at ±4-sigma or better against the tolerance band. A Cpk below 1.0 means the process is producing out-of-spec units as a matter of statistical certainty. We require Cpk ≥ 1.33 on all tier-1 structural and print dimensions, referenced against ISO 3534-2 statistical process terminology.
Caliper consistency across the substrate sheet. For folding carton grades, we specify caliper CV (coefficient of variation) ≤ 2.0% across a minimum of 9 measurement points per sheet. At CV above 3.5%, fold crease depth becomes inconsistent and gluing nip pressure varies enough to cause open seam defects downstream. This holds for SBS, FBB, and coated duplex grades in the 230–400 gsm range. For kraft liner used in corrugated components, we reference TAPPI T411 for thickness uniformity procedure.
Colour delta E under standardised illumination. We specify ΔE ≤ 1.5 under D50/2° observer conditions against approved press proof, measured per ISO 13655. Values between 1.5 and 2.5 go to a brand-approval hold; above 2.5 we reject without escalation. Suppliers who submit COAs with ΔE values but no stated illuminant or observer conditions are presenting unverifiable data — we flag this immediately in our SQA-F04 supplier review form.
Seal strength or burst strength, not just peel force. For secondary cartons, we verify burst strength per ASTM D2929 at ≥700 kPa for standard pharma-adjacent applications, ≥850 kPa for tamper-evidence constructions. Peel force data alone is insufficient — a bond can peel smoothly at acceptable force values while still failing cohesive burst under transport vibration.
Eye mark placement tolerance and reflectance delta. For any substrate going through our automated finishing lines, eye mark centre tolerance must be within ±0.5mm of nominal position, and reflectance delta (mark vs. background) must exceed 60 PRS units under 950nm NIR sensing. Suppliers who specify eye marks only in artwork files without dimensional callouts on the COA have never run the substrate through a servo-controlled converting line — which is itself a qualification signal.
The parameter most commonly omitted from supplier COAs is Cpk. Suppliers provide average values because averages look clean. Cpk exposes variability, and variability is what your inline systems will actually see.
| Parameter | Minimum Acceptable | Rejection Threshold | Measurement Reference |
|---|---|---|---|
| Registration (Cpk) | ≥ 1.33 | < 1.0 (auto-reject) | ISO 3534-2 |
| Caliper CV | ≤ 2.0% | > 3.5% | TAPPI T411 |
| Colour ΔE (D50/2°) | ≤ 1.5 | > 2.5 | ISO 13655 |
| Burst Strength | ≥ 700 kPa | < 600 kPa | ASTM D2929 |
| Eye Mark Reflectance Delta | ≥ 60 PRS | < 45 PRS | Internal spec EMS-09 |
Decision Framework — What to Do When Data Is Incomplete or Borderline #
If the incoming COA is complete with Cpk data and all values sit above minimums, standard AQL 2.5 incoming sampling applies — typically 32 units from a lot of 1,200–3,200 pieces per ISO 2859-1 Level II. Production entry is cleared within 4 business hours of dock receipt.
If the COA has nominal values but no Cpk, the approach changes. We pull a 50-piece stratified sample (first, middle, last of the pallet stack) and measure registration and caliper in-house before the lot touches our converting equipment. This adds 6–8 hours to incoming processing. If that sample returns Cpk ≥ 1.25, we accept the lot under a conditional pass and issue the supplier a corrective action requesting Cpk inclusion in all future COAs. If Cpk comes back below 1.0, the lot goes on Category B hold regardless of nominal compliance.
If burst strength or seal integrity data is missing entirely, we do not pass the lot. This is a hard stop in our incoming protocol. A supplier that cannot provide validated bond data either lacks the measurement capability or is withholding a known failure. Both scenarios disqualify the shipment from production entry. The cost of a delaminating secondary carton reaching retail — particularly in pharma-adjacent or cosmetics categories — outweighs any production schedule pressure.
For new supplier qualification runs, where we expect some gaps, we specify a 500-piece pilot lot before any volume purchase order is raised. The pilot goes through full inline inspection including 100% camera scan for colour, register, and structural integrity, with results logged against our SQA-F04 baseline record. If the pilot passes with fewer than 0.4% defect incidence at AQL 2.5, the supplier moves to standard incoming protocol for production volumes. If defect incidence is between 0.4–1.2%, we allow a second pilot with corrective action evidence. Above 1.2% on the pilot, we do not proceed to volume — not because the numbers are automatically unrecoverable, but because a supplier who cannot hold tolerance on a 500-piece qualification run has not demonstrated process control adequate for 50,000-piece production orders.
One non-obvious recommendation: require your supplier to send the COA before the shipment departs, not bundled with the pallet on arrival. A COA received 48 hours before dock receipt gives you time to identify data gaps and request supplementary measurement before the shipment is in your facility consuming floor space under hold status. Most incoming quality disputes are logistical problems before they are technical ones.
Specification Notes for Brand Partners #
When you brief us on a new component requiring supplier qualification, the minimum we need is: approved artwork with registration targets and eye mark position callouts, the substrate specification (grade, gsm, caliper range), intended converting process (die-cut only, fold-glue, lamination, or combination), and the end-use environment if relevant (cold chain, high-humidity retail, export shipping with ISTA-2A requirement).
The gap we see most often in incoming briefs is missing caliper tolerance. Brands specify gsm but leave caliper open. On a 350 gsm FBB, caliper can range from 390 to 430 microns depending on the mill and coating weight — a 40-micron spread that affects both crease depth settings on our die-cutter and the gluing nip gap on our folder-gluer. Specifying caliper range (e.g., 400–420 microns) in your material brief prevents the first sample iteration being a mechanical adjustment, not a quality failure.
Our standard qualification timeline for a new supplier with a pilot lot is 15–18 working days from sample receipt to production clearance decision, assuming no Category B holds are triggered. If the pilot lot requires re-sampling, add 7–10 working days. Brands on a launch deadline should build a minimum 25-working-day buffer for first-time supplier onboarding.
Does a supplier need ISO 9001 certification to pass your qualification process?
ISO 9001 is a management system standard, not a process capability standard. We’ve qualified suppliers without ISO 9001 who had excellent Cpk data and consistent incoming results, and we’ve received defective lots from ISO 9001-certified suppliers whose COAs were generated from non-representative samples. Certification is a starting point for the paper review, not a substitute for incoming data.
What if our supplier refuses to provide Cpk data?
That response tells you something. A capable supplier running a stable process has no reason to withhold process capability data — it’s a selling point, not a liability. If a supplier declines to provide Cpk on the basis that it’s proprietary or “not standard practice,” you can request raw measurement data from 30 mid-run samples instead and calculate Cpk yourself. If they decline that too, factor it into your risk weighting accordingly.
Our supplier’s COA shows ΔE = 1.2, but colours looked off on the pilot. Which is right?
It depends on which illuminant was used for the measurement. ΔE of 1.2 under D65 and ΔE of 1.2 under D50 are not the same visual result — D65 is a cooler white point and can mask warm-tone drift that D50 exposes. We specify D50/2° because that’s the ISO 13655 standard for graphic arts, and most brand colour references are built in D50. Ask your supplier which illuminant their spectrophotometer was set to. If they don’t know, the COA value is not verifiable.
How many units do you inspect on a standard incoming lot?
For lots between 1,200 and 3,200 pieces, our standard is a 32-unit AQL 2.5 Level II sample per ISO 2859-1. For lots above 10,000 pieces, that rises to 80 units minimum. On any first shipment from a new supplier — regardless of lot size — we pull 50 units stratified across the pallet, not random-selected from the top layer.
Planning a packaging project? Contact our team to request a complimentary specification review and sample quote.
The mid-run drift point hits close — we had a 75,000-unit whisky giftbox run where the supplier’s COA showed registration at 0.19mm, perfectly within spec, and our QC team cleared the lot on document review alone. Caught it at the filling line when the foil stamp on the front panel was walking 0.5mm toward the deboss edge by unit count somewhere past sheet 6,000, visible clustering when we pulled 40 pieces across the run. No closed-loop control on their Heidelberg, same story. We now require mid-run Cpk pulls at sheet 3,000 and 7,000 minimum — single pull-sheet COAs don’t tell you anything about a press that’s still thermalising at job start.
Ran into almost the exact same situation with a Shenzhen litho supplier running 85gsm folding boxboard for a fragrance line — their COA was clean, registration at 0.19mm, but our camera system kept flagging the eye marks on sheets pulled after the 6,000-sheet mark. Turned out they were sampling at press start before thermal stabilisation, same story. We didn’t drop them but mandated mid-run Cpk sampling at intervals of roughly 3,000 sheets, and the drift problem disappeared within two shipment cycles.
Switching to the 30 mid-run sample requirement fixed a similar issue for us, but the part nobody talks about is the qualification lag — getting a Vietnamese corrugated tray supplier to actually implement mid-run CPK pulls added about 11 weeks to our onboarding timeline before we’d trust a full production release for our wet food SKUs.
The Cpk ≥ 1.33 threshold for registration is achievable on flatbed, but we found that on rotary die-cut lines running 350gsm GC2 board for rigid watch box sleeves, the substrate spring-back after the cutting station introduces enough positional variance that even a capable press will fail that target at the inspection gate unless you’re measuring post-die, not post-print. Took us three rejected lots from a Guangdong supplier before we isolated where in the process the measurement reference point was actually sitting.
Seal integrity issue that still annoys me — we ran a 60,000-unit SKU of stand-up pouches for a jerky treat line, 3-ply PET/foil/LLDPE laminate, and our supplier’s COA showed burst strength consistently at 840-860 kPa, well above the 800 kPa floor. We didn’t catch until a retailer flagged it that the bottom gusset seals were failing in transit at roughly 8% of units — the issue traced back to the LLDPE sealant layer running at the low end of its thickness spec (68 micron instead of the nominal 75), which the COA never broke out separately. No Cpk on the sealant layer caliper, just a single composite burst figure that masked the variation entirely.
On the 120,000-unit folding carton lot — what substrate were they running, and did the register drift correlate with caliper variation across the web, or was it purely a mechanical tension issue on the press?