CoA mismatch on a customer's incoming test: retest or escalate
Summary
A customer result that disagrees with your CoA can come from the method, the edition, the sampling or plain reproducibility; escalate only when the product itself is suspected. Check the method, edition and lot first. Then compare the gap with the method's reproducibility and retest the same retained sample. If the gap survives those checks, escalate to the manufacturer.
ASTM D3244-24 is the practice two parties use to resolve a dispute over specification conformance, and it applies only to test methods whose repeatability and reproducibility conform to its definitions (ASTM, 2024).
ISO 4259-2:2026, valid from 9 June 2026 and replacing the 2017 edition, specifies how to decide conformance when supplier and receiver results conflict (ISO 4259-2:2026).
ISO 4259-5:2023 sets a between-methods reproducibility for two different test methods that is expected to be exceeded with a probability of 5 % under correct operation (ISO 4259-5:2023).
ISO 3104:2023, the kinematic viscosity method, replaced ISO 3104:2020 on 7 November 2023, so a CoA and a customer report can cite different editions of one method (ISO 3104:2023).
This answer covers numeric properties of homogeneous liquid lubricants and petroleum products, tested by methods that publish their precision. ISO 4259-2:2026 does not apply to non-homogeneous products, and ASTM D6300-26 notes its precision approach may not suit greases (ASTM, 2026). Pharmaceutical components under US GMP rules are outside this answer.
What does a CoA mismatch on a customer's incoming test look like?
A CoA mismatch shows up as a customer lab report whose value for one property differs from the certificate of analysis (CoA) for the same lot, sometimes by enough to fall outside the specification. There are five possible causes: a different method, a different edition, a sampling or handling difference, normal reproducibility, or a real difference in the product.
The first four point away from the product as delivered, and you can usually close them in house. Only the fifth needs the manufacturer. The table below separates the five by what the documents already show, before anyone retests.
Check | What you see | Points to |
|---|---|---|
Method on both reports | Different method numbers, such as an ASTM and an ISO method | Method difference: retest by the CoA method, or use a between-methods limit |
Edition and specification | Same method, different year or different specification revision | Edition difference: align editions, retest if needed |
Lot and sample | Lot on the report differs from the CoA, or sample taken from a used drum | Sampling or lot mix-up: resample from the right lot |
Size of the gap | Gap smaller than the method's reproducibility R | Reproducibility: close in house with the calculation |
All above match | Gap larger than R on the same lot by the same method | Product suspect: retest retained sample, then escalate |
Only the last row leads to escalation. Every other row has a documented answer the distributor can give the customer. Take a viscosity result reported by an ISO method against a CoA value from an ASTM method. It belongs in the first row: the two numbers are not yet comparable, so the next step is a retest by the CoA method, not a call to the manufacturer.
Do the customer's test method and specification edition match the CoA?
The customer's method and edition match the CoA only when both reports name the same method number, the same edition year and the same specification revision. Settle this before anyone argues about the value, because results from different methods or editions are not directly comparable.
A usable CoA names the test method behind each result. The CASRAI guide to certificates of analysis lists a test method reference for each result as a required field, "so the receiving lab can judge whether the method is appropriate and comparable to its own" (CASRAI, 2026). If your CoA has no method reference, ask the manufacturer for one before you judge the gap.
Two different methods for the same property need their own limit. ISO 4259-5:2023 sets a between-methods reproducibility for two methods that purport to measure the same property, after one has been bias-corrected to agree with the other; the limit is expected to be exceeded with a probability of 5 % under correct operation (ISO 4259-5:2023). If no such study exists for your two methods, treat the comparison as indicative only.
Editions change as well. ISO 3104:2020 for kinematic viscosity was withdrawn on 7 November 2023 when ISO 3104:2023 took its place (ISO 3104:2023). Record the edition on each report. Then check which specification revision the customer tests against, since a tightened limit can turn the same value into a fail.
How large a CoA gap is within the test method's reproducibility?
A CoA gap is within the test method's reproducibility when the difference between the two single results is no larger than R, the method's published reproducibility. Two competent labs testing identical material can differ by that much, so a gap below R is not evidence of a product fault.
Formally, reproducibility is the expected difference between single results from different operators in different laboratories on identical material. In D02 methods for petroleum products and lubricants, repeatability and reproducibility are defined in the precision section of every test method (ASTM D6300-26). One condition monitoring source explains that most ASTM methods report these values at a 95 % confidence level, "meaning the values can only exceed 1 case in 20" (CBM Connect, viewed 2026).
R often changes with the level being measured. The same source gives elemental spectroscopy figures: copper at 10 ppm has a repeatability of 1.0 ppm and a reproducibility of 2.4 ppm (CBM Connect, viewed 2026). It names no method or edition, so take R from the precision statement of the method on your CoA. As an illustration, say a CoA shows 10 ppm copper and the customer reports 12 ppm by the same method. The 2 ppm gap sits inside that 2.4 ppm R.
Use the R for the edition both labs used, at the level you measured, and put the calculation in your reply. When the result sits close to a specification limit, ASTM D3244-24 sets the procedure for an acceptance limit agreed between the parties (ASTM, 2024).
Should the same sample be retested before anyone escalates a CoA mismatch?
Yes: retest the same sample before anyone escalates a CoA mismatch. A retest of the retained sample tells a lab error apart from a real lot difference, and a lot difference is the only cause the manufacturer can act on.
Where they exist, use three samples: the customer's sample from the delivery, a retained sample of the same lot, and if possible an unopened container from that lot. ASTM D4057-22 warns that a spot sample represents only one point in a tank or vessel and that a running or all-level sample represents only the column it was drawn from (ASTM, 2022). So a drum sampled from the bottom, or a container opened weeks earlier, may not show the lot as shipped.
A petroleum inspection firm's note to schedulers gives the order: request a retest first, and when time is short, resample while the original sample is retested (AmSpec, 2011). The same note cites ASTM D56 flash point below 140 °F with a repeatability of 2 °F and a reproducibility of 8 °F (AmSpec, 2011); check those values against the current edition before you use them.
Before any testing, agree the retest plan with the customer: which lab, which method and edition, and which sample. ASTM D3244-24 requires the acceptance limit to be set before testing starts, and it recommends a qualified statistician when the practice is applied (ASTM, 2024).
When does a CoA mismatch point to the product itself?
A CoA mismatch points to the product itself when a retest by the same method and edition, on a correctly drawn sample of the right lot, still gives a gap beyond the method's reproducibility. By then the distributor has ruled out every cause it can close in house, and the question belongs to the manufacturer.
Other signs make the case stronger: the same shift across several lots or several customers, a retained sample at the supplier that also fails, and a moved property that the formulation controls, such as viscosity or additive elements, rather than one that handling changes, such as water or particle count.
A further sign is a customer lab that can show its own proficiency. As a prerequisite for using results, ASTM D3244-24 requires each lab to show proficiency with an agreed check standard within plus or minus 0.7R of the accepted reference value, or a Z-score within plus or minus 2.0 in an agreed interlaboratory exchange program (ASTM, 2024).
The CASRAI guide treats a disagreement between a receiving lab's confirmatory test and a supplier's CoA as "itself the finding" in GMP practice, because it questions the supplier's testing reliability (CASRAI, 2026). Outside GMP the same logic holds for the decision: once a gap survives retest, record it as a finding, not a dispute. For the wider rules on when a distributor hands a technical issue to its supplier, see when a distributor should escalate a technical issue to the manufacturer.
Which facts should go to the manufacturer when a CoA mismatch is escalated?
When you escalate a CoA mismatch, send the manufacturer the facts that let its lab start where your checks stopped: the lot, both results, the method and edition behind each, the samples, and what you have already ruled out.
Product name, grade and lot number, with a photo of the container label.
The CoA value and the customer's value, each with its method number and edition.
The specification and revision the customer tested against.
The reproducibility R you used and the calculated gap.
The retest results: which sample, which lab, which date.
How each sample was taken, and whether a retained sample of the lot exists.
Anything else the customer reports, such as a process problem, marked as reported rather than observed.
Keep facts and opinions apart. Write "customer reports viscosity at 40 °C outside specification" (an illustrative wording) rather than "the lot is bad". Ask the manufacturer to test its own retained sample of the lot by the same method and edition, and to say which acceptance rule it applies.
Lubinpla's escalation decision works from your own CoA, TDS and case records, checks whether the doubt touches the product's formulation, lot or specification, and separates in-house handling from escalation in one cited document that you review before anyone contacts the manufacturer.
Frequently asked questions
Can a distributor issue its own CoA after retesting?
A distributor can issue its own test report for the sample it tested, naming its lab, method and edition. It should not change the manufacturer's CoA or replace it with results from another lot, since a CoA describes one lot as released by the party that tested it.
Who pays for the retest in a CoA dispute?
The supply contract decides who pays. Agree the cost rule before the retest, along with the lab, method and acceptance limit, so the result settles the question instead of starting a second dispute over who owes what.
What if the customer's lab will not share its method or edition?
Ask for both in writing, and explain that without them the result cannot be compared with the CoA. Until the customer supplies them, record the mismatch as unconfirmed and do not escalate it as a product fault.
Does a CoA mismatch on one property mean the whole lot is out of specification?
No. A confirmed gap concerns one property of one lot. Check the other properties on the CoA against the customer's report, and ask the manufacturer to retest only what the evidence questions.