A certificate of analysis arrives attached to a drum, a bag or a tanker. By the time it reaches the plant, the material has already been bought, shipped and received. The document is the only compact record of what was actually tested, on which batch, using which method. Reading it properly takes five minutes and prevents most of the arguments that follow a failed trial.
The headline assay is the easiest number to look at and the least useful on its own. A 99.2% caustic soda result means little if the moisture is out of family, the chloride is high for an agrochemical synthesis, or the COA is linked to a different lot than the bags on the floor. Read the identity and method columns first.
Start with lot identity, not the assay
A COA should carry a batch or lot number that matches the pack label, the invoice and the packing list. It should show the release date, the tested quantity or sample reference, and a signature or analyst code. If any of those are missing, the certificate is not a lot record — it is a specification sheet with a number typed into it.
The eight lines worth reading
- 01 Assay, with its basis. Dry basis, as-is and anhydrous basis are not interchangeable. A 99% result on a wet basis may look lower than a 99% result on a dry basis for the same material.
- 02 Moisture or water content, with the method. Karl Fischer and loss-on-drying answer different questions; the former measures water, the latter measures everything volatile under the test conditions.
- 03 Insoluble matter. Small numbers matter in filtration, catalyst beds and coating systems; a result near the limit deserves a conversation even if it passes.
- 04 Chlorides. A critical line for agrochemical, plating and pharmaceutical customers. Read the limit, not just the result.
- 05 Sulphates. Often overlooked, but relevant to solubility, turbidity and certain reaction pathways.
- 06 Iron. A colour and stability issue in glass, food-grade acid and phosphating baths. The method should be stated.
- 07 Heavy metals, especially lead. For food-grade phosphoric acid, the specification and method should be explicit; ICP-MS results are not comparable with a generic limit without the method noted.
- 08 Appearance and odour. The least quantitative line and the fastest way to catch a contaminated or degraded lot before it enters a process.
| Field | What it tells you | What to challenge |
|---|---|---|
| Lot / batch number | Links the test result to the material in the pack | A number that does not appear on the label or invoice |
| Specification columns | The limits the lot was released against | A result with no limit, or a limit that is not yours |
| Method | How the number was produced | A method missing for a trace-metal or moisture result |
| Basis | Wet, dry or anhydrous calculation | Assay compared with a different basis across suppliers |
| Release date & signature | Who released the lot and when | A date before manufacture or a generic QA stamp |
| Retention reference | Whether a sealed sample exists | No retention statement or archive reference |
Method matters as much as the number
Two laboratories can report the same assay with different methods and produce numbers that are not directly comparable. Titrimetry, potentiometric titration and gas chromatography answer related but distinct questions. The same is true of moisture: Karl Fischer coulometry, Karl Fischer volumetric titration and loss-on-drying can diverge on a hygroscopic salt. A serious COA states the method or references a standard method the buyer can look up.
Method also matters when you are qualifying a second source. If your current supplier reports assay on a dry basis and the new source reports as-is, the apparent difference may be nothing more than water. Normalise the numbers before you decide the new lot is off-spec.
“A certificate is only useful if a stranger can reconstruct what happened to the lot from the fields on the page.”
Release principle, Apex analytical bay
Compare against your specification, not the brochure
Product literature typically quotes a typical value or a grade minimum. That is useful for selecting a material. It is not the limit that governs a release. The limits on your purchase order and supply agreement are the ones that matter, and the COA should report against them.
If the COA only reproduces catalogue typicals, ask for the lot-specific values. If a result sits inside the catalogue range but outside your process window, the material is still not suitable for the application — even though it passed someone else’s release. Specification qualification and lot release are two separate decisions.
What to do when a number is out of family
A single result slightly outside your historical band is not automatically a reject, but it should never be accepted without a documented decision. Quarantine the pack, pull the retained sample, and ask the supplier to explain the result against the process history of that batch. If the result is confirmed, open a non-conformance and agree a disposition before the material is used.
- Check the sample ID and pack label before re-testing; a labelling error is more common than a sudden process shift.
- Re-test with the same method first, then a confirmatory method if the result is close to the limit.
- Ask for the previous five lots of the same grade to see whether the result is a trend or an outlier.
- Do not blend a suspect lot into a larger batch to dilute the problem — that removes the evidence and widens the impact.
The purpose of a COA is not to reassure a buyer at the moment of purchase. It is to give the plant a defensible record when something behaves unexpectedly six weeks later. Read it as a release document, and the rest of the qualification gets simpler.
This note is general technical guidance, not a substitute for the lot COA, current SDS or site EHS procedure.