An editorial resource on analytical testing documentation, written for trade and laboratory readers. Please confirm you are 21 or older to continue.
This page is an editorial and informational resource about certificates of analysis — what the document certifies, which analytical methods sit behind each line, how results are expressed, and what accreditation of the issuing laboratory means. It is written for professional and trade readers who evaluate supplier documentation as part of their work.
xPepShop sells nothing. It is a published editorial resource, not a shop and not a testing laboratory. Nothing here is an offer, and no order can be placed through this site. Certificates discussed on this page are issued by independent laboratories, and the descriptions are general explanations of common analytical practice — not a substitute for reading the specific report attached to a specific lot.
Nothing on this page is guidance on the selection, handling or end use of any material, and no claim of any kind is made or implied about what any material does. This resource is written for readers aged 21 and over.
The document itself: what it certifies, what it deliberately does not, and who signs it.
Chromatography, mass spectrometry, titration — which test answers which question.
Specification versus result, area percent, and why the lot number is the important field.
Nine things that separate a working report from a decorative one, and what to ask for instead.
Every serious supplier will tell you their material is tested. Far fewer will hand you the document that proves it, and fewer still will hand you the document that matches the exact lot in the box. That document is the certificate of analysis, usually shortened to COA, and learning to read one takes about ten minutes — after which you can tell a working report from a decorative one at a glance.
This guide covers what a COA actually certifies, which analytical method sits behind each line on it, how results are expressed and why the phrasing matters, what accreditation of the issuing laboratory does and does not guarantee, and the specific red flags worth walking away from.
A COA is a testing record for one specific batch. Not for a product line, not for a supplier, not for a category — for one production lot, identified by a lot or batch number, tested on a stated date, by a named laboratory, using stated methods.
That scope is the whole point, and it is where most misreadings begin. A COA is not a quality badge a company earns once and displays forever. It is a snapshot: this material, this lot, these tests, this date, these results. A supplier with an excellent COA on lot A has told you nothing at all about lot B. The useful question is therefore never "do you have a COA?" but "may I see the COA for the lot I am being sent?"
A complete certificate carries, at minimum: the material identifier, the lot or batch number, the date of manufacture and the date of analysis, each test performed, the method used for each test, the specification (the acceptance range), the result obtained, a pass or fail judgement per line, and the name and authorisation of whoever released it. Anything missing from that list is a question to ask, not a detail to skip past.
COAs come from one of two places: the manufacturer's own quality control laboratory, or an independent third-party laboratory. Both are legitimate, and they answer slightly different questions. An in-house COA tells you the producer's own QC found the batch within specification. A third-party COA tells you someone with no commercial interest in the result found the same thing. Where the two disagree, the independent report is the more informative one.
The credential worth looking for on the letterhead is ISO/IEC 17025, the international standard for the competence of testing and calibration laboratories. It is a meaningful accreditation: it covers method validation, equipment calibration and traceability, analyst competence, sample handling, and what the laboratory must do when a result falls outside specification.
Two caveats, because this is the credential most often oversold. First, it is scoped — a laboratory is accredited for particular methods on particular sample types, not for everything it can physically run, and that scope is published. Second, an accredited laboratory reports on the sample it received. If the sample was not representative of the lot, or was not drawn under a controlled procedure, the laboratory's accreditation does not repair that. Accreditation raises confidence in the measurement, not in the sampling.
Different lines on a COA answer genuinely different questions, and it is worth knowing which is which — because a report can look thorough while quietly omitting the one test that matters for your purpose.
| Test | The question it answers | Typical method |
|---|---|---|
| Identity | Is this material the one named on the label? | Mass spectrometry (LC-MS), sometimes with NMR or FT-IR |
| Purity | What proportion is the named material rather than related substances? | Reversed-phase HPLC with UV detection |
| Water content | How much residual moisture does the material carry? | Karl Fischer titration, or loss on drying |
| Residual solvents | Is anything left over from processing or purification? | Headspace gas chromatography (GC) |
| Heavy metals | Are trace metallic contaminants within limits? | ICP-MS |
| Appearance | Does the physical form match the written specification? | Visual inspection against a described standard |
Not every material warrants every row. What matters is that the tests present are the ones that make sense for the material, and that each result has a method stated beside it.
This is the single most common confusion. Identity asks whether the material is what it says it is. Purity asks how much of the sample is that material rather than something else. A sample can pass identity and still be badly impure; it can also be extremely pure and be the wrong material entirely. A certificate that reports one without the other is answering half the question — and a high purity figure with no identity test anywhere on the page deserves real suspicion.
Most purity figures come from high-performance liquid chromatography. The principle states in one sentence: the sample is pushed through a column that makes different components travel at different speeds, and a detector at the far end records what comes out and when. The output is a chromatogram — a baseline with peaks rising from it. Each peak is something that came off the column at a particular time.
The purity figure is normally an area percent: the area under the main peak divided by the total area under all peaks, as a percentage. So "99.1% by HPLC" means the main peak accounts for 99.1% of the total detected response — at the detection wavelength used.
That qualifier does real work, and it is why the method column matters. A UV detector only sees what absorbs at the wavelength it is set to. Anything that does not absorb there contributes no peak, and therefore no area — water and residual salts, for instance, are largely invisible to it. This is not a trick; it is a well-understood property of the method, and it is exactly why a serious COA reports water content separately by titration rather than leaving you to assume the HPLC figure covered it.
A well-built COA puts two columns side by side: the specification — the range the material was required to fall inside — and the result, what the instrument actually measured. Reading only the result is how buyers end up satisfied with a number that never had a target to miss.
A specification of "≥ 98.0%" against a result of "99.3%" is a clean pass with headroom. The same specification against a result of "98.1%" is also a pass — but it is a pass at the edge, and when you are comparing two suppliers that gap tells you something the pass/fail column does not. Where results appear with no specification beside them, there is nothing to pass or fail against, and the document has quietly become a list of numbers.
One more piece of phrasing worth knowing: results are sometimes reported as "not detected" rather than zero. Those are different statements. "Not detected" means below the method's limit of detection, and a rigorous report states what that limit was. Zero is a claim no instrument can actually make.
Everything above is worth nothing if the certificate in front of you describes a different batch from the material in front of you. Traceability is the whole mechanism, and it rests on one unglamorous field: the lot or batch number, printed on the container, repeated on the certificate, and recorded against the sample the laboratory retained.
A supplier running this properly can do three things on request. Provide the certificate for the specific lot you received, rather than a representative example. Show that the identifier on the label matches the identifier on the report. And state how long a retained sample of that lot is held, so an independent re-test stays possible if a result is ever disputed.
The strongest version of this is per-lot documentation the buyer can pull themselves — a reference on the label that resolves to that lot's own report, rather than a PDF emailed on request. It removes the step where a document has to be selected by hand, which is precisely the step where the wrong one gets sent.
None of these require a laboratory to spot. They are all visible on the face of the document in under a minute, which is what makes them useful.
Ten checks, most of them a few seconds each. The habit is worth building, because the pattern is consistent: suppliers who publish complete documentation tend to have the systems that produce complete documentation, and suppliers who publish a number tend to have the number.
Who publishes this resource, why it exists, and how to reach the editor with a correction or a question about the guide.
xPepShop is an editorial resource on analytical testing documentation, published by DigiAD Agency LLC, a Wyoming limited liability company. It exists because the questions a buyer should ask about a certificate of analysis are almost always the same ones — what was tested, by whom, using which method, and against which specification — and they are easier to answer once, properly and in public, than one conversation at a time.
