The 30-Second COA Check: Five Failures Buyers Spot First

A certificate of analysis can pass every analytical test it reports and still lose a buyer before the reader reaches the results. The failures that do that damage are not analytical. They are internal inconsistencies, and they are found in seconds without a calculator, which is exactly what makes them expensive.

The reason is straightforward. An analytical problem has to be discovered by work and can be argued on its merits. An inconsistency in the document is discovered by reading, and it speaks directly to whether the results were produced and recorded with care. A reviewer who finds one is now reading everything else with a raised eyebrow.

What follows is the check itself, the five failures it looks for, and the important part: what it does not prove.

Why internal consistency carries so much weight

A certificate is a document that will be filed, quoted, and in some cases attached to a regulatory submission. Its individual numbers are used by people who were not present when the testing was done, which means they have only the document to judge. An inconsistency breaks the document's claim to have been produced by a controlled process, and once that claim is in doubt the specific error matters less than the doubt.

There is a second reason specific to this trade. Bile-acid certificates carry a mixture of chemical values, method references and administrative fields, and the errors that appear are usually administrative. That is reassuring in one sense, because nobody is claiming the chemistry is wrong. It is damaging in another, because administrative errors are the ones a buyer can find without technical knowledge, and buyers who cannot evaluate the chemistry will evaluate what they can.

Five checks that take under a minute

Each of the five is a comparison between two things already on the page. No calculation is required, which is what makes them fast and reliable.

The first is dates. A certificate normally carries a manufacturing date, a release or test date, and a retest or expiry date, and those dates can appear in more than one place, sometimes in more than one language. The check is that every instance of the same date agrees. A date written one way in a header and another way in a footer is the single most common inconsistency on a technical certificate, and it is read as carelessness.

The second is the method statement. Every quantified result should name the method that produced it or reference where the method is described. A page where some tests carry a method and others do not is a page that raises the question of why. The problem is not that a method is missing; it is that the pattern of omission looks selective.

The third is units. Limits and results should both carry their units, including the per cent sign. A limit written without a unit will be read with an assumed one, and different readers will assume differently. This matters more than it appears, because specifications get quoted into agreements, and the quoting tends to drop the context that made the unit obvious.

The fourth is the reference to the standard. A certificate states the specification or standard it is issued against, and that statement should name a document rather than a party. When the reference field contains something that is not a standard, the reader cannot tell what the material was tested against, which undermines every limit on the page.

The fifth is signature independence. Testing, review and approval are three roles. Where the same person occupies more than one of them, the independent review has not taken place, regardless of how many signature lines are filled in. A reviewer who sees the same name on two of the three lines has learned something about the signature block, and it is not a good thing to have learned.

The five checks in the form a reviewer uses

Check	What a failure looks like	What it costs
Dates agree	The same date written differently in two places, or in two languages	Immediate doubt about document control
Methods stated	Some results carry a method, others do not	Results cannot be compared with your own
Units present	A limit without a unit or per cent sign	Ambiguity that propagates into agreements
Standard named	The reference field names a party rather than a document	Every limit on the page becomes unanchored
Roles independent	One name on two or three of the testing, review and approval lines	The review requirement is unmet on its face

None of these five requires the reader to know anything about bile-acid chemistry. That is the point: they are checks a procurement lead can make as easily as a quality manager, and in a review meeting they are made first.

What the check does not prove

A clean consistency check proves that the document was produced carefully. It does not prove that the material is good, and the distinction needs to be held firmly.

A certificate can be internally flawless and carry a result that is wrong. A transcription error that happens to be self-consistent, a calculation performed on the wrong basis throughout, or a method applied outside its validated range will all pass a consistency check and still produce a misleading certificate. That is why the consistency check is the first step and not the whole review.

Conversely, a single administrative slip does not make a batch unacceptable. A date written inconsistently is a reason to ask a question and not a reason to reject material.

The analytical review that follows

Once the document is internally consistent, the review moves to the numbers. That part is slower and is where the real work sits: comparing limits with results, reading the content figure together with the volatile figure, checking that the impurity panel is complete, and placing the whole certificate against the previous batches so that a trend becomes visible.

Reading the lines that carry weight covers that second stage in detail, and batch-to-batch consistency covers what to do with the results once several certificates are on the table. Third-party testing adds a further question, which is whose scope the results fall under when the testing was performed outside the releasing organisation.

What we do about it

We run this check on every certificate before it goes out, and we treat it as a release step rather than as a formality. Dates are written once and referenced everywhere else. Methods are stated against every quantified result. Units are carried through limits and results alike. The reference field names the document, not the party. And the three signature roles are held by three people, with the names printed rather than initialled so that the separation is visible to a reader.

None of that makes the chemistry better. It makes the document worth the trust that gets placed in it, which for a certificate is the whole of its function.

The two-minute version, written out

The consistency check is easier to run than to describe, so it is worth writing as a sequence. Open the certificate and find every date on it; confirm that each date appears consistently wherever it is repeated, including across languages. Read down the results column and confirm that each quantified result names a method or references one. Read the same column again and confirm that each limit and each result carries its unit. Find the field that states the standard or specification and confirm that it names a document. And read the three signature lines and confirm that they are occupied by three people.

That sequence takes about a minute on a single-page certificate and rather longer on one with an appendix. It requires no technical knowledge and no calculation, which is what makes it reliable: there is nothing to interpret, only comparisons to make.

What the sequence produces is not a judgement about the material. It produces either a clean page or a list of questions, and either outcome is a gain because the list arrives before the material is accepted rather than after.

Where these failures come from

Each of the five failures has a characteristic cause, and knowing the cause makes the response proportionate.

Date inconsistencies come from documents assembled by more than one tool. A header generated by a system and a footer typed by hand will disagree eventually, and the disagreement is a process problem rather than a person problem. Method omissions come from templates built around specifications rather than around tests. Unit omissions come from the compression that happens when a table is widened to fit a page. Standard-field errors come from a field that was never defined clearly enough to be filled correctly. And signature independence fails when an organisation is small enough that one person legitimately holds more than one role, which is a structural issue rather than a lapse.

Knowing which cause applies matters because the responses differ. A template problem is fixed by changing the template. A structural problem is fixed by engaging a second signatory or by being explicit about which of the roles is combined. A filing problem is fixed by redesigning the page. Treating all five as carelessness produces a correction that does not hold.

What good looks like on the page

A certificate that passes the check has a recognisable shape. Dates appear once in a defined place and are referenced rather than repeated. Every quantified result has a method beside it. Units are present everywhere they are needed and nowhere they are not. The standard field names a document and, where relevant, its edition. And the signature block has three lines with three names, printed rather than initialled.

None of those features is expensive. All of them are visible to a reader who has never seen the material, which is the audience a certificate is written for. The guidance documents that deal with records and data expectations describe the underlying requirement in terms of accuracy and completeness rather than of layout, and the layout is simply how those two properties become visible.

Where the material is compared against a pharmacopoeial monograph, the naming of the edition follows the same logic, because the European Pharmacopoeia is revised on a cycle and a limit quoted from one edition may not match another. The framework that governs specification content and justification is the wider reference for what belongs on the document in the first place. A certificate that is internally consistent and states its basis is one that can be filed, quoted and defended without a follow-up letter, which is the whole purpose of issuing one.

Why this check belongs in procurement, not only in quality

The five checks can be run by anyone. That fact is worth acting on, because the person who sees a certificate first is usually the person who ordered the material, and the time between seeing a problem and raising it is often longer than it needs to be for want of a habit.

Building the check into procurement has a second effect. A supplier whose certificates consistently pass a check performed by the buyer's own staff gets a different kind of feedback from one who only hears about problems at the quality review stage, and the feedback arrives earlier. The check is not a substitute for the analytical review; it is the reason the analytical review starts from a clean page.

Where the material arrives with a certificate that fails one of the five, the appropriate action is usually a question rather than a hold. A date inconsistency or a missing unit is a reason to ask for a corrected document and to note the pattern if it repeats. The material itself is not implicated by a document fault, and treating the two as the same thing slows down receiving without improving anything.

Frequently asked questions

Why do document errors matter more than analytical ones?

Because they are visible without a laboratory. An analytical problem requires work to find and can be discussed on its merits. An inconsistency in the document is found in seconds and speaks directly to how carefully the results were produced and recorded.

What is the most common consistency failure?

Dates that do not agree within the same document, usually between a date written in one language and the same date written in another. It is the easiest failure to make and one of the easiest for a reviewer to find.

Does an unsigned or single-signed certificate matter?

Yes, when the roles are not independent. Testing, review and approval are separate responsibilities, and where one person occupies more than one of them the independent review has not happened, whatever the paper says.

Is a missing unit symbol on a limit a real problem?

It is a small problem with a disproportionate effect. A limit written without its unit invites a reader to supply one, and readers supply different ones. For a specification statement that is quoted into contracts, the ambiguity is worth removing.

How long should a quick review take?

The consistency check itself takes about a minute. The analytical review, where limits and results are compared and trends examined, takes longer and is a separate exercise. Keeping the two apart is what makes the quick check reliable.

What does a clean consistency check prove?

That the document was produced carefully. It does not prove anything about the material. A certificate can be internally perfect and carry results that are wrong, which is why the check complements the analytical review rather than replacing it.