How to check a certificate for internal consistency

A certificate of analysis can be beautifully formatted and still contradict itself. The fields that make up a certificate — identifiers, methods, results, dates, signatures — were probably entered by hand, and hand entry produces inconsistencies that formatting cannot hide. Checking a certificate for internal consistency is therefore not a chemistry exercise. It is a reading exercise, it needs no instruments, and it catches a surprising share of the problems that later become complaints.

“Internal consistency” is a precise phrase. It does not ask whether the measured value is correct — no reading of the document can settle that. It asks whether the document agrees with itself and with the shipment it arrived with. Every check below is a cross-check between two things that are already on the desk: the certificate, the label, the packing list, the order.

The lot number first

Start with the lot number and read it character for character. Prefixes, leading zeros, hyphens and the letter O standing in for a zero are where transcription errors live. The lot on the certificate should match the lot on the container label and the lot on the packing list. A mismatch does not automatically mean the material is wrong; it means one of the three documents was written by hand, and the finding is worth a corrected copy rather than a shrug.

A certificate that names no lot at all describes some material, but not this one. If the certificate is addressed to a distributor and the lot will be assigned later, the document should say so, and the lot-specific certificate should be traceable from the number that ends up on the label.

The method line

A real method has a name. “HPLC” alone describes a family of techniques; a useful line names the technique, the detector where it matters, and often the pharmacopeial or in-house designation of the procedure. If the purity is reported “by HPLC” and the identity “by mass spectrometry”, both techniques should appear somewhere in the document. A result attributed to a method the document never names is a small flag that grows when other fields are also thin.

Where a published method is cited, the edition matters. Compendial methods are revised, and a certificate citing an edition that predates the current one may still be perfectly good — or it may describe a procedure that no longer reads the way the material was tested. The check is not to verify the citation from memory; it is to confirm that a citation exists and is specific enough to look up.

The result against the specification

The result line and the specification line have to be readable against each other. Units are the first cross-check: a purity of 99.2 against a specification of “≥95%” only works if both describe the same basis — area percent, mass percent, titration. A result in parts per million against a limit in percent is a reading error waiting for a reader. Certificates that report water by Karl Fischer and residual solvents by headspace should keep those figures in their own rows; a combined table that mixes bases is where transcription slips in.

Decimal places are a quiet tell. A value reported to more places than the method can support, or a specification set at a precision the measurement never had, suggests the numbers were composed rather than measured. Neither is proof of anything; both belong in the note that goes back to the supplier if one is written.

Dates that agree

A certificate usually carries at least three dates: when the sample was received, when it was tested, and when the report was issued. They should tell a story in that order. A test date that precedes the receipt date, a report issued the day before the test, a manufacture date later than the test — each is a small impossibility that says the fields were filled from more than one source. Dates that agree do not prove anything; dates that disagree are always a finding.

Who signed

The signatory line should carry a name and a role. “Authorized by” followed by a job title is stronger than a stamp alone, and a name alone is stronger than neither. The laboratory’s own block — name, address, and, where it holds one, its accreditation mark and number — can be checked against the scope that laboratory publishes: the test on the certificate should fall inside it. None of this requires suspicion; it requires the habit of reading the fields against each other.

Reading across the paperwork

Finally, read the certificate against the documents that traveled with it. The product name on the certificate, on the label, on the safety data sheet and on the invoice should be the same product. A certificate describing one grade while the label names another survives in the market because each document is read alone. Two minutes with them side by side finds what a year of isolated reading does not.

Internal consistency is not proof of authenticity — a careful forgery is consistent. But inconsistency is always real, always cheap to find, and always worth writing down. The receiving record can carry the check: read once for the science, once for the paperwork.