Guides / How to read a certificate of analysis
How to read a certificate of analysis, and what it cannot tell you
How do I check whether a certificate of analysis is any good?
A usable certificate states the batch, the method with its compendial reference, the result with its unit, and a numeric detection limit for every quantitative determination. It lists what was not tested. It attaches raw data. Verification proves the document is genuine and unaltered; it does not prove the unit in front of you came from the batch named on it.
Every quantitative determination should occupy one row carrying: parameter, method with compendial reference, result, unit, limit of detection, limit of quantitation, expanded measurement uncertainty, and whether that parameter fell inside the performing laboratory's accredited scope.
The accreditation column is the one most often abused. A single certificate can legitimately carry both accredited and non-accredited determinations, and stating which is which per row is the difference between disclosure and implication.
| Written this way | Rather than |
|---|---|
| Identity by LC-MS, theoretical 1419.5 Da, measured 1419.4 Da | Identity confirmed |
| Purity 98.2 percent, RP-HPLC-UV 214 nm, chromatogram attached | Purity 98.2 percent |
| Net peptide content 78.4 percent, 7.84 mg per labelled 10 mg vial | Contains 10 mg |
| Lead 0.08 µg/g, LOD 0.01 µg/g by ICP-MS | Heavy metals: not detected |
| Microbial enumeration USP <61>, TAMC 40 CFU/g | Sterility: pass |
| Endotoxin 0.4 EU/vial, spike recovery 92 percent | Endotoxin: pass |
First, does the batch number on the document match the batch number on the container in your hand. If it does not, the certificate is about a different product and nothing else on the page matters.
Second, is there a detection limit beside every not-detected. Not detected without a limit is a statement about the instrument, not about the sample, and an insensitive method will return it reliably.
Third, does the test name match the chapter cited. Enumeration presented as sterility, or an X-ray fluorescence screen presented as compliance with elemental impurity limits, are the two substitutions that recur.
Fourth, is there a list of what was not determined. Its absence is not neutral: a reader seeing purity and no mention of endotoxin will tend to read the silence as reassurance.
A verification key that resolves against the laboratory's own server proves the document is genuine and has not been edited after issue. That is worth having, and it defeats the fabricated PDF.
It does not prove the vial in your hand belongs to the batch on the certificate. A supplier can test one batch and continue shipping later production under the same label, and every check will still pass because the document is authentic. The gap is between the document and the object.
Nor does a public results database close it. What appears there is what suppliers chose to publish, so the distribution of results in any public database is shifted upward by selection. The measures that address the gap are recurring batch testing on a schedule rather than on request, and blind acquisition, where the laboratory buys the sample through the ordinary sales channel.
Chromatograms and spectra are the evidence behind the numbers. A chromatogram shows whether the gradient resolved anything, whether the peak is symmetrical, and whether the integration was reasonable. A mass spectrum shows whether deconvolution was clean.
A laboratory unwilling to release raw data is asking to be taken on trust for the one part of the process that can actually be checked.
What does not detected mean on a certificate?
Only that the analyte was below the method's detection limit. Without a stated limit the phrase carries no information, because a sufficiently insensitive method reports not detected on almost anything.
Does a verified certificate mean my product is genuine?
It means the document is genuine. It does not mean the unit in your hand came from the batch named on it. Check that the batch number on the container matches.
Should a certificate include the chromatogram?
Yes. A percentage cannot show whether the separation resolved anything. The chromatogram shows peak shape, resolution and integration.
Why do public testing databases show such high purity?
Because publication is voluntary. What appears is what suppliers chose to publish, which shifts the visible distribution upward relative to the underlying one.
References
- ISO/IEC 17025:2017, clause 7.8 on reporting of results
- ICH Q2(R2), Validation of Analytical Procedures
Sources
Updated 2026-08-24