How to read a certificate of analysis

What a certificate of analysis actually is

A certificate of analysis (COA) is a report in which a laboratory states what it received, what it did to it, and what it measured. That is the whole genre. It is not a licence, an approval, a safety assessment or a quality guarantee, and it says nothing about any container other than the one the tested sample came from.

Read as a document, a COA answers five questions in sequence: who tested, what they tested, how they tested it, what they found, and when. A useful certificate lets you follow that chain without guessing. A decorative one breaks it somewhere — and the break is the finding.

1. Laboratory and client details

Start at the top of the page, not at the percentage. Look for:

• The laboratory’s legal name and address. A logo is not a name. An identifiable organisation can be looked up; a stylised mark cannot.
• A report or certificate number. Unique per report, and referenced on every page.
• The client. Who submitted the sample. This tells you whose sample it was and therefore how independent the exercise really was.
• Accreditation statements. If ISO/IEC 17025 or similar is claimed, expect an accreditation body, a certificate number and — critically — a scope that includes the test being reported. Accreditation is method-specific, not a general badge.

If the laboratory cannot be identified, nothing further on the page can be verified, however impressive the layout.

2. Sample identity and batch number

This is where most weak certificates fail. The report must connect a physical sample to a named material and to a batch code you can find on the label of the material you are holding.

A batch number that appears on the label but nowhere in the report leaves you with two unrelated documents. So does a report whose sample identity reads only “sample” or “client-supplied material”.

3. The method section

The method is what converts a number into a measurement. For a purity result by HPLC, a report that lets you interpret the figure will normally state the technique and mode, the column, the mobile phases and gradient, the flow rate, the injection volume, the detection wavelength and the run time. For mass confirmation, expect the instrument type, the ionisation mode and the observed against expected mass.

Method detail is also what makes a result comparable. Two laboratories reporting 98.4% and 99.1% on the same batch may simply have used different wavelengths or gradients. Without methods, you cannot tell whether you are looking at a disagreement or at two different questions.

“In-house method” is not automatically a problem — most laboratories run validated in-house methods — but an in-house method with no parameters at all is unreadable by design.

4. The chromatogram

If a chromatogram is included, it is usually the most informative part of the document, because it is much harder to fake convincingly than a number in a table. Check that:

• Both axes are labelled and scaled. Time on the horizontal axis, detector response on the vertical. A trace with no axes is decoration.
• The full run is shown. A plot cropped tightly around the main peak hides whatever eluted before and after it.
• The baseline is visible and flat. Drift, noise and a rising baseline all affect integration.
• A peak table accompanies it. Retention time, area, area percent and, ideally, peak width or resolution for each integrated peak.
• The main peak’s retention time matches the results table. They should be the same number, and surprisingly often are not.
• The solvent front is distinguished. The unretained material at the very start of a run is normally excluded from the purity calculation, and a report should make clear what was integrated.

Small peaks are not a red flag in themselves. Synthetic peptides are expected to carry process-related impurities, and a report that shows them is usually describing a real measurement. A perfectly clean single-peak trace with no baseline noise at all deserves more scepticism than a slightly untidy one.

5. Results, specifications and units

A result is only interpretable next to the criterion it was judged against. Expect a table with the test, the specification, the measured result and the units.

Three specific confusions to watch for:

• Specification quoted as result. “Purity: ≥99%” is an acceptance criterion. “Purity: 99.2% (specification ≥98.0%)” is a result. The first tells you what someone wanted; only the second tells you what was found.
• Purity read as content. An HPLC area percentage is not the amount of peptide in the container. Net peptide content and water content are separate tests, and their absence means those numbers are unknown rather than favourable. See peak purity, net peptide content and water content.
• Implausible precision. Chromatographic purity reported as “99.99%” claims a resolution the method does not have. Two decimal places on a peak-area percentage is a presentation choice, not additional information.

6. Dates and authorisation

Dates turn a document into a timeline, and the timeline has to be coherent: a manufacture or batch date, a sample received date, an analysis date and a report date, in a plausible order. A report dated before the batch it describes is a straightforward contradiction. A report several years older than the material in front of you is a historical record — peptides degrade, so the result has an age as well as a value.

At the foot of the document, look for the name and role of the person who authorised the report, a signature or controlled electronic equivalent, page numbering in an x of y form, and a revision or version indicator. Page numbering matters more than it sounds: a single page extracted from a five-page report can be technically accurate and still deeply misleading.

Common warning signs

None of these proves misconduct. Each one identifies something you cannot verify — which is precisely the information you need.

• No laboratory name or address anywhere on the document.
• No method section, or a purity figure with no technique attached.
• A chromatogram with no axes, no scale, no peak table, or cropped around the main peak.
• Sample identity left blank, generic, or unconnected to any batch code.
• A batch code that does not match the label, or a label with no batch code at all.
• Missing, impossible or internally inconsistent dates.
• Specifications presented where results belong.
• Purity quoted to an implausible precision, or as a round “99%” across every batch.
• The same document, unchanged, offered for multiple different batches.
• Visible editing artefacts: mismatched fonts, misaligned text boxes, a number that sits at a different resolution to the text around it, or a flattened image where the rest of the report is selectable text.
• Accreditation claimed with no accreditation body, certificate number or scope.
• Test results with no units, or units that do not match the test.
• Marketing language inside the analytical report itself — a laboratory reports findings; it does not advertise.

A field-by-field checklist

Worked through in order, the checklist usually resolves into one of three conclusions: the document supports its claim, the document is incomplete in ways you can name, or the document cannot be tied to the material at all. Those are genuinely different situations, and keeping them distinct is the point of the exercise.