Most purchase decisions are made on the first line of a certificate of analysis. It says 95%, the spec says 95%, and the lot moves. That is roughly one sixth of the information available on the page, and the other five sixths decide whether your next production run behaves like this one.
1. Assay — and the method behind it
“95%” means nothing without the method. A UV assay against a calibration curve measures absorbance at a single wavelength; overlapping compounds contribute. An HPLC assay with an external standard separates the analyte first. The two can differ by several percentage points on the same material.
Ask three questions: which method, which column and wavelength, and is the result calculated on a dried or as-is basis? A certificate that does not state the basis is telling you it does not matter to them. It matters to your formulation. Area percentage is a relative number — the next note in this series takes it apart.
2. Moisture
Karl Fischer water is not a formality. At 5% moisture, a tonne of powder contains 50 kg of water. It affects net actives per capsule, hygroscopicity, microbial risk and shelf life. Compare the moisture limit against your process — not against the supplier's own previous lot.
3. Ash and acid-insoluble ash
Total ash indicates mineral residue; acid-insoluble ash points to soil or silica contamination that survived extraction. For root and rhizome materials, a high acid-insoluble ash result is a harvest and washing problem, and it usually correlates with a heavy-metal or pesticide finding elsewhere on the page.
4. Heavy metals
Look for the specific elements — Pb, As, Cd, Hg — with limits, not a single “heavy metals: compliant” line. ICP-MS results below 1 ppm for lead on a botanical extract are plausible; a result of “not detected” with no limit of detection stated is not a result.
5. Residual solvents
ICH Q3C groups solvents by toxicity class. Ethanol and ethyl acetate are class 3 and rarely a problem. Toluene, dichloromethane and methanol are class 2 and must be reported with concentrations. If the extraction route uses a class 2 solvent and the COA is silent, the certificate is incomplete rather than clean.
6. Microbiology and the sampling plan
Total plate count, yeast and mould, E. coli and Salmonella should each have a value and a limit. The more useful question is how the sample was taken: a single grab sample from the top of a drum tells you very little about the lot. An AQL-based composite sample does.
What is not on the certificate
Identity is often the weakest section. A COA can report 95% berberine and say nothing about whether the material came from Berberis aristata root, whether the plant part matches the claim, or whether the marker was spiked. Ask for the botanical name, the plant part, the harvest region and the analytical fingerprint — TLC, HPLC-UV diode array or FTIR comparison against a retained standard.
None of this makes a supplier dishonest. It makes the certificate a starting document rather than a closing one. Our botanical release specifications are written to be read this way, field by field.