The certificate was perfect. The material was not.

A raw material once arrived at this factory with documentation that satisfied every requirement. Every specification in range, every signature in place, nothing missing. Testing found that the material was not what the paperwork said it was.

Our chairman tells that story as something the factory learned the hard way, and it is the most useful thing a first-time brand owner can hear about raw materials: the paperwork being perfect is not evidence that the material is. A bad shipment with a bad certificate stops itself at goods-in. A bad shipment with a flawless certificate is the one that reaches your blend.

A certificate of analysis is a claim, not a measurement

A CoA asserts that somebody, somewhere, tested something and recorded these results. Read as a claim it is useful. Read as evidence it is a liability.

Four things go wrong with it routinely, none exotic.

It describes a different lot. The certificate travels with a lot number. Whether that number corresponds to the material in the drum in front of you is an assumption until someone tests the drum.

It was re-issued, not re-tested. Much of the global ingredient trade moves through traders. A certificate retyped onto a trading company's letterhead may be an accurate transcription or may be three transactions removed from any laboratory. It looks identical either way.

It reports only what passed. A specification lists parameters; a certificate lists results. When the two lists are not the same length, the missing parameters are not passes. They are silence.

The method cannot see the problem. This is the subtle one, and the reason sophisticated adulteration survives documentation intact. An assay measures what its method responds to. A protein figure calculated from total nitrogen responds to nitrogen, whatever molecule carries it. A botanical extract assayed against one marker compound responds to that marker — including when the marker has been added to a weak extract deliberately, which is the mature form of the practice.

None of this requires anyone in the chain to be a criminal. Most of it only requires that nobody in the chain is checking.

Four ways an ingredient supply chain goes wrong

Substitution. A different species, plant part or grade shipped under the name you ordered. Root quoted, stem supplied. A closely related species with a similar profile and a fraction of the content. Substitution is attractive because it is defensible — the supplier calls it a sourcing variation.

Adulteration. Something added so the material tests the way it is supposed to: bulking agents to make weight, dyes to correct colour, the assay's own marker compound to lift a number, and at worst a synthetic active in something sold as a plant extract. Adulteration is aimed at the test, not the product, which is why a factory's panel has to be broader than the supplier can predict.

Potency drift. No bad actor at all. Botanical material varies by harvest, region, season, drying method and storage. A 10:1 extract from this year's crop is not last year's material under the same name and the same specification. Potency drift is the most common reason a finished batch misses its label figure, and the least discussed.

Contamination. Heavy metals, which concentrate in roots and leafy material. Microbiological load. Pesticide residue. Residual solvents from extraction. Mycotoxins in nuts, seeds and spices. And allergen carryover from whatever the supplier ran before your order.

What gets tested on arrival

A serious incoming regime is a sequence, not a test.

Quarantine first. Material coming off a vehicle goes to a physically segregated area with a quarantine label and cannot be dispensed from there. Status is a fact about where the pallet is standing, not a field in a spreadsheet.
Goods-in checks. Seals intact, containers undamaged, labels matching both the purchase order and the certificate, net weights as declared.
A written sampling plan. How many containers get sampled, and where in each container — powders segregate in transit, so a scoop off the top is not representative.
Identity. The non-negotiable test, ideally per container rather than per delivery. Appearance, colour and odour, plus an instrumental identification: an infrared fingerprint matched to a reference standard, chromatography for botanicals, melting point, pH, solubility.
Assay. Active content, by the method the monograph specifies rather than the easiest method available.
Purity and limits. Moisture or loss on drying, ash, heavy metals, microbiological limits, residual solvents.
Manufacturability parameters. Particle size, bulk and tapped density, flow. Not safety tests — these decide whether the blend flows, the tablet compresses and the sachet fills. A material that is chemically perfect and physically wrong still stops a line.
Retained samples. A sample of every lot kept to shelf life plus a margin. That is what lets a factory answer a complaint about a two-year-old batch with evidence rather than an apology.

Excipients and primary packaging go through the same process. People skip them because they are "inert". A failed lubricant halts a tablet press exactly as effectively as a failed active.

No factory runs all of this in-house; trace-metal and pesticide-residue work normally goes to a specialist laboratory. What matters is the split: identity and assay in-house, because their turnaround decides whether a line runs this week, and everything else at an accredited laboratory whose scope genuinely covers the test.

What should happen when a lot fails

The procedure matters more than the result, because every factory eventually gets a result like this.

The lot stays where it is. Status changes from quarantine to rejected, physically and visibly. Nothing moves closer to production while a decision is pending.
The failure is investigated before it is accepted. Re-sample, re-test, rule out the laboratory, the sampling and the method. A factory that rejects on one result is as unserious as one that releases on one result.
The decision follows a defined procedure. Reject and return, or reject and destroy. For a borderline parameter unrelated to safety or identity there may be a documented concession route signed by quality — never for identity, never for a safety limit, and never verbally.
The supplier is dealt with, not just the shipment. A wrong certificate is a more serious finding than a wrong lot. The supplier moves to increased testing on every delivery, or off the approved list.
Somebody's launch date moves. This is the part that gets negotiated in practice, and where a factory's character shows. It either tells the client the truth and reschedules, or finds a reason to believe the result was probably fine.

Quality in, quality out

Our chairman states the principle the same way every time he is asked where quality starts.

"Quality in, quality out. The more that goes in with quality, the more the output has quality in it. The beginning is in the production inputs."

He puts something in "inputs" that brand owners usually leave out. Not only the active and non-active materials, but the process conditions attached to them: which materials need granulation, at what drying temperature, for how long. Those parameters are inputs too, and they are specified per material, not per factory. A correct material processed on the wrong schedule is a failed batch with a clean certificate of analysis.

What to ask your contract manufacturer

If you have never audited an ingredient supplier yourself, two habits cover most of the risk: establish whether you are buying from a manufacturer or a trader, and treat a price well under market as information rather than a win. Beyond that, if you are evaluating a factory rather than an ingredient supplier, five questions get you to the substance:

Which incoming tests do you run in-house, and which go out, to which accredited laboratory?
Do you identity-test every container, or sample across a delivery?
Show me the incoming test records for a material you received last month.
What is your procedure when a material fails, and who signs the rejection?
Do you keep retained samples of raw materials, and for how long?

That last one is the quiet one. A factory that retains raw-material samples to shelf life is a factory that expects to be asked questions later and has decided to be able to answer them.

For what happens to those materials once they are released into production, see what quality control actually looks like on a capsule line.