A certificate of analysis arrived with fourteen lines on it. Appearance, colour, flavour, texture by hand, moisture, water activity, total plate count, yeast and mould, coliforms, lead. Everything passed. The buyer forwarded it to a retailer and the retailer asked one question: how much of the vitamin is in the gummy?
There was no line for that. There was a line for "vitamin premix added", which is a statement about the kettle rather than about the product.
This is the most common shape of a weak COA. It tests the factory thoroughly and the formula not at all.

Where the regulation stops
In the United States, 21 CFR 111.75 sets out what a supplement manufacturer must do to determine whether specifications are met. Three obligations matter here.
- Before use —— at least one appropriate test or examination to verify the identity of any component that is a dietary ingredient. Other components can rely on a supplier certificate, but only after the manufacturer has qualified that supplier by confirming the supplier's results.
- In process —— monitoring at the points where control is necessary, to detect deviation before it reaches the finished batch.
- Before release —— verification that the finished batch meets specifications for identity, purity, strength, composition and contamination limits, on every batch or on a statistically justified subset.
Read that list again and notice what it does not contain. It does not name a single test. It does not say how many samples, which method, or which analyte. The regulation requires you to verify that *your* specifications are met, and the specifications themselves are established separately, under 111.70, by the manufacturer.
A specification that says "vitamin C: present" produces a COA that says "vitamin C: present" and passes. A specification that says "vitamin C: 60 mg per piece, 95 to 110 per cent of label at release, minimum 80 per cent at end of shelf life, RSD across ten pieces below 10 per cent" produces a COA that can fail. Only the second one is a test.
Which gate buyers never see
Incoming, in process, release. Buyers see the third and hear about the first two.
Incoming quality control is where most of the risk is retired. Identity testing of a botanical or a premix is the difference between a formula and an assumption, and the supplier certificate is only worth the qualification behind it, which means somebody at the factory has to have re-run the supplier's numbers at least once and documented it.
In-process control is where the batch is actually made. Soluble solids at the cooker outlet, pH before the acid is dosed, deposit weight per piece, seal integrity on the sachet or the induction seal on the bottle, metal detection at the end of the line. None of these appear on a COA because they are not properties of the finished product. All of them decide whether the finished product is uniform.
Release testing is the gate that gets exported, so it is the gate that gets abbreviated.
Potency: the method has to match the matrix
A vitamin in a gummy is not a vitamin in a capsule. The matrix is a hydrated gel with sugar, polyol, pectin or gelatin, acid, colour and flavour, and every one of those interferes with an extraction.
A method validated on a powder will often under-recover from a gummy, because the analyte is bound in the gel network and the sugars caramelise at the extraction temperature. The fix is a method developed against the matrix: a spiked recovery experiment where a known amount is added to a blank gummy and pushed through the whole procedure, plus calibration standards prepared in a matrix-matched solution rather than in water.
Ask for the recovery figure. A potency result without a recovery figure is a number with an unknown bias, and a bias of fifteen per cent is the difference between passing and failing a label claim.

Then there is the question of timing. Potency at release is not potency at the end of shelf life. Most actives decay, some by a few per cent over eighteen months and some by a third. The label promises the end date, so the specification has to hold at the end date, and the release result has to be high enough to allow for the decay. A COA reporting release potency alone has answered a question about the day of manufacture.
Uniformity is the test nobody asks for
Take a three-gram gummy carrying five milligrams of an active. That is one part in six hundred. Now take a batch of a hundred thousand pieces, deposit them from a hopper over four hours, and ask whether every piece carries the same five milligrams.
Nothing in the process guarantees it. The active may be added as a diluted premix or as a straight powder, and a straight powder added to a viscous syrup at the wrong point of the mix will not disperse. The syrup stratifies in the hopper. The deposit weight varies piece to piece, so even a perfectly homogeneous syrup gives pieces with different total active.
Uniformity testing is how you find out, and it is the cheapest test on the list relative to what it protects. Ten pieces from the beginning, middle and end of the run, assayed individually rather than as a composite. A composite hides exactly the variation the test exists to find, because it averages it away.
The output that matters is not the mean. It is the relative standard deviation across pieces, and the lowest and highest single piece against the label. A batch with a good mean and one piece at forty per cent of label is a batch that will generate a complaint from one customer in a thousand and a testing failure somewhere down the chain.
Water activity is the shelf life test you can run today
Water activity is not moisture content. Moisture is how much water is in the piece; water activity is how much of it is free to move, and free water is what microbes and chemical reactions can use.
Gummies are formulated in a fairly wide band, roughly 0.50 to 0.75, with most commercial products held below 0.65 because that is where the microbiology stops being interesting. The thresholds are well established: most bacteria need above 0.85, Staphylococcus aureus manages at 0.85, xerophilic moulds get down to about 0.65, and osmophilic yeasts to about 0.60.
A gummy at 0.60 is therefore not sterile. It is a product in which nothing can grow, holding organisms that were killed by neither the cook nor the dry. That distinction decides your microbiological specification: a total plate count and a yeast and mould count on a low-aw product are testing for survivors and for hygiene, not for growth risk.
Water activity also predicts the texture complaint that arrives in month eight. Above the equilibrium point with the storage environment, the piece takes on moisture and the surface turns sticky. Below it, the piece loses moisture, the gel network tightens, and a chewy gummy becomes a tough one. The number belongs on the specification with a range, not a maximum.
Which heavy-metals limit applies
Two different kinds of limit are in play, and they are not interchangeable.
The European Union regulates concentration. Regulation (EU) 2023/915 sets maximum levels for food supplements of 3.0 mg/kg for lead, 1.0 mg/kg for cadmium, and 0.10 mg/kg for mercury. These are legal limits on the product as sold, enforceable at the border.
California regulates dose. The Proposition 65 maximum allowable dose level for lead is 0.5 micrograms per day, which is a statement about how much a person consumes, not about how much is in a kilogram. A product can sit comfortably under the EU concentration limit and still trigger a Proposition 65 warning if the serving size is large.
The method is the same either way, closed-vessel digestion followed by ICP-MS, and it is the sample preparation that decides whether the result is real. The specification should name the limit, the unit and the market it comes from, because "heavy metals: pass" is not a specification.

What an out-of-specification result should trigger
The instinct is to re-test. The correct response is an investigation, and the difference matters commercially.
An OOS result means either the batch is out or the method failed, and a re-test on a fresh sample cannot tell you which. The investigation asks whether the sample was representative, whether the standard was prepared correctly, whether the instrument was in calibration, and whether the batch itself deviated. Retesting until a passing number appears converts a quality system into a random number generator, and the batch that eventually ships carries the same risk it always did.
For a buyer, the useful question is not "did it pass" but "what happens when something fails". A supplier with a written OOS procedure will answer in a paragraph. A supplier without one will answer with a new COA.
Where our testing stops
Our SGS certificate CN23/00007035, held by Anhui Wanhuacao Biological Technology Co., Ltd., covers GMP against Codex Alimentarius CXC 1-1969 as revised in 2020, with gel candy inside the stated scope. That is a certification of the facility and its system. It is not a result for your batch, and no certificate of that kind ever is.
Release testing is run against the specification agreed for the project. Third-party testing at an accredited laboratory is available per batch, and the analyte list is agreed before the first production run rather than after the first complaint. If a market requires a specific method or a specific limit, say so with the brief, because the method has to be chosen before the samples are taken.
If the format is still open, the functional gummies page sets out what runs on these lines, and gummy shelf life and stability covers the storage programme that decides which of these numbers moves first. The contract manufacturing overview explains what a factory can and cannot put its name to. To agree an analyte list for a project, send the specification.
Ding Shaohua, Foreign Trade Manager, handles specifications and release documentation for this line. The same person answers on WhatsApp, TikTok, LinkedIn and Facebook.
FAQ
Is a COA with fourteen passing lines a good COA?
Only if the lines were chosen against a specification that could fail. A certificate listing appearance, moisture and microbiology has tested whether the factory is clean, which is worth knowing and is not the same question as whether the formula is right. The two lines that decide that are potency by a matrix-validated method and uniformity across individual pieces.
Why does potency differ between the release test and the label?
Because the label describes the product at the end of its shelf life and the release test describes it on the day it was made. Actives decay at different rates, so the release specification has to sit above the label by whatever the decay is expected to be. A COA that reports release potency without saying what it is expected to be at the end date has left the important half out.
How many pieces should be tested for uniformity?
Enough to see the batch rather than the average. Sampling at the beginning, middle and end of the run catches stratification in the hopper and drift in the deposit weight, and assaying those pieces individually rather than as a composite is what makes the test meaningful. A composite result is an average of the variation the test exists to detect.
Does low water activity mean the product is microbiologically sterile?
No. Water activity below about 0.65 stops growth, it does not sterilise. Vegetative cells and spores that survived the cook and the drier remain in the piece and are counted by a total plate count. What low water activity buys is that they cannot multiply in the pack, which is a different and more useful guarantee than sterility.
