A certificate of analysis for a mineral ingredient looks like a page of numbers. The numbers only become evidence when the form, the analytical method, the basis of the calculation and the release limit are read with them. This guide sets out a reading order for a batch certificate, using magnesium bisglycinate as the worked example, and marks the places where a result can be accurate and still answer the wrong question.

Start with the batch, not the assay

A certificate of analysis is issued for a lot, not for a product family. The first pass is therefore administrative: the product name, the chemical form, the lot number, the manufacturing site, the quantity, the specification version and the date of release should all match the purchase order and the drum in front of you. A certificate that is internally consistent but names another lot is not evidence for this delivery.

The certificate is also not the technical data sheet. The data sheet describes the range a grade is expected to meet; the certificate records what one lot actually measured against that range. This distinction matters when a generic certificate is forwarded with a batch number added later. Ask for the certificate issued for the lot, signed and dated, and check that the specification version on it is the one you approved. A short certificate can be complete for a commodity mineral, but a chelate needs more than a metal assay. If the document has no ligand result, no ratio and no hydrate statement, the missing lines are the finding.

Identify the form before reading the percentage

Magnesium is the element; the ingredient is the compound. Magnesium oxide, magnesium citrate and magnesium bisglycinate share the same metal and behave differently in a formula because the anion or ligand is most of the mass. The certificate should name the compound, the hydrate state and, for a chelate, the ligand ratio. Magnesium bisglycinate is a 1:2 chelate of one magnesium ion and two glycine molecules, CAS 14783-68-7; the anhydrous, monohydrate and dihydrate forms are different drums with different elemental contents.

The identity block is where the certificate says what the material is, but the methods behind it decide how much the statement proves. A complexometric titration measures total magnesium and cannot tell whether the magnesium came from a chelate, an oxide or a blend. X-ray diffraction can detect crystalline magnesium oxide above its detection limit. FTIR can show metal–ligand coordination. HPLC can measure glycine. Read the identity block as a set of methods rather than a single tick. If it names only magnesium and no form-specific test, the form is an assumption that belongs to the supplier's paperwork, not to your specification.

Read the assay with its basis

Two percentages are routinely confused. Elemental magnesium is the mass of the metal in the material; compound purity is the mass of the named molecule. The anhydrous chelate is about 14.1% magnesium and 86% glycine by mass, so a 14% magnesium result is consistent with the chelate, not a purity of 14%. A supplier may also quote a purity figure of 98% or 99%; that number answers a different question and cannot be compared with the elemental figure.

The method and the basis have to travel with the number. Complexometric EDTA titration and ICP-OES both measure magnesium, but they have different interferences, different sample preparation and different uncertainty. A result expressed on a dried basis is not the same as a result expressed as-is. The certificate should state the method, the basis and the calculation; without those three, two numbers from two suppliers are not comparable even when they look identical. For a hydrate, read the assay next to the loss on drying: the monohydrate carries about 12.6% magnesium and the dihydrate about 11.8%, so a lower assay may describe the same chemical entity and a different material on the shelf.

The ligand block is the identity check

For a chelated mineral, the ligand result is not a secondary line; it is part of the identity. The certificate should report glycine by HPLC, free glycine and the calculated magnesium-to-glycine ratio. A 1:2 chelate produces a ratio close to two moles of glycine per mole of magnesium; a specification range such as 1.95–2.05 is tight enough to catch a blend and wide enough for normal analytical variation.

The ratio is calculated from two measured values, so its uncertainty is the uncertainty of both assays. It is a consistency check, not a direct measurement. Added magnesium oxide raises the magnesium without adding glycine, so the ratio falls. Added glycine can push the total glycine figure back toward the expected range, but the free glycine rises because that glycine is not bound. Read total glycine, free glycine and the ratio together; each one can be satisfied on its own by a different route. The elemental ceiling is another useful check: the anhydrous 1:2 chelate cannot exceed about 14.1% magnesium by mass, so a certificate showing 18% or 20% is describing a material with another magnesium source in it. The reverse is not true, because a blend can be formulated down to 14%; a 14% result alone does not prove the chelate.

Physical results predict the process

Particle size, bulk density, pH and loss on drying are often read as housekeeping lines. In a capsule or tablet they decide the fill weight, the flow and the compression behaviour. A specification of at least 95% through 40 mesh and a bulk density of 0.55–0.90 g/mL describes a powder that can be filled, but the same figures also show why two lots can behave differently on the same machine.

Each physical result has a method behind it. Sieve analysis gives a single cut; a particle-size distribution gives the fines and the coarse tail that cause segregation or dust. Bulk density depends on whether it was poured or tapped, and the cylinder method should be named. pH depends on the concentration of the suspension; a pH of 8.5–11.5 at 5% is not the same measurement as a pH on a 1% dispersion. Appearance is qualitative and easy to dismiss, but it is a useful first signal. A cream or tan tint can follow a drying step that ran hot, a trace of iron or a reaction with a reducing sugar; the assay can still pass, so the tint is a reason to ask for the batch record rather than a reason to reject the lot on its own.

Contaminants: read the limit, the result and the detection limit

The contaminant block usually covers lead, arsenic, cadmium and mercury by ICP-MS, and microbiology as a separate group. A result below a limit is not a zero, and a “not detected” statement is only meaningful when the detection limit is stated. The certificate should state the limit, the result and, where relevant, the limit of detection or quantification for the method used.

The legal limit also has a context. A raw-material limit is not automatically the finished-product limit, because the ingredient is diluted or concentrated by the rest of the formula. A result that is comfortably inside a limit for the powder can look different when the product is reviewed against the limit for the product as sold. The buyer's regulatory team has to make that translation; the certificate supplies the input. Microbiology has the same structure: aerobic plate count, yeast and mould, Escherichia coli and Salmonella, each with a limit and a method. A certificate is a snapshot of the lot at release, not a prediction of what will happen after a hygroscopic powder spends a month in a humid warehouse.

Trace the paper around the certificate

The certificate should connect to the material in three places: the lot number on the drum, the lot number on the retained sample and the lot number on the document. Add the manufacturing site, the date of manufacture, the date of analysis and the signature of the person who released the lot. A certificate can be genuine and still belong to a different lot; the identifiers are what tie it to this delivery.

The certificate is one document in a pack. The technical data sheet, the safety data sheet, the allergen and GMO statements, the origin record and the residual-solvent statement answer questions the certificate does not. Regulatory status is a third layer: a certificate can prove that a lot meets a specification without proving that the form is permitted in the destination market. Keep the three questions separate. Retesting is appropriate when the source is new, when the first shipment is entering a regulated market, or when the label makes a form claim that the certificate does not prove. The useful retest is not a full repeat of the supplier's panel; it is the set of tests that answers the open question, such as XRD for crystalline oxide, FTIR for coordination and HPLC for free glycine.

A reading order for the next certificate

A consistent order keeps the assay from anchoring the review. Start with the product and lot identifiers, then move to the form and hydrate statement, then to the assay basis and method, then to the ligand block, then to the physical results, then to contaminants and microbiology, and finish with the signatures and dates. If a line is missing at any step, write it down before moving on.

The comparison between suppliers should use the same six lines: form and hydrate, assay and method, ligand ratio and free glycine, loss on drying, particle size and bulk density, and the contaminant panel. A certificate that answers all six is a usable document. A certificate that answers only the magnesium figure is a starting point, and the missing answers are the questions to send back.

  • Match the product name, form, hydrate, CAS number and lot to the purchase order and the drum.
  • Confirm that every result has a method, a release limit and a stated basis.
  • Read the ligand ratio and free glycine before accepting the magnesium assay.
  • Check the loss on drying against the hydrate named on the certificate.
  • Read each contaminant result against its limit and its detection limit.
  • Confirm the dates, signature, specification version and retained sample.

What to take away

  • A certificate of analysis is lot-specific; the identifiers have to match the drum and the purchase order before any result is read.
  • A number without a method and a basis is not comparable with a number from another laboratory.
  • The elemental magnesium assay is not an identity test; the ligand ratio, free glycine and an XRD oxide screen answer the form question.
  • Hydrate state and loss on drying belong together, because they explain why two certificates can report different magnesium figures for the same chemical entity.
  • “Not detected” is only meaningful with a detection limit, and a raw-material limit is not automatically a finished-product limit.
  • A consistent reading order catches the missing line that a quick scan of the magnesium result would pass over.

Go deeper

XRD oxide screen

What an X-ray diffraction screen can and cannot prove about crystalline magnesium oxide in a chelated lot.

Read the reference

Cite this note

Rainwood Nutrition. “How to Read a Mineral COA.” Published 18 November 2026. https://nutrition.rainwoodbio.com/blog/how-to-read-a-mineral-coa/