Certificate of Analysis (CoA) — What It Contains and What It Does Not Guarantee

A certificate of analysis (CoA) is a lot-specific document stating the results of tests that were actually performed on the material you are receiving. It is a record of measurements on your batch. That is its entire claim, and understanding the boundary of that claim is most of what makes a CoA useful.

Three documents get confused with each other constantly, and the confusion costs money.

Document	Scope	Binds the supplier to	Typical failure mode
Datasheet	The product line, generically	Nothing lot-specific. It describes the intent of the product	Buyer plans against datasheet figures that were obtained on a different lot under vendor-optimised conditions
Specification	The acceptance criteria a lot must meet to be released	The release decision, if the criteria are stated numerically and contractually	Buyer assumes criteria exist. Many RUO products are released against internal criteria that are never disclosed
Certificate of analysis	The specific lot in the specific vial	The truthfulness of the stated measurements only	Buyer reads a passing CoA as a functional guarantee, which no CoA is

The single most important sentence about a CoA: it tells you what was tested, and by omission, what was not. A CoA with four lines on it is not a poor CoA of a good product; it is a record that four tests were done. What is absent from the page is absent from the material’s evidence base.

What a CoA typically contains

The following recur across cell and tissue products in our supplier survey. Not all appear on any given CoA, and the ones that do not appear are the interesting ones.

Identity block

Line item	What it means	How it fails to mean what you think
Product name and catalogue number	Which SKU this is	A catalogue number is not a cell identity. Two SKUs can be the same cells with different fill volumes
Lot / batch number	The production run. The unit at which everything else on the page applies	Comparability is not implied between lots of the same catalogue number. This is the most consequential thing on the CoA and it is usually the smallest text
Donor identifier	Which human the material traces to	Frequently coded. Two lots from “the same donor” may be different tissue collections years apart
Cell type / lineage claim	What the supplier says it is	An assertion until an identity test on the page supports it
Passage number	Subculture count	Uninterpretable without a zero point and split ratio. See passage number and population doublings
STR profile	Short tandem repeat genotype for human line authentication	Confirms the line is what it claims against a reference profile. A profile with no reference comparison stated is a genotype, not an authentication

STR profiling is the one identity test that resolves a real and documented problem. Systematic investigation of widely used tumour cell lines has found substantial rates of cross-contamination and misidentification across the field, which is why authentication became a minimum quality-control expectation for banked cell lines rather than an optional extra. Consensus practice, per ANSI/ATCC ASN-0002, is a minimum of 13 loci with an 80 percent match threshold against a reference profile; next-generation sequencing approaches to the same question are established in the literature and give higher-resolution profiles.

Note what STR does not do. It authenticates a human cell line against a reference. It does not confirm the differentiated identity of an iPSC-derived product — an iPSC-derived neuron and an iPSC-derived cardiomyocyte from the same donor have identical STR profiles. For derived products the identity claim rests on marker expression, and marker expression is a much weaker instrument.

Purity and phenotype block

Line item	Question to ask about it	Why
Marker positivity, percent	Which markers, by which method, at which timepoint, and gated how?	“95 percent positive” with no named marker, no method and no timepoint is not a measurement. Flow cytometry and immunocytochemistry give different numbers on the same population, and gate placement moves the figure
Negative markers	Which contaminating populations were excluded, and were they looked for?	A purity figure that only counts positives cannot exclude a co-purifying population that is negative for the marker you stained
Morphology	Was this assessed against a reference image?	The softest line on the page. Meaningful as a sanity check, not as a specification

Purity is where CoAs are most often technically true and practically misleading. A product can be 95 percent positive for a lineage marker and still be functionally heterogeneous in ways that dominate your assay — subtype composition within the lineage is rarely reported, and is frequently the variable that actually differs between lots.

Safety and contamination block

Line item	What it establishes	The limit of the claim
Sterility	No bacterial or fungal growth under the stated method	Depends entirely on the method and sample size. A compendial 14-day test and a “no growth observed in culture” note are not the same evidence. See sterility and mycoplasma testing
Mycoplasma	Negative by the stated method	Mycoplasma testing is separate from sterility testing and neither covers the other. Method and detection limit determine what “negative” means
Endotoxin	Bacterial lipopolysaccharide below a stated limit	Rarely tested at all for RUO cell products. If there is no line, there is no result
Viral / adventitious agents	Screening for specific agents	Ordinarily absent from RUO CoAs. Its absence is normal and should still be noticed
Karyotype or CNV screen	Chromosomal integrity at a stated passage	Meaningless without the passage at which it was performed, since culture-acquired aberrations arise over time

Performance block

Line item	What it establishes	The trap
Post-thaw viability, percent	Membrane integrity of a sample by a stated dye at a stated time	The most over-read number in the industry. See post-thaw viability. Trypan blue reads high on post-thaw samples because it cannot separate cells from debris; time-after-thaw moves the figure substantially
Cells per vial	Count at freeze, or count post-thaw	These differ, and the CoA does not always say which. Ask
Plating / attachment efficiency	The fraction that attaches and stays attached at a defined timepoint	Far more predictive of your experiment than viability, and far less commonly specified — because it is harder to pass
Functional data	Whatever the supplier chose to show	Usually a representative figure from a development lot, not a measurement on your lot. Check which

What a CoA does not guarantee

It does not guarantee function. Nothing on a standard research-use CoA obliges the supplier to deliver a lot that performs in your assay. Viability is a membrane-integrity measurement; a cell can be committed to apoptosis, unable to attach, or phenotypically drifted and still score as viable.

It does not guarantee lot-to-lot equivalence. Two lots that both pass the same tests can behave differently. A CoA is a within-lot record, and comparability between lots is not one of the things it establishes. If your experiment spans lots, plan a bridging comparison.

It does not guarantee that untested attributes are acceptable. The absence of an endotoxin line does not mean endotoxin is low. It means endotoxin was not measured.

It does not guarantee that the material is fit for a regulated purpose. Grade and permitted use are governed by the licence and the product designation, not by the test results. See RUO vs GMP.

It does not guarantee the material survived shipping. A CoA is generated at release. Every excursion between the release bench and your freezer happened afterwards. This is why the temperature record in the shipment matters independently, and why the arrival check is your test rather than theirs.

It does not guarantee a remedy. A CoA states measurements. Whether a lot that under-performs entitles you to a replacement is a matter of the terms of sale, not of the certificate.

Four questions that expose a weak CoA

These are the questions we ask on a buyer’s behalf before an order is placed. They cost one email and they discriminate sharply between suppliers.

1. May I see a redacted example CoA for this SKU before I order? The single highest-yield question in cell procurement. A supplier who will send one is disclosing what they test as standard, before you are committed. A supplier who will not is telling you something. Ask for a real redacted lot certificate, not a template with the fields empty — the template shows the intended tests, the real one shows which fields actually get filled in.

2. Which of these tests are performed on every lot, and which on the master bank only? This is the question that changes what a CoA is actually evidence of. Karyotype, viral screening and detailed characterisation are commonly performed at the bank level and inherited by descendant lots. That is legitimate and economically necessary. But “tested” at the bank and “tested” per lot are different assurances, and a CoA that reproduces bank-level results without labelling them as such invites the wrong inference.

3. What are the acceptance criteria, and what happens to a lot that misses them? A test result without an acceptance criterion is data, not quality control. Ask for the numerical criterion and ask what the disposition procedure is for a failing lot. The answer tells you whether a quality system exists behind the certificate.

4. Will you notify me of a process change between lots? Not strictly a CoA question, but the one that determines whether your CoAs remain comparable over time. For research-grade material the supplier ordinarily owes you no notification and may change medium, coating or protocol between lots without telling you. Asking costs nothing, and it is the most common source of unexplained variance in a long-running programme.

Reading a CoA on arrival: a practical sequence

Check the lot number against the vial label and against your purchase order. A CoA for a different lot is a filing error that becomes an experimental error six weeks later.
Check the passage number against what your protocol assumes, and against the passage at which the supplier’s characterisation was performed.
Read the method column, not the result column. “Mycoplasma: negative” is a result; “Mycoplasma: negative, by PCR” is evidence with a known sensitivity profile.
List what is missing. Write down the tests you expected and did not find. That list is your risk register for this material, and it is the agenda for the supplier call.
Note the date of test relative to the date of freeze. A viability measured at freeze and a viability measured on a retained vial after storage are different claims.
File it against the lot, not against the product. CoAs are lot documents. A programme that files them by catalogue number has thrown away the only thing that makes them useful.

Boundary cases

A CoA is not a batch record. A batch manufacturing record documents how the lot was made, step by step, by whom, with which qualified materials. A CoA documents what was measured afterwards. GMP supply comes with both; research supply comes with, at most, the CoA.

“Tested to specification” without the specification is not a claim. If the supplier states that a lot met specification but will not state the specification, you have a pass/fail with an undisclosed threshold.

Third-party test reports are stronger than in-house ones, but rarer. Some suppliers subcontract mycoplasma or STR testing to an accredited laboratory and attach the report. That is a materially stronger document than an internal result, and worth asking about.

A CoA for a tissue model is a different animal. For engineered tissue products, barrier integrity, histology and viability at receipt matter more than most of the cell-product line items, and shelf life is short enough that the release date on the certificate is directly operational rather than archival.

The one-line summary

A certificate of analysis is a truthful record of the tests someone chose to run on your batch. Read the method column, list what is missing, ask which tests are per-lot and which are inherited from the bank — and never mistake a page of passing results for a promise that the cells will work.

Certificate of analysis: how to read one, and what it does not cover