What 85 peptide COAs show: identity, purity and amount against the label
An original data study of 85 certificates of analysis (COAs) for research peptides, read for three things: whether the lab found the labelled compound, the purity figure it printed, and the amount it measured against the label. Data as recorded through 2026-10-09.
Of 85 peptide COAs from two testing labs, 81 of the 82 reports with an identity verdict confirmed the labelled compound (98.8%, Wilson 95% interval 93.4% to 99.8%). 72 of the 85 printed a purity figure at all, and 69 of those 72 (95.8%) were at or above 98% by HPLC as reported by the lab. Across 81 reports with a usable amount, the median was 104.8% of label, and 77 of the 85 reports were chosen and submitted by sellers.
Summary
This study read and scored 85 certificates of analysis (COAs) for research peptides, issued by two third-party labs between July 2025 and October 2026. 81 of the 82 reports that carried an identity verdict confirmed the labelled compound (98.8%, Wilson 95% interval 93.4% to 99.8%). In one report the identity result did not match the label, and 3 more could not be checked. 72 of the 85 reports (84.7%) printed a purity figure at all, and 69 of those 72 (95.8%) were at or above 98% by HPLC as reported by the lab. The 3 below 98% all came from one protein product reported in IU, the lowest at 97.0%. Amounts ran above label more often than below. Across 81 reports with a usable amount, the median was 104.8% of label (mean 106.1%, SD 10.6). 70.4% fell within 90 to 110% of label, 25.9% were above 110% and 3.7% were below 90%. Missing information was common: 9 of 11 blend reports printed no purity figure. 77 of the 85 reports were chosen and submitted by sellers. Only 3 came from vials we bought anonymously ourselves, too few to compare.
Key figures
| Measure | Count | Share | Wilson 95% interval |
|---|---|---|---|
| Identity confirmed, of reports with a verdict | 81 / 82 | 98.8% | 93.4% to 99.8% |
| Identity confirmed, of all reports | 81 / 85 | 95.3% | 88.5% to 98.2% |
| Report prints a purity figure | 72 / 85 | 84.7% | 75.6% to 90.8% |
| Purity at or above 98% (HPLC, as reported), of reports with a figure | 69 / 72 | 95.8% | 88.5% to 98.6% |
| Purity at or above 98% (HPLC, as reported), of all reports (no figure counts as no pass) | 69 / 85 | 81.2% | 71.6% to 88.1% |
| Amount within 90 to 110% of label | 57 / 81 | 70.4% | 59.7% to 79.2% |
| Amount above 110% of label | 21 / 81 | 25.9% | 17.6% to 36.4% |
| Amount below 90% of label | 3 / 81 | 3.7% | 1.3% to 10.3% |
Methods
Source. Each row is one COA from one of two testing labs: Janoshik (55 reports) or Freedom Diagnostics (30). Reports came from three routes:
| Route | Reports | Who chose the vial |
|---|---|---|
| Seller-submitted | 77 | The seller |
| Buyer-submitted (a buyer’s own test in the lab’s public records) | 5 | A buyer |
| Blind purchase (we bought anonymously and sent the vial to the lab) | 3 | Us |
The 85 reports cover 6 suppliers. Suppliers are not named. This is a study of what goes on a COA, not a ranking of sellers.
Verification. Each report was checked against the lab’s own records, either its report lookup or the report file it issues, and the figures were transcribed into a tracker. One blend report’s lookup is still pending because its report key was hard to read. Lab records were searched only by the company names printed on the COAs.
Identity. “Confirmed” means the lab found the labelled compound: one lab prints an explicit identity line, the other names the compound it found in its findings section. “Did not match” means the lab found a different compound from the one on the label. “Not stated” means there was no label to check against or no readable identity line.
Purity. We used the purity percentage as printed. The threshold of 98% is our own reading line, not a regulatory standard. A report with no purity figure never counts as meeting it. The one undated report’s purity figure is counted in the purity figures.
Amount. Percent of label is measured content divided by labelled content. A blend counts once, as total measured over total labelled. A report is eligible for the amount figures only if it prints both a label and a measured amount, carries a date, and does not repeat another report. 81 of 85 were eligible. The 90 to 110% band is a reading aid, not a specification. Where a lab reported a two-vial mean, we used it as printed. The report whose identity did not match is included at its printed figure (about 97% of label), even though that figure measures a different compound from the one labelled.
Intervals. Shares carry Wilson score 95% intervals, which behave better than the simple normal interval at small counts and near 0% or 100%.[1] The standard deviation is the sample SD.
Findings
1. Identity: one mismatch in 82 verdicts
81 of the 82 reports that gave an identity verdict matched their label. In the other report, the lab found a different compound from the one labelled, even though its purity (above 99%) and amount (about 97% of label) looked unremarkable. A clean purity figure does not show that the vial holds the labelled compound. Only the identity line can confirm that.
Three reports gave no verdict, each for a different reason:
- one report printed no product name or label amount, so there was nothing to check the finding against;
- one was an endotoxin-only report with no content or identity figure;
- one had an identity line that could not be read on the copy available, and it also printed no date.
81 of 82 still allows a true confirmation rate as low as about 93%. And because sellers chose most of these vials, it says little about vials picked at random.
2. Purity: high when printed, and often not printed
| Purity figure (HPLC, as reported) | Value |
|---|---|
| Reports printing a figure | 72 of 85 |
| Median | 99.72% |
| Mean | 99.48% |
| Lowest | 96.98% |
| Below 98% | 3 (all one protein product reported in IU) |
Whether a purity figure appeared varied with the product type and the lab in this set:
| Report type | Printed a purity figure |
|---|---|
| Single compounds, excluding a non-peptide coenzyme product and one protein report measured by immunoassay, with a content figure | 68 of 68 |
| Blends (two or more compounds in one vial) | 2 of 11 |
| Coenzyme product (non-peptide) | 2 of 4 |
| Protein report measured by immunoassay | 0 of 1 |
| By lab: Janoshik | 43 of 55 (78.2%) |
| By lab: Freedom Diagnostics | 29 of 30 (96.7%) |
All 9 Janoshik blend reports and both Janoshik coenzyme reports gave no purity figure. The two Freedom blend reports and both Freedom coenzyme reports did. One protein report measured content by immunoassay and gave no purity figure. If a purity figure matters for your work, ask before testing which method the lab will apply to that product type.
3. Amount against label: centred above 100%
| Group | n | Median | Mean | SD | 90–110% | Above 110% | Below 90% | Range |
|---|---|---|---|---|---|---|---|---|
| All eligible | 81 | 104.8% | 106.1% | 10.6 | 70.4% | 25.9% | 3.7% | 83.5–153.9% |
| Single compounds | 70 | 104.0% | 105.7% | 10.3 | 74.3% | 22.9% | 2.9% | 83.5–153.9% |
| Blends (total) | 11 | 110.0% | 109.2% | 12.8 | 45.5% | 45.5% | 9.1% | 83.9–127.5% |
About three in four eligible reports (60 of 81, 74.1%) measured above label. Underfill below 90% was rare (3 reports). Overfill beyond 110% was common in this set, especially in blends, where 5 of 11 measured over 110%. Blend totals can hide uneven parts: in 3 of the 5 blends with a copper-peptide component, that one component measured 119% to 140% of its own labelled share, as reported and with copper included; one of these reports is still awaiting a lookup check. The single highest reading was a single compound at 153.9% of label.
Excluding the coenzyme product and the four protein reports, which are reported in larger units or in IU, the single-compound figures tighten slightly: n = 62, median 103.8%, with 80.6% within 90 to 110% (Wilson 69.1% to 88.6%).
By lab, the Janoshik median was 106.7% (n = 54) and the Freedom median 103.1% (n = 27). The two labs largely tested different products, so this gap should not be read as a difference between the labs.
4. Missing or unusual fields
| Issue | Reports |
|---|---|
| No purity figure | 13 |
| Product and label not printed | 1 |
| No date printed | 1 |
| Endotoxin-only report (no content figure) | 1 |
| Identical figures under two different lab references | 1 |
| Label amount inferred, not printed as a figure (10 from a product code, 1 from the client name) | 11 |
| Content reported as a two-vial mean | 13 |
The duplicate matters. Two lab references showed identical figures for the same product, which suggests either one vial filed twice or a reissued report. We left the repeat out of the amount figures. It still sits in the identity and purity counts; without it those would read 80 of 81 confirmed and 68 of 71 at or above 98%. If you are tallying a seller’s COAs, check for repeats before counting them as independent tests.
5. By route: too few independent reports to compare
| Route | Reports | Identity confirmed | Purity ≥98% by HPLC as reported, of those printing | Amount median |
|---|---|---|---|---|
| Seller-submitted | 77 | 73 of 77 | 61 of 64 | 104.8% (n = 73) |
| Buyer-submitted | 5 | 5 of 5 | 5 of 5 | 104.8% (n = 5) |
| Blind purchase | 3 | 3 of 3 | 3 of 3 | 103.1% (n = 3) |
The 3 blind-purchase reports confirmed identity, printed purity at or above 98%, and measured 99.3%, 103.1% and 115.6% of label. With n = 3, the Wilson interval for “identity confirmed” runs from 43.9% to 100%, so these reports cannot show whether vials sellers choose differ from vials bought anonymously. The one identity mismatch was, as far as we can tell, a seller-submitted report; its route was inferred from the company printed on it.
What this does not show
- It is not a random sample. Sellers chose 77 of the 85 vials and decided which reports to share. Reports that read badly may never have reached us.
- It does not describe stock. A COA describes what was in the tested vial on the day the lab tested it. It says nothing about other vials, other lots, storage or shipping.
- Seller-submitted vials were not re-tested. We confirmed that each report is listed in the lab’s records and copied its figures. We did not repeat the analysis.
- It is not a lab audit. We did not assess either lab’s methods, calibration or accreditation. One of our own blind submissions listed the wrong label amount on the request form. The label was corrected in the tracker to the vial’s printed amount, a reminder that COA fields reflect what the submitter wrote.
- It covers three things only. Identity, purity by HPLC as reported, and amount against label. Most reports did not cover endotoxin, sterility, heavy metals, residual solvents, counter-ion content or water content. Only 4 reports in the set included an endotoxin test.
- It does not compare suppliers or labs. Pooled figures weight each supplier by how many reports it shared, and the two labs saw different products.
- It says nothing about any use outside the laboratory.
How to read any COA: a checklist
- Lab named, and verifiable. The report names a real testing lab, and that lab offers a public lookup. Open the report from the lab’s own site, not from a screenshot.
- The report matches the vial. The product name and labelled amount are printed and match the vial in hand. A COA with no product name cannot confirm identity.
- Identity method. What method confirmed identity (mass spectrometry, for example), and does the report say the labelled compound was found? A high purity figure on its own does not answer this.
- Purity method and figure. A percentage with its method named, usually HPLC. “Not reported” is not a pass. Blends and some compound types often carry no purity figure.
- Amount against label. Divide measured by labelled. Note whether the figure is one vial or a mean of several, and the unit. Content in IU (international units) comes from a different kind of assay than content in mg and should not be pooled with mg figures without care.
- Date. The analysis or receipt date. An undated report cannot be placed against a lot.
- Who submitted it. A seller, a buyer or an independent buyer. The route tells you who chose the vial.
- Repeats. Identical figures under two report numbers are one test, not two.
- What was not tested. Endotoxin, sterility, heavy metals and residual solvents need their own tests. If the report does not list them, that report did not test for them; ask whether a separate report exists.
Data and how to cite
The aggregate tables on this page are the full output of the analysis. Row-level data is not published because it would identify suppliers. Corrections are welcome and will be dated on this page.
What 85 peptide COAs show: identity, purity and amount against the label. Original data study, October 2026. Figures are drawn from 85 third-party COAs recorded through 2026-10-09; intervals are Wilson 95%.
Questions people ask
How often did a peptide COA confirm the labelled compound?
81 of the 82 reports that carried an identity verdict confirmed the labelled compound, 98.8% (Wilson 95% interval 93.4% to 99.8%). In one report the identity result did not match the label, and 3 more could not be checked. Sellers chose 77 of the 85 vials, so these are not random-sample rates.
Does a high purity figure show the vial holds the labelled compound?
No. In the one mismatched report, the purity and amount looked unremarkable, yet the lab found a different compound from the one labelled. Only the identity line can confirm that.
How many of the COAs printed a purity figure?
72 of the 85 (84.7%). 69 of those 72 (95.8%) were at or above 98% by HPLC as reported, and 9 of 11 blend reports printed no purity figure. Sellers chose 77 of the 85 vials, so these are not random-sample rates.
Did measured amounts run above or below the label?
Above more often than below. Across 81 reports with a usable amount, the median was 104.8% of label; 70.4% fell within 90 to 110% of label, 25.9% were above 110% and 3.7% were below 90%.
Who chose the vials behind these COAs?
Sellers chose and submitted 77 of the 85. 5 were buyers’ own tests in the labs’ public records, and 3 came from blind purchases, too few to compare with the seller-submitted reports.
What does a COA not tell you?
A COA describes what was in the tested vial on the day the lab tested it, not other vials, other lots, storage or shipping. Most reports here did not cover endotoxin, sterility, heavy metals, residual solvents, counter-ion content or water content.
Sources
- Wilson EB. Probable inference, the law of succession, and statistical inference. J Am Stat Assoc. 1927;22(158):209-212.