COA Gap Index · version 0.2 · 26 September 2026

Third party tested? 3 of 50 top-selling saffron and NMN supplements let you check

By JT Root, founder of Numi Naturals and a NASM certified nutrition coach. Published 26 September 2026. Every row checked 26 September 2026. Next recheck planned for December 2026.

Third party tested is printed on most supplements sold in the United States. It tells you someone outside the company ran a test. It doesn't tell you whether you can see the result, or whether the result is for the batch you bought.

The COA Gap Index measures that gap on the products people actually buy. Every row records what a shopper could open on a stated date, and the full dataset is free to download and reuse.

What are the key figures?

• 82%say third party tested on their own website (41 of 50).
• 22%publish a certificate of analysis anyone can open, with no login, email or chat (11 of 50).
• 6%let a shopper match a certificate to the lot on the bottle (3 of 50).
• 73%of the products that claim third party testing publish no certificate at all (30 of 41).
• 4 of 11published certificates don't report a measured amount of the main ingredient.
• 1 of 11certificates includes a separate identity test showing the ingredient is what the label says.
• 0 of 11certificates cover all five basics: a named lot, an identity test, a measured amount, heavy metals and microbiology.
• 5 of 11certificates show no laboratory accreditation number. One was issued by the brand's own quality team.
• 5 of 50offer a certificate only if you ask for it.
• 8 of 60best-sellers have no brand website that could be found, so there is nowhere outside the marketplace to look.
• 8 vs 3NMN brands published certificates more often than saffron brands (8 of 25 against 3 of 25).
• 1NMN best-seller's “View Certificate of Analysis” link opens a report for a different compound, NMNH.

Which best-sellers let you match a certificate to your bottle?

3 of the 50: Nootropics Depot, partiQlar and Renue By Science. Each one lists its reports by lot or batch number, so a shopper can start from the number printed on the bottle and open the report for that batch.

Nootropics Depot lists 10 saffron lot numbers on the product page, each linked to a signed report. partiQlar keeps a list of batch reports on its quality page and tells readers to match the number on the bottle. Renue By Science shows the batch now shipping on the product page, with its reports, and lists earlier batches on a testing page.

Even these 3 don't publish everything. Only the Nootropics Depot report includes a separate identity test, and it has no microbiology. partiQlar's latest report measures NMN and nothing else. The detail is in the certificate table below.

What do the published certificates actually show?

A certificate is only as useful as what it measured. Each of the 11 certificates was read for six things: whether it names a lot, whether it has an identity test, whether it reports a measured amount of the main ingredient, whether it covers heavy metals and microbiology, and whether it shows the laboratory's accreditation number.

7 of the 11 report a measured amount of the main ingredient. The other 4 don't: one gives the saffron amount as “Quantity by Input”, a figure calculated from the batch records rather than measured; one records NMN potency only as “Pass”; and two report contaminants but no amount at all.

Only 1 certificate has a separate identity test, the check that the ingredient is really saffron or NMN. One NMN certificate is titled “Identity by HPLC” but reports only the amount. 2 name no lot, so no shopper can tie them to a bottle. 5 show no accreditation number for the laboratory. One was prepared by the brand's own quality team, and one comes from a laboratory the brand's own site says was set up to test its products.

Two findings stood out. On one NMN product page, the link labeled “View Certificate of Analysis” opens a manufacturer's report for NMNH, a different compound. Another NMN brand's metals and microbiology report is headed “Preliminary Results Only”.

Brand names link to the certificate that was read. Where a brand publishes several reports, the most recent one for the product was read.

Which brands are in the index, and what did each one show?

Saffron supplements, 30 best-sellers

NMN supplements, 30 best-sellers

Brand names open the page that was checked, or the Amazon listing where there was no brand website. Links are marked nofollow, the convention for an audit, and carry no judgment about the site.

How were the 60 products chosen and checked?

The sample is what shoppers are shown first. On 26 September 2026, Amazon.com was searched for saffron supplement and NMN supplement, sorted by Best Sellers, and the first four pages of results were saved. Sponsored placements were removed. Products that don't contain saffron or NMN, including listings sold as NMN alternatives and pet products, were excluded. The first 30 unique brands in each search were kept, one product per brand, its highest-ranked listing.

Each product was then checked on the brand's own website, never on Amazon. Three things were recorded. Does the product page, or a quality page linked from it, say third party tested or independently tested? Can a certificate of analysis be opened without a login, an email form or a chat? And can a shopper start from the lot number on the bottle and reach the report for that lot?

When nothing was found, the search followed a fixed path and was recorded: the product page and its links, the header and footer, the site's own search, and the usual addresses for certificate pages. The time allowed for each brand was capped, so every brand got the same search.

Every certificate found was then read for its lot, identity test, measured amount of the main ingredient, heavy metals, microbiology, laboratory and accreditation number. No products were bought and no laboratory tests were run.

How reliable are these figures?

50 products is a sample, so each percentage has a margin. At 95% confidence, the true share claiming third party testing is between 69% and 90%. The share publishing a certificate is between 13% and 35%, and the share allowing a lot match is between 2% and 16%. Even at the top of that last range, most best-sellers don't let a shopper check their own bottle.

20 of the 60 rows were checked a second time without sight of the first result, and every certificate was read twice. The two checks agreed on 57 of 60 recorded answers, or 95%. The 3 disagreements were settled by opening the pages again. One claim had been missed because it was written as 3rd party. After that, every product recorded as making no claim was searched again, and one more claim was found.

Who ran this, and what is the competing interest?

Numi Naturals ran it. We sell a saffron supplement, Saffron360, and an NMN supplement, NMN360, so this is a plain conflict of interest rather than a footnote.

Three things were done about the conflict. The sample comes from Amazon's own ranking rather than our choice of brands. No Numi product is in it: Saffron360 appeared in the saffron results only as a paid placement, which the method excludes. The method, every row and the raw data are published so anyone can redo the work.

Our own certificates don't pass all five basics either. The Saffron360 report confirms identity but doesn't measure the saffron amount. The NMN360 report measures NMN at 533 mg against the 500 mg on the label but has no separate identity test. Both name the lot and include heavy metals and microbiology.

Which limitations does this version carry?

Amazon personalizes results, so another shopper on the same day may see a slightly different order. The results were captured from a signed-in US account and the saved lists are kept with the data. The ranking reflects Amazon's Best Sellers sort on the day it was captured, not total sales.

For 5 brands the exact Amazon product couldn't be matched on the brand's website; 3 don't list it at all. For those, the brand's site-wide statements were scored, and each is flagged in the data. 4 brands appear in both categories and are counted once in each.

A website can change between the check and the reading. Brands add certificate pages and sometimes remove them, which is why every row carries its date and why the plan is to recheck every quarter on the same method.

The COA Gap Index doesn't test whether a published certificate is accurate. It records what a shopper can see.

How can this data be used and cited?

The data and charts are published under a Creative Commons Attribution 4.0 license. You can reuse, chart and republish them, including commercially, as long as you credit the source.

What has changed, and what has been corrected?

• Version 0.2, 26 September 2026. Rebuilt on a stated sample: the top 30 brands per category in Amazon best-seller results, 60 products in all. Added the certificate content checks, a blind second check of 20 rows, confidence ranges and this citation section.
• Version 0.1, 21 September 2026. A pilot of 17 hand-picked products, including 2 of our own. Superseded by version 0.2, whose sample and method differ, so the figures should not be compared directly.
• Corrections. None so far. Brands can send a correction with a link to hello@numilove.com.

Common questions about certificates and lot numbers