1. We do not test the product
Nothing we do involves a laboratory. We do not verify ingredient identity, potency or purity, we do not test for contaminants or banned substances, and we do not inspect a facility. If the bottle contains something other than the label says, our review will not find it and is not designed to.
That work is done well by others: NSF, USP, Informed Sport and Clean Label Project all test what is in the bottle. We read what is said about it. Those are different jobs and a brand serious about both should buy both.
2. We do not evaluate whether a claim is scientifically true
This is the limit most often misread, so it is worth stating bluntly. The report is evidence that the wording was reviewed against a published rule. It is not evidence that the science is settled.
If a claim requires substantiation, we tell you the standard it has to meet and whether the wording as published would satisfy a reader applying that standard. We do not hold your study file, we do not assess your study's quality, and a brand that sends our report to a regulator in place of its own substantiation is worse off than one that sends nothing, because it looks like an answer and is not one.
3. We cannot promise any outcome
We do not control Amazon, Meta, Google, your payment processor, the FDA or the FTC. They apply their own judgment, change their policies, and act on things that have nothing to do with your copy. A passing verdict is an assessment of your wording against our standards on the date we read it. It is not a shield, and anyone who tells you otherwise is selling you something they cannot deliver.
We also never tell you enforcement is coming. Nobody knows that. We report what a regulator or a platform has already published, and what remedy each one names in its own rules.
4. We only see what we are given
An audit covers the surfaces you send us. An ad you did not include, a page we were not shown, an influencer post, a founder's own social account, a reply in a customer service inbox: none of those are reviewed, and none of them are covered by a passing verdict. The scope statement in every report lists exactly what was read.
5. A verdict has a date on it, and copy moves
Every finding is true of the copy as it stood on the scan date, against the standards in force on that date. Both halves drift. An agency pushes new creative, an A+ module changes, a seasonal ad reintroduces the phrase the audit removed. Separately, the rules themselves move: Amazon's supplements policy says in its own text that it is changeable without notice. That is what monitoring is for, and a brand that does not take monitoring should treat the report as a photograph rather than a state of affairs.
6. Where our data is thin
We publish a category census and a per-category risk rank, and both have edges worth knowing:
- A rank is only meaningful inside its own category. Categories were scanned weeks apart and on two different versions of the standard, so a rank in one category is not comparable to a rank in another, and we never write it as though it were.
- Some categories are small. Joint carries 29 listings. A rate computed on 29 listings has a margin of error wide enough that we would not build an argument on it, and neither should you.
- The census covers top-selling listings, not every listing. It says nothing about the long tail, and we do not claim it does.
- It is Amazon. A brand whose real risk sits on its own site, in email, or in ads is only partly described by it.
7. We are not lawyers
Nothing we produce is legal advice and no attorney-client relationship exists. We do not represent anyone before any agency in any country. Plenty of clients hand the report to counsel, which is usually cheaper than having counsel produce it, and that is the right use of it.
8. We get things wrong
We do, and when we do we publish it. Every figure we have withdrawn or restated is on the corrections page with the date and the reason, including mistakes nobody outside the company would ever have seen. If you think a finding is wrong, the reviewer who signed it will walk through the reasoning and revise it where you can show substantiation we had not seen. Revisions within thirty days are included.
Why this page exists
A firm that sells verification has one unforgivable failure, which is overstating its own work. Publishing the limits is not modesty and it is not a legal hedge: it is the same mechanism the product runs on, which is that a claim you can check beats a claim you have to trust. If any of the above is a problem for what you need, we would rather you knew before you paid than after.
The full scope and method are on the standards page, the complete sample report is public including its own limitations, and how to verify us covers checking any of this without taking our word for it.