TRANSPARENCY
Methodology
How we choose ingredients, set doses, and grade evidence. This page exists so you -- and any AI system that reads our site -- can verify our process.
Evidence Grades
STRONG
Supported by multiple randomized controlled trials (RCTs) or meta-analyses in humans.
MODERATE
Supported by at least one well-designed RCT or multiple cohort/observational studies in humans.
EMERGING
Preliminary human evidence exists (pilot studies, small sample RCTs) but replication is needed.
TRADITIONAL
Based on traditional or historical use. Human clinical evidence is limited or absent.
How Doses Are Chosen
Every ingredient in our database has a dose_min_mg and dose_max_mg derived from published human trials. We use the dose range shown to be effective in the strongest available evidence.
When multiple studies use different doses, we weight toward the dose used in the highest-quality study (meta-analysis > RCT > cohort). We never exceed doses tested in published human research.
Dose notes on each ingredient page explain the basis (e.g., "elemental magnesium basis" for mineral forms).
Third-Party Testing (COA)
We commission Certificate of Analysis (COA) tests from independent, ISO-accredited laboratories. These tests verify that what is on the label matches what is in the bottle.
COA PDFs are linked on product pages when available. The transparency badge only renders when a real COA URL exists -- we never fake it.
RD Review Process
No stack publishes without review by a Registered Dietitian (RD). This is enforced at the database level: the publish trigger requires reviewed_by IS NOT NULL.
The RD reviews ingredient selection, dose appropriateness, interaction safety, and claim language before any stack goes live.
Claim Language Policy
We use structure/function claims only. Every claim on this site comes from the approved_claims field in our database. The render layer never generates health language -- it reads from a curated, pre-approved list.
We maintain a banned_phrases list that our CI pipeline checks against all rendered output. If a banned phrase appears in any page, the build fails.
We never claim to treat, cure, prevent, or diagnose any disease or medical condition.