A claim's grade is the minimum of its seven basis dimensions — never the average, never the maximum. One weak dimension caps the whole claim. This is the opposite of how most evidence products grade: they score strengths and let volume of studies carry confidence. Here, 39 studies cannot lift a claim past a VERY_LOW confidence dimension.
Averaging rewards bad evidence when it's plentiful. A meta-analysis sitting on poorly-reported trials would grade HIGH on volume alone. Under weakest-link rules, the claim cannot out-read its weakest assessed dimension — the grade is the most the data can honestly support, not the most the field happens to have produced. Graders who disagree can't inflate; they can only fail to submit.
Below is the actual basis grid of a released turmeric claim (CLM-6a6b7e456bdbc1b7). Design evidence is HIGH — randomized and observational designs are bound. But extraction confidence is VERY_LOW: too little verified extraction detail survives audit. The ceiling lands on the weakest cell.
You can trust the label. A VERY_LOW grade is not a shrug — it is an audited statement that the evidence, as it exists and as it was reported, cannot support more. Products that grade everything HIGH are telling you about their marketing. A ceiling tells you where the honest edge of knowledge sits — which is the information a formulation, dossier, or regulatory team actually needs.