LOADINGLoading your Pro surface...
Every Scan Dose verdict is a 0–100 composite of five axes. The same five inputs power every score, every time. No paid placement, no editorial reranking. The inputs are deterministic given the product label and the user's profile.
Full source-list at /our-data. Full benchmark + grader rubric at /methodology.
Per-ingredient evidence grade (A / B / C) drawn from RCT count, meta-analysis presence, and NIH ODS classification. Ingredients with strong-evidence claims at the trial-effective dose contribute positively; ingredients with debunked or limited evidence reduce the score.
Source: 834+ researched ingredients · NIH ODS · PubMed
Each active ingredient is compared against the minimum effective dose from clinical trials. Underdosing is the most common reason a scan loses points. Hidden 'proprietary blends' are flagged when the per-ingredient amount cannot be verified.
Source: Per-ingredient clinical-trial dose ranges from the research corpus
When the user's profile lists medications, every ingredient is checked against the verified interaction database. Severe interactions (e.g. 5-HTP + SSRIs, St. John's wort + birth control) cap the score regardless of other axes.
Source: 177 published interactions · 128 unique PMIDs
Pulls from independent lab-test data (ConsumerLab + LabDoor + USP). Products that have failed third-party content or contamination tests are penalized. Products without any third-party data carry a small uncertainty discount.
Source: 373 failed products in surfaced corpus · independent lab testing
Cross-references the FDA recall and CAERS adverse-event databases. Products in active or recent recall windows are flagged. Brands with disproportionate adverse-event signals carry a brand-level penalty.
Source: 2,338 FDA recalls · 2,061 tainted-product records · 772 CAERS events
Documented serious-adverse-event or contraindication risk. Combinations like 5-HTP + SSRI (serotonin syndrome) or red yeast rice + statin (rhabdomyolysis). Verdict caps + a clear safety-first surface.
Clinically significant interaction or absorption issue with established literature. Examples: levothyroxine + calcium (4-hour separation), warfarin + high-dose fish oil (bleeding risk).
Mechanism-plausible interaction with limited or mixed clinical data. Worth flagging; not a stop sign. Surfaces a discuss-with-prescriber prompt without overstating risk.
Theoretical or rarely-clinical-grade interaction. Logged but not surfaced as a primary warning unless the user explicitly asks.
Every claim backed by a Scan Dose interaction or evidence grade is tagged with the strongest tier of supporting evidence we can verify. Higher-tier evidence weighs more heavily in the final score.
Each scan returns a numeric confidence score derived from how many of the five axes had high-tier supporting evidence. When the evidence is genuinely contested, mixed, or missing, Scan Dose says so, explicitly, instead of generating a confident-sounding answer.
This rule is consistent with the safety-critical-omission hard-fail in the public benchmark. See /benchmark for the rubric.
Snapshot date: 2026-05-03. Weights re-tuned at most quarterly with version notes published at /methodology.