Insyder Score v0.3
Fixed rules, public label data, and a complete accounting of what each point means.
Universal scoring
| Axis | Points | Computation |
|---|---|---|
| Dose adequacy | 40 | Label dose (NIH DSLD) vs. the clinical reference floor/ceiling (data/clinical-doses.json). Full points at or above the floor; linear below it; capped at the ceiling — mega-dosing earns no bonus. A proprietary blend that hides the dose scores 0 on this axis. |
| Form quality | 15 | Per-ingredient form hierarchy (e.g. magnesium glycinate/citrate outranks oxide; creatine monohydrate is the reference standard — "fancier" forms don't score higher, but do cost more, which shows up in the value axis instead). B12 form is not scored — NIH ODS: absorption doesn't vary by form. |
| Third-party verification | 25 | NSF Certified for Sport / Informed Sport / USP = 20 pts, checked against the certifier's own database, never brand copy; +5 for a published batch Certificate of Analysis; 0 if none found. NSF (and, for omega-3, IFOS) are checked automatically; Informed Sport and USP are both structurally site-blocked to automated checks today and are verified manually for each category's top 10 before a full top-3 publishes. |
| Label transparency | 10 | Full label disclosure = 10; a proprietary blend that obscures amounts = 0–3; undisclosed "other ingredients" issues are deducted. |
| Value (cost per effective dose) | 10 | Price ÷ servings-at-clinical-dose (not per label serving), percentile-scored within the category's priced cohort and shown as a dollar figure on the product's receipt. The price join (Shopify Global Catalog MCP) runs over each category's top-ranked candidates, so coverage is partial and thins with rank — priced today: protein powder 28 of 1,182, creatine 33 of 246, fiber 20 of 228, and omega-3 33 of 438. A product with no confident catalog match shows "unverified" here rather than a guessed score, as does everything in electrolytes and multivitamin: those prices are joined, but cost-per-effective-dose needs a single clinical dose target to divide by, and neither multi-active category has one (v0.2 changelog below). |
An additive, per-category overlay scoring the judgment a generic supplement ranking doesn't make: GLP-1-specific tolerability and clean-label signals (protein's sugar-alcohol/thickness penalties, electrolytes' clean Na/K/Mg band, fiber's soluble-first preference, and so on). Baseline 10; +3/+5 for a positive signal, −3/−6 for a negative one; clamped to the 0–20 range.
Clinically informed, product-validated. Our electrolyte band is bounded using WHO/UNICEF oral rehydration guidance (the clinical gold standard for GI-fluid-loss electrolyte replacement, scaled down for a chronic/daily rather than acute-rehydration context), NIH dietary reference intakes (daily sodium, potassium, and magnesium ceilings/allowances), and ACSM exercise-fluid-replacement guidance (a different context, used only to anchor a sensible floor). No clinical guideline addresses GLP-1-specific daily electrolyte-drink dosing directly — this band is our own reasoned synthesis across adjacent literature, not a single clinical lookup. It is cross-checked against, not copied from, established clean-label products like LMNT. Full derivation and source table: research/electrolyte-band-derivation-2026-07-15.md.
Each category's candidate set comes from the NIH DSLD's own relevance-sorted search results per search term, taking up to the top 300 hits per term (6 pages of 50). This cap is uniform across every category and applied before scoring, never after — results beyond it are deep-tail fuzzy matches our brand-normalization filter would reject anyway.
Label data: NIH Dietary Supplement Label Database (DSLD), a free, public, structured API (api.ods.od.nih.gov/dsld). Price/availability: Shopify Global Catalog MCP, joined for each category's top-ranked candidates (partial coverage — see the value row above) — price affects only the value axis, and only after every other axis is frozen. A small number of top-selling brands with zero DSLD hits are scored from manually double-transcribed labels instead (capped at 5 products per category), marked "label transcribed from brand site" everywhere they appear.
Every score on this site is computable from the two public data sources above plus the axis definitions and weights on this page — you don't need access to our systems to check our work. Pull a product's label from DSLD, apply the dose/form/transparency rules above against data/clinical-doses.json's published reference values, check the certifier's own public database for verification status, and you'll land on the same number we published.
Version history
Dose-floor gate. A product whose measured dose does not reach the clinical reference floor for its category is no longer eligible to be shown as one of our picks, whatever its total score. No score changed: every axis and total computes exactly as before, and the full scored table still lists these products with their real numbers — this affects only which products can be presented as a recommendation. The change follows a case where an omega-3 carrying 900mg EPA+DHA against a 1,000mg floor rose to the top of its category, ahead of a product carrying 2,020mg, because it was the only certified one. Leading with a product that misses the floor contradicts the reason we publish the floor. A dose that is absent, hidden inside a proprietary blend, or has no clinical floor defined is treated the same way, and says so — not verified and underdosed are different findings, and we show which one applies.
Potassium band widened 200→[200,400]mg per an independent derivation (WHO/UNICEF oral rehydration salts' sodium:potassium ratio, ~2.2:1 by mass, vs. the prior band's 5:1 ratio) — products between roughly 201–400mg potassium per serving that previously scored as overshooting the clean-label band now score as fully within it. Electrolyte band provenance relabeled "clinically informed, product-validated" (from "benchmarked to established clean-label products, not a clinical guideline"); sodium and magnesium numbers are unchanged. Evidence: research/electrolyte-band-derivation-2026-07-15.md. No axis-weight changes.
Dose axis is N/A ("unverified") for electrolytes and multivitamin — no single clinical dose target exists for either multi-active category; their signal lives in the GLP-1 overlay instead. Removed the B12 methylcobalamin-over-cyanocobalamin form preference (NIH ODS: absorption doesn't vary by form). Multivitamin's "methylated Bs" overlay bonus is suspended pending independent verification. Overlay magnitudes use a documented scheme (baseline 10, +3/+5, −3/−6, clamped 0–20). No axis-weight changes.
Initial proposal.