Science you can inspect.
Validus is an evidence-informed educational system. It evaluates food quality and the context around a food, and it never claims to know the effect a food will have in your body. This page shows how a conclusion is reached, so you can judge the reasoning and not just the result.
Where the evidence is strong, we say so. Where it is thin, we say that too.
The model in layers
Five layers, each doing one job. Meal Balance is now live through Meal IQ; Eating Pattern remains in build and is labeled that way everywhere on this site.
| Layer | What it does |
|---|---|
| Food Quality Live | Evaluates the food itself: ingredients, degree of processing, and nutrition composition. This layer is locked to the product and does not move with who is looking at it. |
| Personal Fit Live | Re-weights the same quality signals against the health goal you pick, so the fit can change while the food's quality stays fixed. Health goals are a PRO feature. |
| Meal Balance Live | Describes how the foods in one logged meal combine, which parts carry it, and where one adjustment could make it stronger. |
| Eating Pattern Upcoming | Will summarize repeated choices across days and weeks, instead of reading anything into a single scan. |
| Next Best Action Live | Turns the result into one practical next step: pair it, time it, swap it, or keep it. A written next step comes with every scan. |
Evidence standards we publish
We weight conclusions according to evidence strength, relevance, dose/context, and consistency. Stronger and more directly relevant human evidence moves a conclusion further than early or indirect work. Consensus guidance is context around a conclusion, not a substitute for the studies behind it.
High
Consistent findings across multiple human studies. Ingredient analysis in the app labels this "Well Studied".
Moderate
Real human evidence with gaps: mixed results, narrow populations, or doses unlike normal eating. Labeled "Moderately Studied".
Emerging
A very new area with preliminary findings only. The direction is plausible and far from settled. Labeled "Emerging Research".
Insufficient
Too little to support a confident conclusion: few human studies, mostly observational or animal data. We publish that instead of guessing. Labeled "Limited Evidence".
What we are uncertain about
These are published limits of the data, not fine print. Uncertainty is part of the output.
- Missing quantitiesLabels list ingredients in order, not amounts. When a quantity is not published, we treat it as unknown rather than assume one.
- Formulation changesBrands reformulate. Product data can be older than the package in your hand, which is why every product page accepts a correction.
- Serving contextA score reflects the serving as published. Portion size, the rest of the plate, and timing are context a barcode cannot carry.
- Individual variabilityStudies describe averages across groups. Individual responses vary, and a scan cannot measure yours.
Versioning
Scoring is a model, and models change. The version behind every result is public, and so is what changed.
Current scoring model: v1.0, effective April 24, 2026.
- v1.0 · April 24, 2026Matrix-based dimension scoring, a per-signal baseline anchored to the Validus Standard goal, an exception that preserves whole foods, and a category-fit principle so a food is judged against its own category.
Prior versions predate the public changelog. If a product or a score looks wrong, corrections are accepted in-app on every product page, or by email to support@validushealth.ai.
Going deeper
Seed oils, emulsifiers, sweeteners, degree of processing, and certification labels each need more room than a method page. Those sit in Learn, starting with seed oils.
