Scoring methodology

Planned Phase 2 model · Not active in the current release
Algorithm version: Not released · Reference-data version: Not released · Reviewed: 7 August 2026

This page publishes the proposed scoring contract before the feature ships. It must be updated from the tested @shrayk/scoring package—and its fixtures must match the app—before any score is shown to users. The description below is therefore a specification, not a claim that current products are being rated.

The model rates a packaged product from available label data. It does not rate a person’s diet and is never medical, nutritional, allergy, or dietary advice.

Output and bands

The planned output is Shrayk’s own score from 0 to 100. Higher values indicate a more favourable result under this model, not that a product is “safe,” “healthy,” suitable for unlimited consumption, or appropriate for an individual medical need.

| Score | Band | Meaning in the interface | | ------ | --------- | ----------------------------------------------------- | | 75–100 | Excellent | Most favourable band under this model | | 50–74 | Good | More favourable than the lower bands under this model | | 25–49 | Poor | Material factors reduced the result | | 0–24 | Bad | Strongest reduction under this model |

The interface always shows the label with the colour. It never communicates a band by colour alone.

Components and proposed weights

The planned total combines three versioned components:

1. Nutritional quality — 60%. The specification currently calls for the original Nutri-Score nutritional computation, including its beverage, added-fat, and cheese variants, transformed into Shrayk’s own 0–100 component.
2. Additive context — 30%. A reference table assigns a documented risk tier based on named regulatory and scientific assessments.
3. Organic status — 10%. A verified organic status adds 10 points within the component design.

Shrayk will not display Nutri-Score letter grades or the Nutri-Score logo. Nutri-Score is a separate front-of-pack system and trademark. The current specification’s reference to its original computation must be reassessed before launch because the official algorithm has since been updated.

Additive penalties and cap

The proposed additive tiers are:

| Tier | Proposed effect | | ---------------- | -------------------------------- | | None identified | 0 | | Limited context | −6 | | Moderate context | −15 | | High context | −30 and final score capped at 49 |

A tier is not a diagnosis and does not mean an additive makes a product “dangerous.” Hazard classifications, exposure assessments, acceptable daily intakes, data gaps, and regulatory restrictions answer different questions. The reference table must record the source, date, assessment scope, rationale, and uncertainty for every classification.

Organic component

A verified organic designation contributes the proposed +10 component. It does not override nutrition or additive context, does not prove an individual product is healthier, and cannot remove the high-tier cap.

Not enough data

Shrayk must return Not enough data rather than estimate a score when required nutrition or ingredient fields are missing or inconsistent. The following categories are unratable under the planned model: alcohol, supplements, infant formula, and pet food. Other exclusions can be added only through a versioned methodology change.

Deterministic implementation

The score will be produced by deterministic TypeScript code, never by a generative model. The same input and the same algorithm/reference versions must return the same result in the app, on the web, and in tests. Automated models may help extract label text in a future consent-based contribution flow; extracted values must be reviewed and the model cannot choose a score.

Each factor sheet will show the algorithm version, reference-data version, available inputs, component effects, cap if applied, and links to additive sources.

Governance and conflicts

No brand can pay to alter a product score, suppress a factor, or move an alternative. Shrayk has no paid score placement, ads, or affiliate ranking. Material changes require a dated changelog, updated test fixtures, and an explanation of whether existing products are recalculated.

Source policy

Reference data will prioritise primary assessments from EFSA, IARC, ANSES, and EFSA’s OpenFoodTox database. IARC classifications identify carcinogenic hazard; they do not by themselves measure the risk from consuming a particular amount of a food additive. The sources page explains how each source is used and links to the official material.

Before this can launch

The final @shrayk/scoring implementation, test fixtures, additive reference table, legal review, and app/web parity check are not present in this website repository. Accordingly, no score demo on this site computes a product result today.