The Vitagri Ontology

Commitment 01

Nutritional quality is a property of a consignment, not a description of a category

“Nutritious” is usually an adjective attached to food types. Broccoli is nutritious; crisps are not. In Vitagri's model a nutritional value belongs to a specific consignment of a specific commodity, from a specific place and time. It carries the nutrient, the unit, the basis (fresh or dry weight, edible portion), the method, the laboratory, the uncertainty and the date of measurement. Two loads of the same variety from adjacent fields may differ, and the model can show by how much.

A result is only ever claimed for the material it represents. The model traces field to harvest lot to consignment to sample to result, including where lots are split or blended.

The terms are used precisely:

• Nutrient content — the measured amount of a named nutrient per unit mass, on a stated basis (for example, mg zinc per kg dry matter).
• Nutrient density — nutrient content relative to energy (for example, per 100 kcal), used when comparing foods of different energy content. Within a commodity of near-constant energy, such as wheat, higher content means higher density.
• Nutritional quality — a defined profile of nutrient contents, measured on a consignment and assessed against a stated reference for its commodity, such as a benchmark range or contract specification. Never a single unqualified score.

None of these is bioavailability (how much the body absorbs) or a health effect (what eating the food does). Each of those needs its own evidence.

Adjectives cannot be verified, traded, audited or improved. Values with a method can.

Commitment 02

Soil to health is one connected model, not six separate literatures

Farming practice, soil biology, crop and animal production, food composition, diet, human health. Most institutions treat these as six fields with six literatures, six conferences and six funding streams. Vitagri holds them in one model: distinct entities, linked by explicitly stated pathways, with the evidence for each link catalogued in the same governed structure in Pulse Brain.

Connecting the disciplines does not establish a causal chain. Its value is the opposite: it shows, link by link, where the evidence is strong, where it is weak and where it is missing. Strong evidence from soil to crop and strong evidence from diet to health do not add up to a demonstrated effect from soil to health. The model makes that gap visible rather than papering over it.

Commitment 03

Evidence is an entity, and its strength depends on the question

A citation is not evidence. In Vitagri's model a source carries its study design, graded as an evidence tier. A finding extracted from it carries what the source actually reported, with an effect size where one exists and a record of how it was extracted. A claim — made by Vitagri, a customer or anyone else — is a separate entity again, holding a verdict against the catalogue.

Three things are kept apart:

• Study design — what kind of study it was. A property of the source.
• Extraction confidence — whether the catalogue has recorded the finding correctly.
• Certainty — how far the body of evidence supports a specific claim about a specific outcome. This depends on risk of bias, consistency, directness, precision and publication bias across studies, not on any one study's tier.

A well-designed trial can give weak support for a claim it did not test. Representing that is what allows Vitagri to say how strongly it knows something, and where it does not know. An organisation whose product is verification must hold its own uncertainty as data, not as a caveat in prose.

Commitment 04

Properties belong to classes, and a marker is valid only where it has been validated

Conjugated linoleic acid and odd-chain fatty acids trace largely to microbial activity in the rumen. They speak to ruminant products. Reading them against a wheat sample is not a wrong answer that better data would correct; it is a category error. Other markers are studied across classes but mean different things in each: nitrogen isotope ratios are researched for both crops and livestock products, and each must be read against its own reference.

So every marker in the model carries the classes and conditions under which it has been validated. Outside them, the model returns no reading rather than a weak one.

Vitagri's marker discipline is often described as caution. It is a design property: the reason outputs inside their validated scope can be relied on.

Commitment 05

Farming-system labels are recorded, not used as explanations

Regenerative, organic, conventional, agroecological. These labels differ in kind. Organic is a legal standard with defined rules and certified inspection. Regenerative and agroecological have no settled definition in UK law, and different people use them for different bundles of practice.

Vitagri records a label as a fact about a farm or consignment: which label, under which standard or definition, certified by whom, valid when. It does not treat a label as the explanation for an outcome. Explanations are sought in the specific practices, conditions and measurements the label bundles together. Whether a certified system as a whole is associated with a different outcome is an empirical question, answered by the evidence like any other.

This is the structural reason Vitagri does not take a side in the argument about farming systems. Independence is not a value the company asserts and might trade away under commercial pressure. It is a property of how the data is held.

Corollary

The model is versioned, never silently rewritten

Every definition, field and grading rule carries a version. Additions are the normal case. When something must be corrected, deprecated or redefined, the change is made as a new version, with a documented reason and a migration, and the earlier definition is kept. Every historical record can still be read in the meaning it was created under.

A catalogue that can be silently rewritten backwards is not an evidence base; it is an opinion with a database behind it. The same principle governs the Growing Health Report: a dated synthesis, superseded by later editions rather than quietly edited. It governs this page too: every version stays available, and the changes are listed below.