Every section below appears on every product. This explains what each one is, and — more importantly — what it does not mean.
This bag is invented, but every claim on it is real — each one appears on products sold today. Here is what this app can and cannot do with each.
The app reads the parts of a label that can be checked, says plainly where it is estimating, and turns the rest into questions worth asking the manufacturer.
What it claims is out of 100. It measures how much a label tells you about the food and how precisely it says it — where the protein comes from, how the fat and carbohydrate are structured, whether ingredients are named down to the species and tissue.
What it proves is out of 7. It counts specific measurements a manufacturer has published: an amino acid profile, a digestibility study, a typical analysis, a feeding trial.
What it does mean: a food scoring 85 and 0 has written a detailed, precise label and published no measurements. A food scoring 55 and 4 says less but has proven more of it.
What it does not mean: that 85 is better than 55. These are not grades. A carefully made food sold with a sparse label scores lower than an ordinary food sold with a detailed one, and that is the measurement working, not failing.
The guaranteed analysis, exactly as printed, with the label's own ≥ and ≤ kept.
What it does not mean: that these are the real values. Protein and fat are legal minimums, fibre and moisture legal maximums.
Arithmetic on the declared values — dry-matter conversions, grams per 1,000 kcal, carbohydrate.
What it does not mean: precision. Because moisture is a maximum, every dry-matter figure is a range rather than a point. And carbohydrate is not starch — it is what remains after everything else is subtracted, which includes sugars, soluble fibre and measurement error.
Where a manufacturer has published a measurement, that number is shown. Where they have not, a range is estimated from the processing format using published feeding trials.
What it does not mean: that an estimated range describes your bag. It describes foods of that type. And "90% digestible" on a package is ambiguous — protein, fat, energy and dry-matter digestibility can differ by eight points in the same food.
Total essential amino acids per 100 g of dry matter — measured where published, estimated otherwise.
What it does not mean: protein quality in the full sense. This is density. It says nothing about which amino acid is in shortest supply, and nothing about whether heat processing left them available to the dog.
Facts about how the food is built: the fat source, whether there is a marine oil, how many organ types, whether artificial colours or synthetic preservatives are present. No score.
What it does not mean: that any of it is good or bad. Whether a marine oil matters depends on your dog.
Everything a label cannot establish, each with a question written out for you to put to the manufacturer.
What it does mean: these are answerable. Most are one number the maker already has.
What it does not mean: that the food is poor. Nothing in this section reduces any score. A missing study is an open question, not evidence.
This page is separate from the product screens on purpose. A shopper comparing two bags does not need it. Anyone deciding whether to trust the numbers does.
It is not a nutritional quality score, a safety score, or a recommendation. It measures what a label declares and how precisely — nothing more. A carefully formulated food sold with a sparse label will score lower than an ordinary food sold with a detailed one, and that is the score working as intended rather than failing. It cannot tell you how digestible a food is, whether its amino acids survived processing, or how your dog will do on it. Those live in evidence confidence and in the questions, and mostly they are not knowable from a label at all.
The two numbers are shown side by side and never combined for this reason. A high architecture score with an evidence score of zero describes a food that looks good on paper and has proven none of it.
Label architecture describes what the label declares the food to be. It is not a health claim, and a high score proves nothing about how a dog will do on the food. Evidence confidence counts how many of seven specific disclosures the manufacturer has published. They are kept apart because a food can be architecturally excellent and evidentially silent — combining them would hide exactly the thing worth seeing.
Every product page carries a section called formulation choices. It reports whether the food contains a marine oil, which fat source it uses, how many organ types are in it, and whether artificial colours or synthetic preservatives are present. None of it carries a score.
That is deliberate. Whether those things matter depends on the dog in front of you. A marine oil is valuable if your dog needs the omega-3 and irrelevant if they are already getting it elsewhere. Organ variety matters more in a diet with fewer other sources of micronutrients than in one that is heavily supplemented. Artificial colouring does nothing for any dog, but it is also not the reason a food would or would not suit yours.
So the app tells you what the label says and stops there. It scores what can be read off a label — how the protein is sourced, how the macronutrients are structured, how precisely the ingredients are named — because those are facts about the label. What those facts are worth is not a question a label can answer, and not one an app should answer on your behalf.
An earlier version did put points on these. It was wrong to, and the points are gone.
We do not rank ingredients by presumed quality, because the canine literature does not support it. Fresh poultry out-digested poultry meal in one trial while fresh beef and beef meal came out level; the advantage disappeared entirely after extrusion in another; and soy protein concentrate out-digested poultry meal for most indispensable amino acids in ileally cannulated dogs. What a label genuinely tells you is specificity — whether an ingredient names a species and a tissue — so that is what is scored.
Protein and fat are minimums; fibre and moisture are maximums. Where products in this catalog have been measured in published trials, protein came in at ×1.165, ×1.124 and ×0.934 of the guarantee. One did not reach its own label. Because moisture is a maximum, every dry-matter figure carries a range.
Carbohydrate is calculated by difference and includes sugars, soluble fibre, organic acids and
analytical error. In one trial a frozen raw diet calculated to 8.59% carbohydrate and measured 0.00%
starch, while an extruded diet calculated to 45.19% and measured 45.64%. Carbohydrate is reported as a range and
starch is treated as unknown unless a manufacturer publishes it.
Nothing in the unknown section reduces any score. A brand that has not published a digestibility study has not proven anything bad — it has left a question open, and the app writes that question for you to ask.
These are questions a reasonable person would want answered before buying a bag of food. None of them can be answered from a label, and this app does not pretend otherwise. Each says what would answer it — which is usually one number the manufacturer already has.