The BARC Protocol — Open-Science Specification Series, Release B-2.1.0: Cover Note, Shared Conventions and Disclosure Boundary

Cover note, shared conventions and disclosure boundary

Author. Prasanna M, Growlrr Foods Pvt Ltd, Chennai, India. Correspondence: support@growlrr.com. ORCID: 0009-0003-6812-347X.

DOI. 10.5281/zenodo.22647606 — the concept DOI, which always resolves to the current version of this record · Published. 8 September 2026

Licence. Creative Commons Attribution–ShareAlike 4.0 International (CC BY-SA 4.0).

Status. A framework standard and validation record published by Growlrr Foods Pvt Ltd and shared openly for citation and scrutiny. Not peer-reviewed as of the date of publication, 8 September 2026, and not a consensus position of any external body.

Competing interests and funding. The author is the founder of Growlrr Foods Pvt Ltd, a pioneer and proponent of the BARC Protocol for pet foods. BARC stands for Biologically Appropriate, Residually Corrected, and is conceptualised as an open protocol under a Creative Commons Attribution–ShareAlike licence, for like-minded collaborators around the world to formulate and develop. Growlrr Foods Pvt Ltd develops and commercialises a BARC-based residual-correction product, BowlBalancer. The work was funded entirely by the company; no external funding was received. No independent laboratory, institution or reviewer has verified the results reported here. Readers should weigh these facts alongside the analysis, which is published in full — including its limitations and open items — precisely so that it can be checked.

Framework. The manuscripts in this series apply BARC — Biologically Appropriate, Residually Corrected — an open protocol released under a Creative Commons Attribution–ShareAlike licence (CC BY-SA 4.0) at growlrr.com/barc-protocol (concept DOI 10.5281/zenodo.22648597) for others to formulate and develop against. The framework is not deposited here; this series is one implementation of it.

Plain-language material. The case for correcting a home-prepared bowl at all — the problem §1 sets out — is put non-technically at growlrr.com/why-balance-homemade-pet-food; how the correction works is at growlrr.com/how-it-works; the reference list behind the public claims is at growlrr.com/pet-nutrition-references; the nutrition hub is at growlrr.com/pet-nutrition; and licensing and citation terms are at growlrr.com/license.

Abstract

This is the cover note to a ten-record open deposit specifying and validating a micronutrient residual correction for fresh and home-prepared cat and dog diets, together with the open protocol it implements.

Published evaluations of home-prepared pet diets find them nutritionally incomplete in the great majority of cases: of 200 canine recipes assessed against National Research Council allowances, five met every essential nutrient requirement; of 94 feline recipes, none did. The standards themselves are not the obstacle — the NRC, FEDIAF and AAFCO all publish quantitative daily requirements. The obstacle is that correcting a fresh bowl to those requirements demands microgram-scale accuracy across roughly two dozen nutrients, which a household cannot achieve with organ meats, botanicals and an array of separate supplements.

The eight manuscripts deposited alongside this note specify one response to that problem and report its validation: a low-phosphorus calcium architecture, an allometric energy model, the process engineering of a high-dilution unit dose, a stochastic nutrient-risk model run to 1,000,000 simulated feeding days, a tissue-kinetic treatment of chronic safety, a deterministic over-dose margin analysis, a two-sided calcium gate for large-breed growth, and the validation methodology including a paired-arm ablation. The separately deposited BARC Protocol states the framework they implement.

This note carries what the eight records share and should not each restate: the reading conventions for every reported figure, a glossary of the statistical and nutritional vocabulary used, the release lineage, the reproducibility position, the disclosure boundary and the reasons for it, and the limitations of the programme as a whole. All quantitative results in the series are model outputs. No animal feeding trial is reported, and as of the date of publication no independent laboratory, institution or reviewer has verified them.

Keywords: BARC Protocol · residual correction · homemade dog food · homemade cat food · home-prepared diet · companion animal nutrition · nutritional adequacy · fresh feeding · raw feeding · open science · specification series · NRC 2006 · FEDIAF · AAFCO

1. The problem this work addresses

Home-cooked feeding has grown worldwide as an alternative to industrially produced kibble and wet food, for reasons that are not unreasonable: owners want fresh, identifiable ingredients and control over what their animal eats. The difficulty is that the resulting bowls are, in the overwhelming majority of published evaluations, not nutritionally complete.

The evidence is unusually consistent. Stockman and colleagues evaluated 200 home-prepared maintenance recipes for dogs, 129 of them written by veterinarians: five provided adequate concentrations of all essential nutrients against National Research Council recommended allowances, and nine met the AAFCO profiles [1]. The corresponding feline study analysed 94 recipes and found that none met the NRC recommended allowances for all essential nutrients; most were short on three or more, some on as many as nineteen, with several falling below half the recommended allowance for choline, iron, zinc, thiamine, vitamin E and manganese [2]. Stockman et al. also tested the most common defence of home feeding — that rotating through varied recipes averages out to completeness — and found that it did not [1].

The standards themselves are not the gap. The NRC, FEDIAF and AAFCO publish quantitative daily requirements as a function of body weight, life stage and activity [3,4,5]. The gap is one of access and precision. The nutrients most often missing are the ones least forgiving of approximation: the calcium-to-phosphorus relationship, iodine, taurine, thiamine, and vitamins A and D. For the household, correcting them accurately is largely out of reach. In practice owners reach for organ meats, kelp and other botanicals, advice passed between owners, and discussion forums, while attempting to juggle a dozen or more separate supplements at doses no domestic scale can resolve. That is burdensome for an owner who is trying to do better than an industrial feed they have lost confidence in, and it carries real risk for the animal over years of feeding.

Why the ancestral answer does not close it. Raw and ancestral-pattern feeding — the BARF tradition — meets these requirements through the prey animal itself: bone for calcium, liver and kidney for vitamin A and copper, tripe, kelp for iodine, fed as a daily rotation of organs. That works in the setting it was derived from. It transfers poorly to an urban pet, because the mechanism it depends on is energy throughput, not the recipe. A free-ranging carnivore eats a substantial multiple of a sedentary pet's maintenance energy — on receipted estimates roughly 1.6 to 2.3 times, and higher in working and free-ranging populations (BARC Protocol record, 10.5281/zenodo.22648597, section 2.3) — and because prey composition is fixed per kilogram, eating that multiple delivers a corresponding multiple of every micronutrient. A neutered, indoor, ageing animal active for a few hours a day, and predisposed to weight gain, eats a small fraction of that. The same ancestral bowl, served against a small calorie budget, no longer carries the micronutrient load; the low-density nutrients fall through first.

The consequence is that for a modern pet the diet must be more micronutrient-dense per kilocalorie than an ancestral diet, not merely more natural. That is the requirement this work sets out to satisfy, and it is what BARC is for.

A caveat that belongs at the front, not in a footnote. The correction proposed here is performed with purified, pharmaceutical-grade actives rather than with organs and botanicals — for the same reason the problem exists in the first place: kelp, liver and similar sources vary by species, season, site and lot, and that variation is not declared per lot, so an owner correcting with them cannot know the dose they gave. But precision cuts both ways. Cholecalciferol, selenium compounds and iodine sources are dosed in micrograms and have a narrow interval between the daily requirement and a harmful one. They are not kitchen-counter materials. The position taken throughout this series is that the accuracy must live in a high-precision blending facility and reach the household as a fixed unit dose — a count rather than a measurement. §5 states what follows from that for what is and is not published here.

1.1 BARC, and what this series is

BARC — Biologically Appropriate, Residually Corrected — is an open framework for feeding cats and dogs. It requires two things at once: a fresh-food base that is biologically appropriate for a carnivore, and a calculated residual correction that closes the measured gap between that base and a published nutrient standard, dosed to the individual animal. Neither alone qualifies.

The framework is deposited separately, and cited separately. BARC has its own record — concept DOI 10.5281/zenodo.22648597 — under a Creative Commons Attribution–ShareAlike licence (CC BY-SA 4.0), so that anyone may formulate and develop against it. It is deliberately a separate record rather than part of this series: the framework is not owned by any implementation of it, including this one, and citing BARC should not mean citing this product. A living, human-readable copy is kept at growlrr.com/barc-protocol; the concept DOI always resolves to the current version, so the framework can be revised as others contribute without breaking citations already made to it.

This series is one such implementation, described technically. Eight manuscripts specify and validate a particular residual correction — its calcium architecture, its energy model, its manufacturing method, its chronic-safety and over-dose margins, and the methodology used to test it. They are the engineering, not the framework.

The distinction matters for reading them. Where a manuscript states a BARC boundary or definition it is applying the open framework, not asserting a new one; where it states a delivered milligram, a scaling exponent or a reliability figure, that is this implementation and is the thing being put up for checking.

2. The eight manuscripts

Each is standalone and separately citable. Read in this order for a first pass; each may be read alone.

Where to start, by interest. For the statistics: 2 and 8. For the formulation engineering: 1, 4 and 7. For the safety case: 5 (time axis) and 6 (dose axis). For the energy model and dosing: 3.

How the manuscripts depend on one another

Each record declares these as cites relations, isPartOf this cover note, and isSupplementTo the BARC Protocol page, so the set resolves as a graph rather than as eight loose files.

3. Release lineage

Two identifiers recur and mean different things. 3077c230 is the engine build of release B-2.1.0. 842353e5 identifies the formulation of the adult corrections, fixed 3 August 2026 and unchanged through this release. Both are internal build identifiers, quoted only so that a specific number can be traced to the exact code and formulation that produced it. A result computed on the 842353e5 block with the bone-mineral calcium system belongs to release A-1.0.0 and is labelled as such wherever it appears — most consequentially in Manuscript 6 §3.

4. Reading the reported figures, and the terms used

1. Every figure carries its species, release, draw count and confidence basis. Results from different releases and draw counts are not repeated measurements of one quantity and are not comparable as such.

2. A bound obtained with zero observed failures is N-limited. It reduces to n/(n + z²), a function of draw count alone, and the series therefore reports such results as zero failures observed in N draws, Wilson-95 lower bound X, N-limited. A bound of this kind is not a reliability rate, and a larger draw count does not indicate a more reliable formulation.

3. The physical quantity converges long before the bound does. Worst-nutrient delivery is stable from roughly 5,000 draws. The value of the 1,000,000-draw run is its replicate dispersion and its audit corpus, not extra decimal places.

4. One confidence convention runs through the series. Every Wilson bound is the lower limit at z = 1.96, the two-sided 95% interval. Where a C99.99 basis (z = 3.891) is quoted, it is named explicitly and applies only to the inherited 100,000-draw baseline.

5. The gate is C/R 95/99 — the Wilson-95 lower bound must be ≥ 99% of days at ≥ 90% of requirement, and a batch fails if any nutrient reaches ≥ 80% of its safe upper limit. A day passes only if every floor is cleared: a joint criterion, not an average.

4.1 Glossary

The series assumes some familiarity with animal-nutrition and statistical vocabulary. This is what each term means as used here; where a term has a broader meaning elsewhere, the narrower sense below is the one intended.

5. Disclosure boundary

Disclosed across the series. The BARC definition and taxonomy; the appropriateness boundaries; the allometric energy model in full, with coefficients, exponents, multiplier bands, anchors and the delivery buffer; the unit-count and floor-scaling equations and the exponent-matching rule; the structural principles of the calcium, growth-calcium and homogeneity architectures with their outcomes as percentages and ratios; the validation methods in enough detail to reproduce the procedure; the complete per-arm ablation result set; the pooled and per-run results of the 1,000,000-draw validation and its 200,000-row audit corpus; the family-configuration results; the reservoir, kinetic and deterministic-escalation results; and verified reference lists.

Withheld, and the first reason is safety rather than commerce. The per-unit composition by weight is not published. Residual correction in this framework is performed with pharmaceutical-grade actives — calcium iodate, a selenium compound, retinol and cholecalciferol, copper and manganese salts — several of which are dosed in micrograms and have a narrow interval between the requirement and the toxic dose. Vitamin D, selenium and iodine are the clearest cases: a weighing error of a few milligrams, invisible on any domestic scale, is the difference between a correct dose and a harmful one. These are not kitchen-counter materials, and publishing a percent-by-weight recipe would invite exactly the practice this framework exists to replace — an owner triturating a microgram-scale active by eye at home.

This is the same reasoning that leads the framework to prefer purified actives over botanicals and organ tissue as correction sources (Manuscript 4 §4). A kelp or liver route is not safer for being natural; it is less controlled, because its active content varies by species, season, site and lot and is rarely declared per lot. The framework's position is that if a nutrient is worth correcting deliberately, it should be corrected with a material of known assay, blended under high-precision conditions in a facility built for it, and presented to the household as a fixed unit dose — so that the precision lives in the factory and what reaches the kitchen counter is a count, not a measurement. That is a design commitment, not only a disclosure policy.

Withheld — also proprietary, not part of this deposit and not licensed. The correction's bill of materials: ingredient identities beyond those already published on growlrr.com, the salt and chelate form of each active, carrier fractions, dilution quantities and blend ratios, the process parameters that realise them, and supplier and contract-manufacturer specifications. And the source of the balancing engine.

What is derivable from what is disclosed. The series reports per-nutrient delivery as a percentage of requirement, at stated unit counts and a stated unit mass, because no adequacy or safety claim here can be checked without it. Elemental mass per unit therefore follows from published requirement tables, and that is intended: the nutritional claims must be checkable. What it yields is a nutrient specification — what reaches the animal — and not a formulation: it does not give the materials, salt forms, carriers, dilution scheme or process that deliver those masses. It is emphatically not a preparation method, and nothing here should be read as instructions for compounding these actives outside a controlled facility. Any reproduction of the methods represents the correction abstractly, as a mass that closes the measured gap. Nothing in this series should be read as a formula.

Two limits on what the licence does. CC BY-SA 4.0 covers this documentation and its data; it does not license the formulation as an invention, which is separate intellectual property and is not disclosed. And it acts forward only: material that was public and unlicensed before this deposit is not retrospectively bound by it.

6. Reproducibility

What a reader can do. Re-derive every energy target, unit count and nutrient floor from the equations in Manuscript 3. Re-check every reported aggregate against the per-run figures tabulated in Manuscript 2s, including the 200,000-row audit corpus, which is itself reproducibly sampled: rows are drawn without replacement per run by a deterministic sampler independent of the engine's own streams, then merged across all ten seeds and shuffled. Check every reported figure against its stated release, draw count and basis. Check every reference against the resolvers.

What a reader cannot do. Regenerate new draws. That requires the balancing engine, which is not deposited. The run design is fully specified — ten independent runs of 100,000 draws, seeds spaced 1,000 apart, pooled on exact pass and total counts — so the procedure can be reimplemented, but the correction vector it operates on is proprietary and any reimplementation must represent it abstractly. The 1,000,000 draws are accumulated from ten seeded runs rather than executed as a single process, which is a memory constraint of the implementation rather than a property of the method.

6.1 Computational tooling and use of AI assistance

The validation model is implemented as TypeScript modules executed under Node.js, with no external numerical or statistical dependencies: the pseudo-random generator (mulberry32, seeded and deterministic), the Box–Muller lognormal sampler, the Wilson interval and the aggregation are all implemented in the codebase rather than delegated to a library. Manuscript 2 §3.2 gives the detail.

Claude, a large language model developed by Anthropic, was used as a coding and analysis assistant throughout the development of this work and the preparation of these records — in writing and refactoring implementation code, in auditing the manuscripts for internal consistency and unsupported claims, and in checking reported figures against the engine's output.

It was not used to generate results. Every quantitative value in this series was produced by executing the model. The author is responsible for all content.

7. Scope of the whole series

• Healthy animals only. Adult and growing cats and dogs at maintenance, growth or reproduction. Not a clinical or therapeutic system; no disease claims; not a substitute for a veterinary therapeutic diet. Animals with diagnosed disease should be fed under veterinary direction.
• Protection is not treatment. The low-phosphorus architecture enables a lower, diet-governed phosphorus load, which is preventive nutrition for a healthy ageing animal. It does not treat kidney disease.
• Models, not guarantees. Every reliability figure is a confidence bound at a stated draw count under a stated model. The word "guarantee" is not used.
• No claim that fresh feeding beats commercial feeding. The argument is scoped to completeness, never to freshness. There is no consensus that switching to a raw or fresh diet improves outcomes, and this series does not assert one.
• No brand comparison. Category arithmetic computed from public composition data appears in Manuscript 1 and is deliberately brand-agnostic. No brand-level audit is published, and none should be inferred.
• Wild carnivores are not a welfare benchmark. The throughput argument uses wild energy expenditure to show why an ancestral recipe under-delivers at domestic intake; it does not claim wild diets are optimal, and it quotes the smaller, receipted multiplier rather than the larger, convenient one.
• Not peer-reviewed as of the date of publication (8 September 2026), and the framework does not claim to have invented the category. Manuscript 1 §3 names its predecessors.

8. Limitations and future work

These apply across the series and are stated once here rather than repeated in each manuscript, which carries its own method-specific limitations.

Empirical validation is outstanding. No feeding trial has been conducted, no serum or tissue concentration has been measured, and no finished production lot has been assayed against the design specification. Every figure in this series is the output of a model whose inputs are published food-composition tables and requirement standards. The models establish internal consistency and quantify behaviour under assumed variability; they do not, and cannot, substitute for measurement in animals.

A non-invasive household observation study is planned to begin closing this gap. The intent is to collect, from animals fed the corrected diet under ordinary home conditions, the class of measures an owner can record without clinical intervention — the kind of longitudinal observational data that can corroborate or contradict a modelled expectation in a real population rather than a simulated one. Its protocol, endpoints, sample size and analysis plan are not yet fixed, and no results exist. Results will be reported when they do, and this series will be revised against them. Until then the work is a model and is presented as one.

Composition inputs are tabulated, not assayed. Ingredient nutrient contents are taken from published composition databases rather than from analysis of the lots actually used, and the coefficients of variation in the noise model are assumed rather than estimated from observed lot-to-lot data. A systematic error in the underlying tables would propagate through the entire analysis undetected.

The validation is partly self-referential. The engine computes a residual gap and sizes a correction to close it, then evaluates whether the corrected diet meets that requirement. The paired-arm ablation [Manuscript 8] mitigates this by demonstrating that the uncorrected base fails, which is not a foregone conclusion, but both arms draw on the same composition inputs.

The energy exponent rests on a single study. The canine exponent of 0.64 derives from one calorimetry dataset of 22 dogs [Manuscript 3]. A larger pooled meta-analysis reports a materially different exponent, and the fixed-composition architecture depends on the lower value. Manuscript 3 §8 states the consequence; a sensitivity analysis across the plausible exponent range would strengthen the position.

Toxicokinetic parameters are cross-species. The load-bearing absorption parameter in the kinetic model is a human tracer value applied to dogs and cats, as are several whole-body half-lives [Manuscript 5].

Scope boundaries. Healthy animals only; adult and growth life stages; the canine energy model is empirically grounded over 5.8–48.8 kg and extrapolated outside it; safe upper limits are drawn from secondary summaries of the primary requirement volume; and the deterministic escalation is computed on an earlier formulation release, conservative for the nutrient it affects.

Independence. This work is authored, funded and published by the company that develops the product it describes. It has not been peer-reviewed as of the date of publication (8 September 2026), and no independent laboratory or institution has verified the results.

9. The records in this deposit

Fill each slot once at deposit and every cross-reference in the series resolves.

10. References

1. Stockman J, Fascetti AJ, Kass PH, Larsen JA. Evaluation of recipes of home-prepared maintenance diets for dogs. J Am Vet Med Assoc. 2013;242(11):1500–1505. doi:10.2460/javma.242.11.1500
2. Wilson SA, Villaverde C, Fascetti AJ, Larsen JA. Evaluation of the nutritional adequacy of recipes for home-prepared maintenance diets for cats. J Am Vet Med Assoc. 2019;254(10):1172–1179. doi:10.2460/javma.254.10.1172
3. National Research Council (US), Committee on Animal Nutrition. Nutrient Requirements of Dogs and Cats. Washington (DC): The National Academies Press; 2006. doi:10.17226/10668
4. FEDIAF. Nutritional Guidelines for Complete and Complementary Pet Food for Cats and Dogs. Brussels: FEDIAF; 2024. https://europeanpetfood.org/self-regulation/nutritional-guidelines/ (This is the edition on which the nutrient tables used here are based.)
5. Association of American Feed Control Officials. Dog and cat food nutrient profiles. In: AAFCO Official Publication. Champaign (IL): AAFCO; 2016.

11. How to cite

Cite the individual manuscript for a specific claim, and this cover note's concept DOI where you mean this body of technical work as a whole. To cite BARC itself — the framework rather than this implementation — cite growlrr.com/barc-protocol.

Prasanna M. (2026). The BARC Protocol — Open-Science Specification Series, Release B-2.1.0: Cover Note, Shared Conventions and Disclosure Boundary. Growlrr Foods Pvt Ltd. Zenodo. https://doi.org/10.5281/zenodo.22647606

12. Contact and contributions

Correspondence: support@growlrr.com. The framework is published to be checked. Datasets, test batches, corrections and disagreements are welcome, and the public repository linked from these records is where to send them.