Our Methodology: How PetFoodRankings Scores Dog & Cat Food
Most pet food rankings hand you a score with no explanation. PetFoodRankings does the opposite. Every number on this site comes from a published rubric, and this page is the rubric.
Our scores use a hybrid model. Three dimensions — protein quality, ingredient transparency and value — are computed by algorithm from structured label data, so they are reproducible and consistent across every product. Three dimensions — nutritional standards, recall history and owner feedback — require human judgment and are editorially assessed. Editors can apply bounded adjustments to algorithm scores, and every adjustment is logged on the product page.
Independence & funding
PetFoodRankings is independently owned. No manufacturer, retailer or PR agency can pay for placement, a higher score, or removal from a ranking. We do not accept sponsored content disguised as reviews.
We earn revenue through affiliate commissions when readers buy through links on this site. Commissions are identical regardless of score, and no brand is told in advance whether it will be included. If a conflict ever arises, the correction is published here rather than quietly made.
The 100-point scale
Every product receives an overall score from 0 to 100, calculated from six weighted dimensions. Scores are rounded to the nearest whole point; ranking order uses the unrounded value.
The scale is deliberately uncompressed. A food in the 80s is genuinely good; we reserve the 90s for products with no significant compromise we could identify.
Algorithm vs. editorial — who scores what
Two kinds of information go into a score. Facts that can be computed the same way for every product (ingredient tiers, dry-matter protein, price per calorie) belong to an algorithm. Judgments that require context (how serious a recall was, whether a formula fits a life stage) belong to editors. Mixing the two is how review sites end up with unexplainable numbers.
Algorithm dimensions: protein quality (30%), ingredient transparency (20%), value for money (10%). These are computed from the first five ingredients, the guaranteed analysis and the price — the same inputs produce the same output every time.
Editorial dimensions: nutritional standards (18%), recall history (14%), owner feedback (8%). These are assessed against the rubrics below and reviewed by a second editor.
Editors may adjust an algorithm score by a bounded amount when the algorithm misses something real (for example, a documented batch-consistency problem). Every adjustment is shown on the product page as an explicit editorial-adjustment line and recorded in the score history. An adjustment without a logged reason should not exist — if you spot one, tell us.
The six scoring dimensions
Named meat first, species-appropriate protein levels
Named ingredients, published guaranteed analysis, no hidden blends
AAFCO adequacy statement, feeding trials, WSAVA alignment
Recall record over the past 10 years and company response
Fair price for the quality delivered, per 1,000 kcal
Aggregate owner ratings and reported palatability issues
Protein quality & source — 30 points (algorithm)
Dogs are facultative carnivores and cats are obligate carnivores, so the source of protein matters more than any marketing claim. The algorithm reads the first five ingredients and the guaranteed analysis, then computes a score from four inputs.
Every product starts at 50 before ingredient and nutrient adjustments.
First ingredient tier: named whole meat (deboned chicken, fresh salmon) +25; named meat meal +16; named organ meat +10; whole grain or named fat +6; neutral 0; plant protein isolate −10; named by-product −14; unspecified species −18.
Each additional named animal ingredient in the first five earns +4 (whole meat), +2 (meal) or +1.5 (organ). Each plant protein concentrate costs −6, each by-product −8, each unspecified ingredient −10.
Protein is converted to a dry-matter basis (see below) and compared against species and life-stage targets: puppy 28%, adult dog 24%, senior dog 24%, kitten 30%, adult cat 28%, senior cat 28%. The gap from target adds or subtracts up to ±8 points.
Why dry-matter basis: wet food at 78% moisture shows 10% protein on the label, but roughly 45% once water is removed. Comparing as-fed percentages across forms is mathematically meaningless, so every protein comparison on this site happens on a dry-matter basis.
Ingredient transparency — 20 points (algorithm)
A label you cannot evaluate is a label you cannot trust. The algorithm scores what the first five ingredients reveal about how much the brand is telling you.
Every product starts at 70.
+5 if the first ingredient is a named animal source (whole meat, meal or organ); −10 if it is not.
−8 for each by-product or unspecified-species ingredient in the first five (poultry by-product meal, meat by-products, meat and bone meal, animal fat). −5 for each plant protein concentrate (pea protein, corn gluten meal, wheat gluten, soybean meal).
+5 if the first five ingredients contain no penalty-tier ingredients at all.
Current limitation, stated plainly: we score the first five ingredients because that is what our structured data covers today. Full ingredient-panel parsing (proprietary blends, additive flags, guaranteed-analysis completeness) is on the roadmap and will raise the ceiling of this dimension.
Nutritional standards — 18 points (editorial)
We verify the AAFCO nutritional adequacy statement on every label — the statement that tells you how the food was validated, not just what it claims to contain.
AAFCO feeding-trial validated for the stated life stage (the food was actually fed to animals and shown to sustain them), and the formula aligns with WSAVA guidance: named protein source, moderate-to-low carbohydrates, appropriate caloric distribution.
Formulated to meet AAFCO nutrient profiles rather than feeding-trial tested. Meets the letter of the standard, but was never fed to a live animal.
Life stage mismatch (e.g., an “all life stages” formula that is too energy-dense for large-breed puppies), missing adequacy statement, or a formula built around a trend rather than species needs.
AAFCO sets a floor, not a ceiling. Meeting the profile is the minimum for inclusion on PetFoodRankings; exceeding it thoughtfully is what earns points here.
Recall history & safety — 14 points (editorial)
A single recall does not disqualify a brand — what matters is frequency, severity, and how the company responded. We review the past 10 years of recalls (mandatory and voluntary) and public statements.
No recalls in the past 10 years, or only minor labeling recalls handled transparently with prompt customer notification.
One older, contained recall (e.g., a single lot) with a clear, fast public response.
Multiple recalls, or any recall involving pathogen contamination (Salmonella, Listeria) or nutrient excess/deficiency that reached consumers.
Recurring recalls across multiple years, or a pattern of slow/incomplete customer notification.
We report recalls as facts with dates and scope — never as speculation. If a product is currently under recall or investigation, its page carries a warning banner and it is excluded from rankings until resolved.
Value for money — 10 points (algorithm)
Cheap food is not automatically good value, and expensive food is not automatically bad value. What matters is the price of the calories relative to comparable products — so the algorithm computes it.
Price per 1,000 kcal = price per pound × 2.20462 ÷ kcal per kg × 1,000. Calorie content comes from the label where we have it, otherwise from a form-based estimate (dry 3,550 / wet 1,000 kcal/kg) — the estimate is disclosed on every product page.
Within each species (dog food or cat food, wet and dry together) the cheapest product per 1,000 kcal scores 100 and the most expensive scores 15, with everything in between scaled linearly. We deliberately do not split the group by form: price per 1,000 kcal already absorbs the calorie-density difference between wet and dry, so a canned food that costs six times more per calorie than a kibble should not score 100 merely for being the cheapest can in a small group. Value is explicitly relative, never absolute.
Editors can adjust when the algorithm misses context (e.g., a budget food that is cheap because it is nutritionally thin). Adjustments are logged.
Owner & pal feedback — 8 points (editorial)
We aggregate owner ratings from major retailers and veterinary forums, with extra attention to palatability reports (a nutritious food a dog refuses to eat helps no one).
Consistently high owner ratings with few recurring palatability complaints.
Good average ratings with a noticeable minority reporting picky eating or digestive transition issues.
Mixed ratings, or recurring reports of refusal, stool changes, or formula inconsistency across bags.
This is our lowest-weighted dimension on purpose: anecdote is not evidence, and it should never outweigh what is actually in the bag.
The ingredient dictionary
The protein and transparency algorithms read a published ingredient dictionary that assigns every ingredient a tier. The dictionary is part of the methodology — if you disagree with a tier, you are disagreeing with a documented rule, not a vibe.
| Tier | Class | Examples |
|---|---|---|
| Tier 3 | Named whole meat | deboned chicken, fresh salmon, bison, whole herring |
| Tier 2 | Named meat meal | chicken meal, salmon meal, herring meal |
| Tier 1.5 | Named organ meat | chicken liver, beef liver |
| Tier 1 | Egg / named fat / whole grain | egg product, chicken fat, brown rice, barley, oatmeal |
| Tier 0.5 | Neutral | peas, sweet potato, beet pulp, flaxseed, canola oil, broth |
| Tier -1 | Plant protein concentrate | pea protein, corn gluten meal, wheat gluten, soybean meal |
| Tier -1.5 | Named by-product | poultry by-product meal, chicken by-product meal, meat by-products |
| Tier -2 | Unspecified species | meat and bone meal, animal fat |
Ingredients not yet in the dictionary are flagged at build time and scored as neutral until they are classified — we would rather under-score an unknown ingredient than guess.
How we gather data
Every product page is built from: (1) the manufacturer's current label and guaranteed analysis, (2) the AAFCO adequacy statement, (3) FDA and publicly reported recall records for the brand and its co-manufacturers, (4) aggregated owner ratings from major retailers, and (5) veterinary-nutrition references including WSAVA guidance.
We do not operate a laboratory and we do not run feeding trials ourselves. Where a claim cannot be verified from public records, we say so on the product page rather than guessing.
Update cadence, corrections & score history
Rankings are re-verified at least every six months, and immediately after any major recall, reformulation, or label change. Each page shows its last-updated date.
Algorithm dimensions recompute automatically whenever ingredient or price data changes — no silent edits. Every product page carries a score history log showing what changed, when, and why. Editorial adjustments appear in the same log.
If we get something wrong, we correct it, update the date, and note material corrections in plain language. Readers can flag suspected errors at tungchi [at] yeah.net — correction requests are reviewed by the editorial team, not by marketing.
What this methodology cannot do
No scoring system captures your individual animal. Breed, size, age, activity level, allergies and health status all change what “best” means. Use our rankings to shortlist, then confirm with your veterinarian — especially for puppies, kittens, seniors, or pets with diagnosed conditions.
We score the first five ingredients, not the full panel, and we do not lab-test. Both limitations are stated where they matter. PetFoodRankings provides general information, not veterinary advice.
Try your own weights
Every score on this site is a weighted sum you can recompute yourself. Drag the sliders to set your own weights — the rankings below update live using the exact sub-scores published on each product page. That is the point of a public methodology: you do not have to trust our weighting, you can test it.
Top dog foods under your weights
Top cat foods under your weights
Computation runs locally in your browser using the exact sub-scores published on each product page; nothing is uploaded.
Methodology FAQ
Which scores are calculated by algorithm and which are editorial?
Protein quality (30%), ingredient transparency (20%) and value for money (10%) are algorithm-computed from structured label data — same inputs, same output, for every product. Nutritional standards (18%), recall history (14%) and owner feedback (8%) are editorially assessed against published rubrics. Editors can make bounded adjustments to algorithm scores, and every adjustment is logged on the product page.
Does PetFoodRankings accept payment from pet food brands?
No. Rankings and scores are never for sale. We earn commissions when readers buy through affiliate links, but commission rates are identical across products and no brand can pay for placement or a higher score.
Why is there no single 'best pet food' winner?
Because context wins. The best food for a Great Dane puppy, a senior indoor cat and a dog with IBD are three different products. Our category pages exist precisely so each situation gets its own ranking.
Do you test the food yourselves?
We analyze labels, guaranteed analysis, AAFCO statements, recall records and aggregated owner feedback. We do not operate a lab or run feeding trials — we say so plainly rather than implying laboratory testing we did not perform.
How do you handle products under recall?
Any product currently under recall or regulatory investigation carries a warning banner and is removed from rankings until the issue is resolved, at which point we re-evaluate it from scratch.
Can I suggest a product to review?
Yes — email tungchi [at] yeah.net with the brand and specific product. We prioritize reader requests alongside our own research calendar.