A score you're allowed to argue with.

What this is

insPAWction reads pet product labels and produces a Product Score out of 100. It covers two families: everyday care products — shampoos, wipes, ear cleaners, toothpaste, balms, litter, household cleaners — and food and treats.

The name matters. This is not a health score. We deliberately avoided that word, because a health score implies a claim about an animal's wellbeing, and we are not in a position to make one. What we can measure is a label: what's in the bottle, whether the maker told you, and what published references say about those ingredients. That's a product score.

Currently: 22 products, 121 ingredients, and 63 blocked search terms that trigger the medication wall.

How the score is calculated

There are two engines, not one. A shampoo and a bag of kibble are not answering the same question, and pretending otherwise is how a scoring model quietly becomes nonsense. "Potential irritants" measures nothing when nothing touches skin. "Fragrance load" reads Natural Flavor as perfume. So food gets its own categories.

Both engines work the same way: six categories, each scored 0–100, combined as a weighted average. No hidden adjustments, no editorial thumb on the scale.

Care products — shampoos, wipes, litter, cleaners

Food and treats

Different questions, different categories.

Ingredient splitting

Worth calling out, because it's the one trick here a label reader can catch that a shopper usually can't. Ingredients are listed by weight. If a maker splits one base ingredient into fractions — peas, pea protein, pea starch, pea fiber — each fraction weighs less on its own, so each sits lower in the list. Combined, they might outweigh the meat sitting at the top. The engine groups the fractions and says so out loud.

Why those weights?

Both engines put 55% on the two categories that actually describe the contents — what's in it, and what's in it that shouldn't be. Transparency carries 15% in both, because a maker who won't tell you what's in the package has made a choice worth penalising. The remaining three carry 10% each: real signals, but not ones that should swamp the rest.

These weights are a judgment call. They are defensible, not objective. If you'd weight them differently, you'd get a different score, and that's a legitimate disagreement rather than a bug. They live in one constant per engine — SCORE_WEIGHTS in includes/scoring.php, FOOD_WEIGHTS in includes/scoring_food.php — precisely so they can be argued about and changed.

The colour bands

• 90–100 — Excellent. Clean, fully disclosed, well documented.
• 70–89 — Good. Solid, with something specific holding it back.
• 50–69 — Mixed. Read the receipt before deciding.
• 30–49 — Poor. Significant gaps in the label, the formula, or both.
• 0–29 — Very poor. Tells you almost nothing, or contains something that needs real care.

What we don't evaluate

This is the most important section on the page, and it's a hard line in the code, not a policy note.

insPAWction does not evaluate prescription medications, over-the-counter medications, supplements, CBD, melatonin, vitamins, medical devices, or anything intended to diagnose, treat, cure, or prevent disease. Searching for one returns a wall, not a score.

The reason is simple. Scoring a shampoo on transparency is a consumer question. Scoring a medication implies a clinical judgment about whether it should be given to an animal, and that judgment belongs to a veterinarian who has met your pet. No amount of careful wording gets around that.

We also don't cover beds, bowls, harnesses, collars, leashes, clothing, or toys. There's no ingredient list to read.

Note the line this draws. A bag of kibble that lists glucosamine among its ingredients gets scored — it's a food. A glucosamine joint chew sold to support a condition does not — it's a supplement. The wall is about what a product is sold to do, not which molecules appear in it. That's why searching glucosamine here returns an ingredient page, and glucosamine supplement returns the wall.

What a score does not mean

A high score does not mean a product is right for your pet. An individual animal can react badly to an ingredient that scores well here. Scores describe formulas in general; your pet is not a general case.

A low score does not mean a product is dangerous. Most low scores come from a maker being vague, not from anything hazardous being present. A product your vet recommends may score poorly here — follow your vet, not us.

An ingredient flagged "use with care" is not automatically bad. Concentration and context decide almost everything. Tea tree oil at a trace in a rinse-off product is a different proposition from tea tree oil applied neat, and our flag can't tell the difference. That's why the flag links to a page that explains itself instead of just going red.

Where the data comes from

Ingredient entries are compiled from publicly available product labels, manufacturer websites, and general reference material about topical ingredients in animals. Each ingredient carries an evidence level — high, moderate, or low — reflecting how much published information exists about it.

This is a proof of concept. The catalogue is small and hand-entered. The scoring logic is the artefact worth looking at; the data set is a demonstration of it. In production, the ingredient database would need veterinary review, and the product entries would need verification against physical labels rather than marketing pages.

Known limitations

• No concentrations. Labels list ingredients in order, not by amount. An ingredient appearing tenth is treated the same as one appearing second. This is the single biggest weakness in the model.
• Species blindness. Cats and dogs metabolise many compounds very differently. The current model flags this in ingredient text but doesn't score for it.
• Transparency rewards effort, not honesty. A maker who writes a thorough, accurate-looking list scores well. We cannot verify that the list is true.
• Small catalogue. An absent product means it isn't entered — not that it doesn't exist or wasn't worth entering.
• Food scoring reads the bag, not the bowl. An AAFCO statement is a claim about formulation, not proof of quality. We do not test food, model digestibility, or account for an individual animal's needs. A food that scores well here could still be wrong for your pet.
• Additives are counted, not measured. A trace preservative and a heavy dose read identically on a label, and so they score identically here.

What's next

• Barcode scanning and OCR label reading from a phone camera
• User accounts, saved favourites, and pet profiles with known sensitivities
• Recall notifications pulled from public safety notices
• Concentration-aware scoring where labels permit it
• Species-specific weighting, especially for cats and essential oils
• Guaranteed-analysis parsing, so protein and fat ratios score directly

Privacy

insPAWction stores no accounts and collects no personal information. "Where you left off" on the homepage is a single cookie holding the last four product slugs you opened. It never leaves your browser except to ask this site for those products. Clear your cookies and it's gone.

Terms

insPAWction is an educational proof of concept provided as-is, with no warranty of accuracy or fitness for any purpose. Scores are opinions derived from publicly available information using the published formula above. Nothing here is veterinary advice, a diagnosis, or a treatment recommendation.

Brand and product names in this build are illustrative examples created for demonstration. Any resemblance to real products is for the purpose of showing how the model behaves.

If you have a medical concern about your animal, call your veterinarian. If you think your pet has been poisoned, contact an emergency veterinary service or a pet poison control line immediately. Do not consult a website.