The rate a product earns comes from one number: the kilograms of CO₂e behind 100g of the protein it delivers. Everything below is how that number is arrived at, and where it is contestable.
Reward is inversely proportional to carbon intensity. A reference intensity of 2.0 kg CO₂e per 100g of protein — roughly soy — earns the full 1.00×, and anything above it earns proportionally less.
There are no bands. An earlier version bucketed the multiplier into tiers, which meant cheese and pork shared one: 21.0 against 7.6 is a 2.8× difference in footprint, and a shared bucket pays both the same. Deriving the rate from the footprint directly removes the boundary entirely, so the ratio in the reward is the ratio in the footprint — nothing is protected by the category it sits in, and no bucket edge can compress or invert anything.
The cap exists so that cleaner-than-reference sources do not pay a premium — pea protein at 0.4 would otherwise earn 5×, and would run away with the pool. The tier names still shown beside a result are labels only; they carry no weight in what anyone earns.
Published footprints are usually given per kilogram of food. That comparison flatters anything watery: tofu, milk and yoghurt all look excellent per kilo, largely because most of the kilo is water. Per gram of protein delivered is the comparison that matters to someone buying protein, so every figure in this app is converted to that basis before it is used.
This is also why a product needs enough protein per 100g for protein to be the point of the food. Below that, rewarding it per gram of protein would mostly be rewarding the rest of the packet.
Poore & Nemecek (2018), Science — a meta-analysis covering roughly 38,000 farms in 119 countries and 40 products, which is the broadest consistent dataset published. The figures used here are global medians, not brand-specific and not farm-specific.
That is a real limitation and it cuts both ways: a genuinely low-impact producer is not credited for it, and neither is a bad one penalised. Within a product category the spread between the best and worst decile is often larger than the gap between two categories. Per Gram rewards the choice of protein source, which is the part a shopper controls in the aisle; it does not claim to measure the particular farm behind the particular packet.
Product data — ingredients, protein content, brand — comes from Open Food Facts, a community database. Coverage is better in Europe than in the US, and a product that is missing there cannot be scanned here.
A dairy herd produces milk. That milk becomes several things, and the herd's emissions have to be divided between them. There is no physically correct answer to how, and the two methods in common use give very different numbers for whey.
Economic allocation apportions emissions by the value of what the process yields. Whey is a by-product of cheese making: historically it was a disposal problem, and even now it is worth a fraction of the cheese it comes off. So it carries a fraction of the herd's footprint rather than a full dairy share.
Mass-balance accounting divides emissions evenly across everything the herd produces, by mass. On that basis whey carries the same burden per kilo as anything else the milk becomes.
| Source | This app | Mass balance |
|---|---|---|
| Whey or casein | 3.6 | ~8–9 |
| Milk protein | 9.5 | ~9.5 |
| Cheese | 21.0 | lower |
Under mass balance whey lands nearer 8–9, close to milk protein itself — which would move it from the second tier to the third and roughly halve what a whey product earns. Both methods appear in published work, including in Poore & Nemecek. Neither is a mistake.
What would be a mistake is applying one at one end and the other at the other. If whey is discounted as a by-product, the discount has to land somewhere, and it lands on cheese. Cheese is costed here at roughly double milk protein because it takes about ten litres of milk per kilo of cheese, and because it carries the share of the herd's footprint that whey is discounted for. Costing whey by economic allocation and cheese by mass would flatter both.
Transport, packaging, refrigeration and cooking are outside these figures. They are real, but for most protein they are small next to production: for beef, farm-stage emissions dominate the total by an order of magnitude, and air-freighted produce is the exception rather than the rule. Including them would add precision the underlying data cannot support.
Land use change is included in the source figures where the dataset attributes it. Carbon opportunity cost — the emissions avoided if the land were not farmed at all — is not; it would raise every animal figure substantially and is not settled enough to reward or penalise anyone on.
