Complete Food and Home Cooking
A whitepaper comparing the environmental footprint and health evidence of vegan complete foods — powders formulated to be a whole diet, such as Huel, Jimmy Joy and Mana — against a regional, seasonal, whole-food vegan diet in Germany. Both sides are vegan, which holds the largest dietary variable constant and leaves a narrower question: given that someone eats plants, does the form those plants arrive in change anything?
Headline result
Both options land in roughly 1–3 kg CO₂e per 2000 kcal, against about 10 kg for a meat-heavy diet. The spread within each option is larger than the gap between them, which is the finding that actually matters: the choice between a powder and a cooked meal moves less than the choices made inside either one.
What drives a complete food’s footprint is ingredient refinement rather than manufacturing. Dry blending accounts for 0.2 % of the total, while soy protein isolate costs about six times its parent bean, with 84 % of that coming from processing.
What is in it
- A reconstructed bottom-up cradle-to-gate estimate for a complete food, since no independent product assessment exists for any brand in the category
- Two 2000 kcal day models for the whole-food comparator, one optimised for regional and seasonal produce, one with imported and greenhouse items
- An ingredient-level refinement penalty traced through bean → meal → protein isolate
- The health evidence: the ultra-processed food literature and its subgroup heterogeneity, and the vegan micronutrient and fracture data
Data and method
Whole foods come from Agribalyse 3.1 via the ADEME data API, cradle-to-shelf, split by life-cycle stage, with seasonal and off-season entries listed separately. Industrial ingredients come from CarbonCloud ClimateHub verified product reports at factory gate, split into agriculture, processing and transport. Measured diet-group footprints come from Scarborough et al. 2023 in Nature Food. Manufacturer sustainability pages were read directly, and every effect size and confidence interval is taken from its primary publication.
Every number was verified against its primary source rather than a search summary.
Text under CC BY 4.0. Research and drafting were assisted by an AI tool; every cited source was checked against the primary document, and responsibility for the content rests with the author. Corrections are welcome — open an issue.