Large-scale mapping of retail food and beverage products to environmental sustainability metrics

Current dietary patterns are suboptimal for both human and planetary health (1,2) . With growing consumer and business concerns around food sustainability, estimating the environmental footprint of foods and diets is pertinent. In many countries, supermarkets are the primary provider of foods and beverages; therefore, supermarket purchasing records represent a novel source of population dietary data that offers advantages over traditional methods (3) . We developed a method for mapping greenhouse gas emissions (GHGE) to food and beverage products from a high-street retailer’s portfolio, to enable the estimation of the environmental footprint of population diets when linked with sales information. We used data from the food and beverage portfolio of a high-street retailer in the United Kingdom (UK), including product name/description, categorisation, ingredients, and weight. We mapped these products to GHGE (kg CO 2 -eq/kg) using a global database on the average environmental footprint of food commodities (4) . This mapping process involved three stages utilising different mapping approaches, guided by product sales data, which we extracted from the retailer’s loyaltycard transactions for Yorkshire and the Humber (UK) region during 2022. Stage 1 involved categorising the products into Living Costs and Food Survey food categories and mapping each category to GHGE, where possible (food-category approach). Stage 2 involved splitting selected food categories (based on complexity, necessity of a better mapping, and sales) and creating a sub-category-specific mapping based on an indicator product, which was selected as most popular using sales data (food-sub-category approach). The indicator-product mapping represented a weighted average GHGE value calculated using information on product ingredients and their estimated proportions (ingredient approach). Stage 3 utilised word-searches in product descriptions to distinguish further between product types within selected prioritised subcategories. We used the estimated product-level GHGE (mapped GHGE × product weight) and sales data to estimate food-category contributions to total GHGE and assess how these estimations change by mapping stage. Of >28,000 products, 77.7%, 98.0% and 98.6% were mapped to GHGE at the end of stages 1, 2 and 3, respectively. Of the final product mappings, 40% were at a food-category level and 60% at least at a sub-category level. We calculated 153 product-specific GHGE using ingredients information for prioritised indicator products. When using mappings from stage 3 vs 1, the contributions of ‘savoury snacks’ and ‘chocolate’ to total GHGE were approximately four and two times higher respectively, due largely to improved mapping that accounted for product sub-category and ingredients. Mapping environmental sustainability metrics to a retail product dataset is feasible when using a staged approach, guided and prioritised by sales data. However, mapping approach and the estimations’ variability should be considered. This method could be used for estimating the environmental footprint from food purchasing data, helping to inform responses towards promoting healthier and more sustainable diets.