This case study is published as part of the Innovate UK BridgeAI programme, under the Independent Scientific Advisor (ISA) offer delivered by The Alan Turing Institute. The ISA offer provides transformative, evidence-based support to SMEs across BridgeAI’s priority sectors, empowering them to harness AI for strategic growth and practical impact. We gratefully acknowledge the contributions of Carla Jouavel, Founder and CEO of TrackCarbon, Carola Zanoletti, Newcastle University PhD student and Professor Tom Jackson, Independent Scientific Advisor for BridgeAI at The Alan Turing Institute, whose insights and engagement were invaluable to the development of this case study. We would like to extend thanks to Alina Abramova, Data Analyst at TrackCarbon who assisted Carola with data collection and analysis. We would also like to express our appreciation to Alexandra Araujo Alvarez, Senior Research Community Manager for BridgeAI; Dominica D'Arcangelo, Programme Manager; and Kathryn Hockman, Project Coordinator, for their leadership and support throughout this work. We further acknowledge Stuart Gillespie for his role as technical writer for this and other case studies within the ISA offer. This work is led by Dr Vera Matser, Head of Strategic Capabilities and Principal Investigator for BridgeAI at The Alan Turing Institute. Abstract This case study explores how UK-based startup TrackCarbon is leveraging data science and machine learning to assess and reduce the environmental impact of food products. In a sector characterised by complex global supply chains and multi-ingredient products, accurately measuring environmental impact remains a significant challenge. TrackCarbon addresses this by developing a scoring system that evaluates products across key indicators, including greenhouse gas emissions, land use, and water consumption. The company was supported by an Independent Scientific Advisor and intern, bringing in a PhD researcher and an expert from The Alan Turing Institute. This collaboration aimed to design and implement a scalable, transparent methodology for environmental impact assessment. The intervention focused on integrating diverse environmental datasets, addressing data quality challenges, and developing a statistical modelling framework to generate consistent product-level scores. This included the application of advanced techniques such as hierarchical modelling and the creation of a weighting system to relate environmental indicators to broader economic proxies. The result was a Python-based pipeline capable of processing, analysing, and scoring both simulated and real-world recipes through a repeatable workflow. The case highlights the value of interdisciplinary collaboration and targeted technical support in accelerating AI adoption within SMEs. It offers practical insights into building data-driven sustainability tools, demonstrating how robust data pipelines and applied statistical methods can enable more transparent, comparable, and actionable environmental impact assessments in the food industry.