Current chemical process development aims to improve sustainability. Decision-making thus needs to assess potential environmental benefits. However, reliable life cycle inventory data is often unavailable at early design stages. Without process information, life cycle assessment practitioners usually estimate inventories solely based on the reaction stoichiometry and proxies for energy and utility demands. However, the quality of these proxies has not been tested on a comprehensive data set. In this study, we compare and benchmark stoichiometry-based estimation methods that employ proxies for the yield and utility demands from the literature to a new benchmark database of 474 processes. This benchmark data set is based on industrially validated processes from the Process Economics Program (PEP) yearbook. Most estimation methods are shown to underestimate the global warming impact. We found that the yield range assumed by Geisler et al. ( 2004) closely reflects the actual raw material demands, while the average process energy demands, calculated by Kim and Overcash (2003), perform best as a proxy for energy demands. Thus, we propose to combine both proxies to improve predictions of the inventory data and the overall global warming impact.