Given the increasing significance of sustainability in investment decisions and regulatory frameworks, Environmental Social and Governance (ESG) ratings for companies are becoming increasingly relevant in the decision-making processes of stakeholders. While large listed companies are mandated to disclose ESG information, the same cannot be said for Small and Medium Enterprises (SMEs). SMEs are not obligated to provide either sustainability information or their own ESG ratings, leaving them susceptible to potential disadvantages in securing capital and attracting investments. Moreover, ESG rating agencies source all the necessary data from the very companies they are meant to assess, leading to an evident conflict of interest.

In this paper, we propose a comprehensive solution to urgently address this gap in ESG disclosure. Leveraging the unique capabilities of Neural Networks (NN) to comprehend and replicate intricate patterns, we train a NN using available environmental and rating data from large companies. The NN learns how to replicate ratings based on the available information. Once the network is adequately trained, we employ it to generate ratings for SMEs that would otherwise lack any form of rating. Another point of innovation is represented by the type of data used, i.e. we utilize data acquired through satellite observations within the European Union (EU) Copernicus Program, ensuring an impartial means of gathering information on environmental activities. Our NN is fed with satellite observations, with the target being the ratings recognized by supervisory agencies. Once the network has been satisfactorily trained and can accurately reproduce the target set of ratings, it is directly applied to the same dataset for a group of SME companies. In doing so, we establish a methodology for consistently rating SMEs’ environmental performance in alignment with the methodology used for larger companies.