Identify Potential Diversification to Companies through Collaborative Filtering

The industrial fragility of European countries has been a major issue during the latest economic and health crises. Governments have become aware of the partial desertification of our industry but also of the astonishing capacity of the players to reinvent themselves, innovate to find solutions, and show resilience. Everywhere in our society, shaken from all sides, we have seen the emergence of diverse initiatives to overcome the shortages of necessities. In this study, we focus on the ability to diversify from the point of view of a company, which literature has shown to be a major factor in improving industrial resilience. We are interested in the proximity of industrial know-how between two product classes in the HS nomenclature, independent of the country or territory observed. Our goal is to evaluate the ability of a firm that produces product A to adapt its production to produce product B. We analyzed thousands of French companies' websites to label the products they manufacture. From the collected data we built a Recommender System (RS) for diversification based on collaborative filtering (CF). The results show that our Recommender System outperforms methods from macro data analysis, such as co-export analysis on the Product Space or semantic analysis of nomenclatures. We formalize an indicator of a company's agility based on its diversification capabilities. Finally, this work offers new perspectives on the formalization of a measurable Resilience Index (RI).