The diversification-performance (DP) relationship, a cornerstone of strategic management, has garnered significant attention. While the literature predominantly suggests an inverted-U shape, indicating that an optimal level of diversification exists, direct empirical support for this relationship has been scarce. This is partly due to the absence of continuous measure that accounts for industry-relatedness. To address this gap, this study introduces a novel measure, the Weighted Average of Industry Relatedness (WAIR), which leverages advanced language models to integrate both the degree and the type of diversification. WAIR allows for a comprehensive analysis of single-business, related, and unrelated diversification strategies. Our findings provide empirical evidence for the curvilinear DP relationship in the short term. Moreover, we suggest potential long-term advantages of conglomerate diversification. These results underscore the significance of relatedness in comprehending diversification mechanisms, validate the curvilinear DP relationship, and demonstrate the practical utility of AI methods in strategic management research.
