Key factors and configuration paths of digital innovation in manufacturing enterprises under the TOE framework

Digital innovation is crucial for manufacturing enterprises to enhance competitiveness and achieve leapfrog development. This paper examines Chinese A-share listed manufacturing enterprises from 2012 to 2022, analyzing the effects of individual factors and multi-factor linkages on digital innovation through the TOE framework. It integrates the CatBoost and SHAP machine learning algorithms with a multi-period QCA. The findings are as follows: (1) The CatBoost regression model demonstrates strong explanatory capacity. The SHAP analysis identifies six key determinants of manufacturing digital innovation: market competition, R&D investment, firm size, workforce size, absorptive capacity, and R&D personnel. (2) The multi-period QCA reveals distinct configurational pathways across different stages of digital transformation. In the initial exploration phase, seven configurations emerge, primarily characterized by technology-driven and technology-environment synergies. In the high-speed development phase, eight configurations are identified, highlighting technology-organization and organization-environment synergies. In the acceleration phase, ten configurations are found, illustrating the interaction among technology-driven, technology-organization, technology-environment, and organization-environment factors. (3) Technological and organizational factors remain core conditions for high digital innovation throughout all periods. As digitalization progresses, environmental factors play an increasingly important role. • CatBoost and SHAP capture nonlinear effects and factor interactions on digital innovation. • CatBoost-SHAP integration boosts interpretability and identifies key factors. • Machine learning with multi-period QCA reduces subjectivity, uncovers influence pathways. • TOE framework employs 47 variables in 13 sub-dimensions to enrich multi-factor linkages.