Balanced combinations of technological and non-technological innovations, or ‘complex innovation strategies’, are associated with better firm performance. However, the mechanisms through which different internal and external knowledge sources influence the resulting innovation profiles of firms are underexplored. This paper addresses these mechanisms, by establishing a theoretical framework and empirically assessing the direct effect of the main knowledge sources, as well as the indirect effect via absorption and spillovers, on the likelihood of adopting simple or complex innovation strategies. Using data from the three most recent waves of the Italian Community Innovation Survey (2014, 2018, 2020), we find that all sources of knowledge contribute to increase the likelihood of being a complex innovator along an inverted U-shaped curve. Moreover, we find a significant impact of the absorption of knowledge embodied in machinery, equipment and software, while no evidence of spillovers from external applied knowledge, here measured by extramural R&D, significantly emerges.