Innovation performance in corporations is strongly influenced by how firms are embedded within collaboration networks, according to new research analyzing 22 years of patent data from the global 3D printing industry. The study, published in Frontiers of Engineering Management in 2025, reveals that both within-community and cross-community embeddedness promote higher innovation output, but collaboration complementarity plays a crucial moderating role that explains previously inconsistent findings in innovation research.
The research analyzed global patent data from 6,109 organizations over 22 years, constructing collaboration networks based on co-patenting activities to identify innovation communities using topological clustering methods. Researchers from Beijing University of Posts and Telecommunications, Tsinghua University, the Higher School of Economics in Russia, and collaborators published their findings with DOI:10.1007/s42524-025-4188-x, providing new evidence for how network structure drives innovation outcomes in technology-driven industries.
The study distinguishes between two types of embeddedness: within-community embeddedness representing ties to peers inside the same innovation community, and cross-community embeddedness reflecting ties that bridge multiple communities. Both types significantly enhance innovation output as measured by annual firm-level patent counts. Within-community connections provide trusted access to shared knowledge, allowing faster resource integration and reducing coordination costs, while cross-community ties offer diverse expertise and non-redundant information that broaden innovation perspectives.
However, the research reveals that collaboration complementarity—the degree to which partners' knowledge and resources complement each other—plays a pivotal contingency role. When complementarity is high, firms gain more from within-community relational embeddedness, but the innovation benefits of cross-community collaboration weaken due to integration complexity and resource absorption costs. This finding clarifies why previous studies reported inconsistent results regarding network embeddedness and innovation, creating a theoretical gap regarding when embeddedness enhances innovation and when it becomes a burden.
The analytical approach used rolling five-year windows to construct global collaboration networks, identifying innovation communities through Louvain topological clustering. Both relational and structural dimensions of embeddedness were evaluated using negative binomial regression models. The complete research is available at https://doi.org/10.1007/s42524-025-4188-x, providing detailed methodology and findings.
These insights have significant implications for firms pursuing innovation advantage in increasingly interconnected ecosystems. Companies embedded deeply in innovation communities may strengthen internal ties to leverage complementarities while selectively bridging external communities to maintain diversity of ideas. The research suggests that innovation is not only about forming partnerships but about forming the right partnerships in the right network positions, with dense internal ties accelerating trust and knowledge transfer while cross-community ties introduce novel perspectives.
For policymakers, the study offers guidance for industrial cluster development, promoting cross-sector collaboration, and designing incentive mechanisms for innovation-driven industries. The analytical framework is also applicable to emerging domains such as artificial intelligence, new materials, and biomanufacturing, suggesting that optimizing embeddedness and collaboration patterns can accelerate technological progress and enhance competitiveness in global innovation ecosystems.
