Mapping patent integration strategies in a multinational firm

Figure 1. Co-invention Network across Countries and Technological Domains

This network visualisation depicts cross-border co-invention ties in Bosch’s AI-related patents filed between 2017 and 2023. Each node represents a country, coloured by dominant technological domain (e.g., propulsion, embedded systems). Directed edges indicate co-patenting flows from subsidiaries to the German headquarters (DE). Node size reflects centrality in the network, and edge thickness indicates the frequency of co-invention ties. The network highlights Germany’s role as the central hub integrating diverse regional contributions, especially in transversal technologies such as artificial intelligence and advanced manufacturing.

Figure 2. Hierarchical Clustering of Inventor Home Countries Based on Co-invention Patterns

This dendrogram shows a hierarchical clustering of inventor home countries based on their involvement in Bosch’s AI-related co-invention activity from 2017 to 2023. The clustering uses Ward’s linkage and rescaled distance metrics to identify similarities in national co-patenting profiles. Each country represents the residence of inventors rather than patent ownership or filing location. The results indicate distinct groupings, with countries such as Germany, Austria, and France forming a tightly connected core, while others like Japan, Vietnam, and Israel appear more peripheral. The analysis captures how geographically distributed inventors are integrated into the firm’s internal innovation network.

This dataset contains structured patent and inventor information used in the analysis of transnational co-invention and patent integration strategies within Robert Bosch GmbH’s innovation network. It includes 380 patents filed between 2017 and 2023 that are classified under artificial intelligence and related transversal technologies.

Variables include:

• Patent ID (anonymised)
• Filing Year
• Jurisdiction of Filing (e.g. EPO, WIPO)
• IPC Technological Classification (e.g. Artificial Intelligence, Embedded Systems)
• Inventor ID (anonymised)
• Inventor Country (home country based on affiliation at filing time)
• Number of Inventors per Patent
• Co-inventor Ties (binary format for network construction)
• Filing Authority
• Assignment Location (corporate entity receiving ownership)

R Script for Estimating Exponential Random Graph Models (ERGM) in Co-Inventor Networks
Description:
This script contains the R code used to construct and analyse co-inventor networks based on patent data. It includes procedures for:

• Importing and formatting adjacency matrices and inventor attribute data
• Building network objects using the network and igraph packages
• Assigning node-level attributes such as inventor affiliation, technological classification, and team size
• Estimating Exponential Random Graph Models (ERGM) using the ergm package
• Testing the significance of network structure, inventor-level effects, and filing-related covariates
• Running diagnostics to evaluate model fit and convergence