International analysis of academic-industry collaboration across major research funders

About the client

The National Centre for Universities and Business (NCUB) is a membership organisation that sits at the intersection of universities and business in the UK. Through initiatives such as their annual State of the Relationship report and Collaboration Progress Monitor, they provide the evidence base that informs how the UK approaches university-business collaboration — work that is particularly vital at a time when post-pandemic pressures are reshaping the landscape of collaborative research and innovation.

Challenge

NCUB wanted a large-scale comparative analysis of academic-industry collaboration across major international research funders, using data available in the open infrastructure. The analysis would contribute evidence to their developing Global Benchmarking Framework for University-Business Collaboration. The framework aims to benchmark the UK’s collaboration performance against other countries, identify transferable practices, and create a repeatable evidence base for decision-makers. Part of this analysis was used to inform the Research Partnerships pillar in Chapter 3 of NCUB’s State of the Relationship 2025 report, where it helped assess the UK’s collaborative competitiveness against a number of comparator countries.

Approach

Drawing on open grant and publication data spanning multiple countries and funders, I:

• Compiled and analysed over 600,000 research grants (2018–2024) from eight public funding repositories across the US (NIH, NSF), UK (UKRI, NIHR), EU (European Commission), France (ANR), Switzerland (SNSF), and Sweden (multiple funders including Vinnova), alongside 1.7 million associated publications retrieved from OpenAlex and Europe PMC.

• Developed a machine learning classification model achieving 93% accuracy to identify which organisations across the grant dataset were in the private sector — a critical step given that most funding agencies do not systematically categorise recipient organisations by type. The model was trained on over 118,000 entries from the Research Organization Registry (ROR) and used when direct funder classifications or exact name matching were not available.

• Mapped the thematic landscape of the grant dataset using SPECTER2 vector embeddings and UMAP dimensionality reduction, overlaid with field-of-study classifications from Semantic Scholar’s open ML model, revealing where business-led research concentrates across disciplines.

• Conducted a citation impact and collaborative trneds analyses demonstrating that academic-industry collaboration is associated with approximately 25% more citations per year and a 65% uplift in field-weighted citation impact (FWCI), with the strongest effects observed for NIHR (+56%), NIH (+34%) and UKRI (+28%).

• Synthesised findings into strategic implications for NCUB, including recommendations to advocate for collaboration-by-design in UK funding programmes, to extend project timescales to enable deeper academic-industry integration, and to press for improved participant-level data reporting across UK funders.

Outcomes

The analysis uncovered significant differences in how funders shape collaboration. Innovate UK’s expansion of business-led grants coincided with a decline in the proportion of those grants that include university partners, while the European Commission’s shift to fewer, larger business-led awards saw a fivefold increase in the rate of academic-industry collaboration per grant. The findings also revealed that when industry participates as a collaborator rather than leading the research, teams are larger, more cross-institutional, and produce higher-impact outputs — evidence that embedding firms within academic networks, rather than simply increasing business leadership, may maximise the diffusion and impact of collaborative research. Part of this analysis was used to inform the Research Partnerships pillar in Chapter 3 of NCUB’s State of the Relationship 2025 report, where it helped assess the UK’s collaborative competitiveness against a number of comparator countries.

The results contribute to a growing body of evidence that the design of funding programmes — not just their budgets — plays a decisive role in determining how effectively academia and industry collaborate. This project was a fascinating opportunity to work with a dataset of unusual breadth, spanning multiple countries and funding systems, and to apply a combination of machine learning, bibliometrics and thematic analysis at scale.