Using data models to grade assets for investment isn’t new. Both private investors and government funding agencies use analytics to make investment decisions. Research Bridge Partners’ data analytics program has taken these best practices and applied them to principal investigators (PIs) and their labs. The result is a data model and set of algorithms that gives us unique insight on not only the research but the personality, productivity, and commercial potential of universities, PIs and their labs.
Academic research is roughly a $72 billion per year industry, and the Administration recently proposed an exponential uptick in funding. We bring data analytics to it, and we share this analysis with academic researchers and their institutions.
Our unit of analysis is the PI, not the university, but we grow the model by “island hopping” from university to university. We started with 14 universities in 2017 and now have over 100. As of Spring 2021, we had about 75,000 PIs in our model, had used the model to develop approximately 400 targets, and have interviewed about half of those to determine our first set of fellows.
We receive two questions about our data analytics program often enough that it makes sense to address them here.
The first question: “Why are we using a proprietary search method instead of the conventional university channels?” In addition to their education and research programs, American universities have built licensing offices, startup incubators and accelerators, and corporate partnering offices. The staff who support these programs work hard to find and adequately resource academic inventors.
We use these channels, both algorithmically and on the ground. Research Bridge Partners’ algorithms weight participation in these programs, and we make sure to talk with the translation professionals and academic leadership wherever we go. In addition, we use our model to independently search for faculty entrepreneurs. Research Bridge Partners’ principals used to run “access point” programs, and none of us were immune from blind spots caused by institutional politics, personal interest, personality fit, limited time, or whatever.
The second question: “What are the underlying features collected on each lab, and how do you assemble your algorithms?”
What we can share are some insights, including the following:
- Collaboration really, really matters. We looked at this from several angles and have developed an analytic perspective on collaboration that generates high correlations with other objective and subjective measures of entrepreneurial potential.
- Gender-related access to commercialization still needs to be addressed at many universities. Academia is now hiring and promoting many more women PIs than in the past. But even after we adjusted for field, career stage, and research quality, women PIs are still commercializing at a much lower rate, overall, than their male peers.
- Research quality is key. The most important thing that an academic co-founder can do, is to invent something amazing. One of the reasons that Stanford is “good at” commercialization is that it has 156 members of the National Academy of Sciences on its faculty. Great research innovation is the prerequisite for great commercialization.
- Both human and artificial intelligence are needed. Algorithms are good at identifying salient trends in vast swaths of research, but they have their limitations. Understanding the disparity in the keywords alternatively preferred in academia and industry, for example, but also from campus to campus. For these and other essential tasks, only human intelligence will do.
We cannot be successful without Research Bridge Partners’ proprietary analytics, but our success does not depend on them alone. Analytics are essential because they enable us to identify individual researchers who have the rare potential to be the scientific co-founders of a success biomedical startup. However, understanding that potential, let alone realizing it, can only be achieved through a deep, collaborative relationship. In other words, our people pick up where our data leave off.