A Deep Dive into NSF Regional Innovation Engine Applications: How We Can Use NSF’s Interactive Application Visualization to Learn More About the Heartland and Innovation 

Jonas Crews


A few weeks after the deadline to submit concept outlines for the National Science Foundation’s (NSF) Regional Innovation Engines program, the economic development world received a pleasant surprise: NSF published an interactive visualization displaying the 679 submissions that advanced to the next round in the process. (Note: this visualization is embedded at the bottom of this page)

From a data perspective, this visualization provides uniquely easy access to the characteristics of advancing applications, helping those of us in the economic development world understand the topical areas in which regions across the country intend to innovate. From an innovation perspective, the associated NSF Engines Concept Outline Contact Form reduces the barriers to collaboration among submitters with similar concept outlines, as well as for anyone else intrigued by one of the submissions; by simply clicking on a concept outline within the visualization anyone can directly contact the submitters of a given concept outline.  

Some readers might be asking, “What is the NSF Engines program?” The NSF Engines program will help regions across the U.S. create thriving innovation ecosystems consisting of strong relationships among for-profit companies, non-profit organizations, government institutions, and academic institutions. Think Silicon Valley, but in locations throughout the U.S. and with the focal point of innovation varying to suit a given community’s unique strengths, needs, and interests. NSF is offering two types of awards: The Type-1 award provides up to $1 million in funding over a one-to two-year period to create an innovation engine. The Type-2 award provides up to $160 million over 10 years to expand and mature innovation ecosystems that are already standing and that are primed for growth. NSF has a stated target of up to 50 Type-1 awards and up to five Type-2 awards for this initial round of applications, but they have publicly stated that they hope to run a future call for Type-2 proposals. The application process is multi-step, starting with concept outline submissions and each subsequent step requiring a more detailed proposal from the teams that advance to that step.¹  

Now that we have a general understanding of the NSF Engines program, let’s go back to the interactive visualization’s collaboration mechanism. Why is it important that potential collaborators have an easy way to learn about and contact submissions? The interactive visual and its associated contact form allow for the cross-pollination of ideas and diversification of perspectives that generally would be missing from the application process. In most cases, little is known about applicants until awards are announced, and submitting teams are made up of members with similar perspectives (i.e., a team of researchers from the same university). This program was designed to allow submitting teams to explore the other 678 concept outline submissions, learning from geographically and/or topically similar submissions, and possibly identifying teams to consider combining with before the next submission deadline. Additionally, non-submitting parties with a vested interest in a submission, or who may be able to bring another prospective, are able to contact concept outline submitters about joining their team or at least providing valuable input before the next submission is due. Through this iterative and collaborative process, NSF expects that the funded proposals will be more comprehensive and consist of more diverse, cross-sector leadership teams than previously funded projects. There’s significant value in team diversity, as a community developer will have a much different perspective on a project’s impact than a university professor, and each of those perspectives is needed to create an optimal proposal. While Type-1 proposals have already been submitted, Type-2 proposals don’t have a deadline until late January. If you believe you could help shape a team’s next submission, I encourage you to use the contact form to reach out! 

Moving to the value of the data beyond the collaboration process, we can use the visualization to understand the distribution of advancing submissions across the country and across topics. This helps us understand the future of innovation both regionally and nationwide, at least from the prospective of the 679 submission teams. It also allows us to see how well represented a region is in the advancing submissions. At Heartland Forward, our first question was whether the heartland was well represented. Thus, I used the “Search by Theme (and more)” tab to isolate advancing concept outlines led by a submitter from a heartland state. There are 263 (out of 679) that meet my criteria, which is roughly 39%; that also happens to be the heartland’s population share.² So, the heartland has its “fair” share of advancing concept outlines, but does that hold true when we consider shares by award type? Looking at only Type-1 concept outlines, a heartland organization was the lead submitter on 182 of the 488 (37.3%), again close to the population share. The heartland’s Type-2 share (42.4%; 81 submissions) is a bit above its population share. Overall, heartland organizations have put forth quality concept outlines at a per-capita rate equal to the non-heartland, and NSF has demonstrated a willingness to look beyond the coasts for innovation investments.  

One quick note before we go into common topical themes across submissions: there are 15 submissions that mention “heartland” somewhere in the outline, and all 15 are aimed at only states we include in our 20-state heartland definition. We must have done a decent job defining the region!  

Our second question was whether the most common topical themes of heartland submissions differ from non-heartland submissions, particularly for Type-2 award submissions. I used the word cloud within the “Search by Theme (and more)” tab to answer this question; the word cloud reflects the frequency of keywords in the filtered list of submissions below it, and we can hover over a term to see the number of times it occurs in the filtered list of submission keywords. The heartland’s top theme for Type-2 award submissions is “Bioeconomy” (15 occurrences out of 81 submissions) followed by “Advanced Manufacturing,” “Artificial Intelligence,” and “Health” (each occurring 12 times). These results align with heartland strengths, advanced manufacturing and agriculture, and needs, improved health outcomes and wider spread deployment of advanced technologies. The non-heartland’s top themes are relatively similar, but also different in some interesting ways. “Advanced Manufacturing,” “Artificial Intelligence,” “Bioeconomy,” and “Health” are among the most-used keywords. However, two other themes were the non-heartland’s most used: “Climate Change” (23 occurrences out of 110 submissions) and “Carbon Reduction” (18 occurrences). Each of these terms does show up at a relatively high frequency in heartland submissions, but their heartland frequency is roughly a third of their non-heartland frequency. This is reasonable, as the West continues to deal with annual water crises, droughts, and forest fires, and Florida has experienced significantly more direct hits from hurricanes than any other state.³ Overall, the differences in keywords serve as a reminder that different parts of the country face unique opportunities and challenges, although the heartland and non-heartland submission themes are more similar than we might expect.  

This post is the first installment of multiple dives into the NSF Regional Innovation Engine data. Future posts will dive deeper into the characteristics of submissions and how they differ across regions and award types. Hopefully, we will soon have the results from further rounds of application vetting and will be able to see how much collaboration this visualization tool spurred.  


1) https://beta.nsf.gov/funding/initiatives/regional-innovation-engines

2)  I use 2020 census data for my population share calculation: https://data.census.gov/cedsci/table?g=0100000US,%240300000&tid=DECENNIALPL2020.P1

3) https://www.nhc.noaa.gov/paststate.shtml