Research
My research studies how knowledge, innovation, discovery and entrepreneurship are organized at massive scale. This focus has led me to platform organizations and the digital infrastructure that supports them. On the one hand, I study conventional platforms, such as software and app ecosystems, innovation contests and digital marketplaces. On the other hand, I study the institutions of science and universities as scalable, digitally enabled organizations built around platform hubs. I tend to begin with fundamental questions of organization and governance, and then work to understand how novel practices and infrastructure, often digitally enabled, can better organize knowledge, innovation, discovery and entrepreneurship. Algorithms, data and AI have therefore long been embedded in this agenda. I have a keen interest in innovating the modern organization itself, whether firms and platforms or universities and scientific institutions.
The participants include the developers who build products on digital platforms, the solvers who enter innovation contests, the scientists who choose collaborators and projects, and the students and early-career workers who decide whether to found firms. In each setting a platform owner, contest sponsor, university or employer makes design choices, and I study how those choices shape who enters, how much effort entrants invest, and what gets produced. Entry and selection sit at the center of nearly all of this work, which is why it belongs as naturally in an entrepreneurship department as in strategy or innovation.
About half of my research uses field experiments that my coauthors and I designed and ran inside live organizations, including the NASA Tournament Lab, Harvard Medical School, a large university engineering co-op program, and platforms built and operated at Northeastern by teams I assembled. The other half is largely econometric, with an emphasis on novel research designs applied to large observational datasets, such as those on software and mobile-app markets, complemented by historical analysis, formal theory and structural analysis.
The work is also grounded in experience of leading technology and strategy work. Before entering academia, I worked in strategy consulting at Deloitte, was an early employee at Qualcomm leading in-country M&A teams in Latin America, and built and led the Western European technology advisory practice of the Economist Intelligence Unit. I later directed the IoT Open Innovation Lab at Northeastern. That experience of building and running organizations is why I prefer to study them from the inside, through experiments with partner organizations and on platforms built for the research by teams I assembled. The work has appeared in Management Science, Organization Science, the Strategic Management Journal, Research Policy, the RAND Journal of Economics, the Review of Economics and Statistics, Science and Nature Biotechnology. It has received the INFORMS Technology, Innovation Management and Entrepreneurship Best Paper Award (2017), two Copeland Best Paper Awards (2017, 2022), and a Management Science Best Paper finalist designation (2013). I have been a Research Associate of the NBER since 2016 and an Associate Editor of Management Science since 2011.
Every developer who builds on a platform makes an entry decision, often as a very small firm or as an individual, so platforms are in large part markets for entrepreneurial entry. My early work showed that the way a platform is opened governs this entry. Across 21 handheld computing systems from 1990 to 2004, granting outside developers access to the platform was associated with up to a fivefold acceleration in new device development, while giving up control of the platform itself added an effect an order of magnitude smaller (Boudreau 2010, Management Science). Adding large numbers of application producers to handheld platforms generated more software variety, but later cohorts produced less compelling software, and crowding among similar applications made the average effect on innovation incentives negative (Boudreau 2012, Organization Science). Where complementors are unpaid, as on 85 online multiplayer game platforms, attracting more of them had a net zero effect on ongoing development and failed to produce network effects (Boudreau and Jeppesen 2015, Strategic Management Journal).
A second thread asks who these entrants are. In mobile-app data, small changes in platform design, including the minimum cost of building an app and the non-pecuniary payoffs of doing so, can cause the "bottom to fall out" of a market to amateurs. That flood of low-quality entrants did not crowd out the development activity of top developers and coincided with more of the highest-quality products (Boudreau 2018, NBER Working Paper 24512). My current paper in this line, "Long Tails on Platforms: Structural Conditions and Differences in Participation and Innovation," is at the revise-and-resubmit stage at Organization Science. Using 501 product-market niches in a mobile platform ecosystem, it finds that low minimum development complexity, high intrinsic task appeal and high revenue opportunity each draw more participation at the extreme periphery, yet they lead to different development patterns. High task appeal is associated with more development effort, deeper iteration and more differentiated offerings, whereas low development complexity is associated with more entry and less subsequent development. With Lars Bo Jeppesen and Milan Miric, I have also studied how app developers protect and profit from their innovations (Research Policy 2019, 2022) and how freemium strategies interact with network effects to favor market leaders over followers (Strategic Management Journal 2022).
A new platform venture faces a chicken-and-egg problem that is central to entrepreneurship and hard to study causally. In a field experiment in which invitations to join a newly launched platform went to 16,349 individuals, I randomized statements about the size of the expected future installed base (Boudreau 2021, Management Science; Copeland Best Paper Award). Simple, uncommitted and nearly costless statements about the future influenced adoption more than disclosures of the true current installed base during early adoption, and they ceased to have any effect once the true installed base grew large. Early ventures also need money. In "Crowdfunding as Donations to Entrepreneurial Firms" (Research Policy 2021, with Jeppesen, Reichstein and Rullani), we theorize and show that reward- and donation-based crowdfunding has the payoff structure of a public-goods contribution problem. As a result, the tangible value of a project's outputs has little influence on contributions, and the non-pecuniary motivations of funders do much of the work.
Contests are among the oldest ways to organize innovation outside a firm's boundary, and they give unusually clean evidence on entry and effort. Using 9,661 software contests, Nicola Lacetera, Karim Lakhani and I showed that adding competitors reduces every competitor's incentive to invest but raises the chance that at least one finds an extreme-value solution, so that larger contests perform better on more uncertain problems (Management Science 2011). With random assignment of 2,775 contestants to 755 contests, Lakhani, Michael Menietti and I found that most contestants respond negatively to added rivals while the most skilled respond positively, and we used the estimates to evaluate contest design policies (RAND Journal of Economics 2016). Related work on disclosure and cumulative innovation (Research Policy 2015) and on prize-based contests for computational biology problems (Nature Biotechnology 2013) grew out of my role as Chief Economist of the NASA Tournament Lab from 2011 to 2015.
Several of my field experiments ask who takes part in innovative and entrepreneurial work when the terms of participation change. With Nilam Kaushik, I invited 97,678 adults to join a product development platform and randomized whether the opportunity was framed as competitive or collaborative (Organization Science 2023). Outside STEM fields, men and women differed in their willingness to participate under competition, as prior laboratory studies predict; among those in STEM fields, we found no statistical gender difference. With Sarah Bana, I ran a field experiment with 4,465 jobseekers on a matching platform, varying whether the same candidate–job match was attributed to an AI algorithm, a human employer, or no source. AI disclosure reduced engagement by about 26 percent, and the decline was concentrated among better-matched jobseekers, which changed the composition of the pool that engaged (under review at Management Science). In a working paper with Matt Marx, I use the randomized assignment of engineering students to an early or late first work term to estimate how the timing of professional exposure affects long-run retention in STEM careers.
The same questions of search, matching and selection arise in science, where new knowledge is produced by independent researchers who choose their own projects and collaborators. In a field experiment at Harvard Medical School, randomly assigned 90-minute information-sharing sessions increased the probability that a pair of scientists co-applied for a grant by 75 percent (Review of Economics and Statistics 2017). In a grant process we designed with 2,130 randomized evaluator–proposal pairs, evaluators gave systematically lower scores to proposals close to their own expertise and to highly novel proposals (Management Science 2016; Copeland Best Paper Award). My 2026 NBER chapter, "Field Experiments in the Science of Science: Lessons from Peer Review and the Evaluation of New Knowledge," takes stock of this experimental literature.
I have run more than a dozen large-scale field experiments with government, business and university partners. Each one required an intervention that answered a research question without degrading the partner's operations. At Northeastern, I led the multidisciplinary teams that built the digital infrastructure for four platforms, including the IoT Open Innovation Lab and the NUgig experiential-work platform, with funding from the Kauffman Foundation, GE and D'Amore-McKim. Earlier work was funded by Google, Microsoft Research, the Sloan Foundation and NASA. Building these platforms put me on the operator's side of the launch, participation and governance problems that my papers study.
My agenda for the next decade centers on two large questions, with several projects already under way in each.
We are in the early stages of a major general-purpose technology shock, and the first question is how it will reshape platforms, the modern corporation and new ventures. My working hypothesis is that both established firms and the born-digital organizations we have come to know, platforms included, will be substantially transformed within this decade, and far more over the next ten to twenty years: in how they are organized, where their boundaries lie, and how they create and capture value. The questions run from which layers of the technology stack will remain open to entry, to how work, expertise and matching are reorganized, to what new forms of organization emerge. Several projects are first steps. My 2026 MIT Sloan Management Review article, “Building on AI’s Unfinished Foundation,” argues that the architectures needed for AI to act as a general-purpose technology have not yet settled, and a companion paper examines how general-purpose technologies become platforms. The Bana experiment shows that disclosing AI changes who engages with a matching platform, and with Thomas Jungbauer and Cole Williams I am developing a theory of how AI changes search and matching in labor markets.
The second question is how the institutions of science and the university can be organized to have greater impact. Universities organize knowledge, talent, work experience, technology transfer and new venture formation at enormous scale, and like the platforms I study, they can be redesigned. “Entrepreneurship Apprenticeship: Startup Work Exposure and the Decision to Found,” with Matt Marx and Luna Wu (Cornell), asks whether working at a startup during a student work term raises the likelihood of founding a firm, using the quasi-random assignment of about 45,000 students to placement coordinators who differ in how often they place students at startups; the estimates are preliminary. With Ina Ganguli (UMass Amherst), I am completing a field experiment in which faculty were randomly invited to a multidisciplinary or a standard call for proposals; the multidisciplinary framing reduced interest, and the paper now asks who selects into multidisciplinary work anyway. With Nilam Kaushik, a team experiment that randomly assigned 872 participants to 218 teams finds that relevant expertise raised performance mainly when two or more members shared a domain. I have also developed ideas for platforming university technology transfer and for experimenting with how it is organized, which I have not yet put into practice.
AI raises a further question for these institutions: where human expertise keeps its value. In a theory paper begun in 2026, “The Comparative Advantage of Human Expertise: Evidence and Theory from Peer Review,” I argue that the comparison between experts and AI is one of comparative rather than absolute advantage: experts hold the advantage in setting the standard for genuinely novel claims that cannot yet be formalized, and peer review draws its value from the diversity of independently committed judgments, whereas AI excels at fast, scalable checking against the existing record. The paper proposes empirical tests, including a design that uses the fixed training cutoffs of AI models. With Fabrizio Dell’Acqua, Karim Lakhani and coauthors, I am also comparing the current wave of AI adoption in science with earlier waves.
The Long Tails paper leaves open why task appeal matters so much for who enters. With Laura Dudley and Adam Ma, I am completing “Institutional Differences across Platforms and Variation in Complementors’ Non-Pecuniary Motivations,” a meta-analysis of survey-based studies of contributor motivation across five platform regimes: open-source software, Wikipedia, crowdsourcing contests, citizen science and app developers. The paper asks whether institutional differences among platforms are associated with differences in the sources of complementors’ non-pecuniary motivation, and it examines how robust the literature’s measured differences are once the underlying coding is re-verified against the original sources.
Every paper referred to here is listed in full, with its methods, data and a downloadable PDF, on Papers & Projects. How each study was carried out is set out on Papers by Methodology.