Why Startups Fail to Scale:
It’s Not the Product,
It’s the Organizational Design.

When Do Startups Scale? Large-scale Evidence from Job Postings

Lee, S. & Kim, J. D. — Strategic Management Journal, 45(9): 1633–1669, 2024 Peer Reviewed

Scaling at the right time is a crucial challenge for startups. Conceptualizing “scaling” as the entrepreneurial process of acquiring and committing resources to implement the core business idea and expand the customer base, this study examines how scaling early may decrease imitation risk at the expense of increasing commitment risk. As startups typically hire managers and sales personnel when they begin to scale, we propose that this timing can be empirically measured by when startups first post these jobs. Leveraging a dataset of job postings, we find that early scalers are more likely to fail, but no evidence of a countervailing benefit in terms of successful exit. Additional analyses suggest that the commitment risk in scaling early outweighs the benefit of reducing imitation risk.

The Effect of Flatter Hierarchy on Applicant Pool Gender Diversity: Evidence from Experiments

Hurst, R., Lee, S., & Frake, J. — Strategic Management Journal, 45(8): 1446–1484, 2024 Peer Reviewed

This article investigates how job seekers’ perceptions of an employer’s formal hierarchy affect the size and gender composition of its applicant pool. Building on the literature on gendered organizations and organizational design, we develop opposing perspectives on these relationships. To arbitrate between these perspectives, we first conduct a field experiment in partnership with a hiring firm. We find that featuring a flatter hierarchy in recruiting materials does not significantly affect the size of the applicant pool, but significantly decreases women’s representation within it. Our follow-up survey experiment identifies several potential mechanisms (e.g., perceptions of career progression, informality, workload, and fit). Our findings imply that firms’ growing tendency to adopt flatter hierarchies could inadvertently undermine efforts to attract a greater proportion of women applicants.

The Myth of the Flat Start-up: Reconsidering the Organizational Structure of Start-ups

Lee, S. — Strategic Management Journal, 43(1): 58–92, 2022 Peer Reviewed

There has been an ongoing debate over whether start-ups should be “flat” with minimal hierarchical layers. To reconcile this debate, this article distinguishes between creative and commercial success (i.e., novelty vs. profitability), and examines how these outcomes are variously influenced by a start-up’s hierarchy. This study suggests that while a flatter hierarchy can improve ideation and creative success, it can result in haphazard execution and commercial failure by overwhelming managers with the burden of direction and causing subordinates to drift into power struggles and aimless idea explorations. I find empirical support for this trade-off using a large sample of game development start-ups. These findings offer one resolution to the debate by sorting out the conditions under which hierarchy can be conducive or detrimental to start-ups.

Cognitive and Structural Antecedents of Innovation: A Large-Sample Study

Lee, S. & Csaszar, F. A. — Strategy Science, 5(2): 71–97, 2020 Peer Reviewed

This paper studies how cognitive and structural antecedents affect adaptation to disruptive innovations. We do so by analyzing how video game firms adapted to the free-to-play business model around the period of disruption (2012–2015). Our data set (which contains 461 firms, collectively employing 83,157 individuals) allows us to characterize each firm’s organizational structure and each employee’s experience profile; it also captures the performance of firms under the existing and new technological regimes (that is, firms that do and do not adopt the disruptive innovation). We show that adoption, implementation under the existing regime, and implementation under the new regime are affected by cognitive and structural antecedents in different and often opposite ways. We also point out conditions under which cognitive and structural antecedents can compensate for each other. Overall, our study contributes to a better understanding of how firms should organize to face disruptive innovations.

Target Startup's Organizational Structure and Acquirer's Integration-Separation Decision

Kim, J. D. & Lee, S.

To reap the benefits of startup acquisition, acquirers strategically choose whether to integrate or separate their targets. This study examines how this post-acquisition organization design choice between integration and separation depends on the target startup’s pre-acquisition organizational structure. Leveraging employee-employer-matched data to construct a novel measure of post-acquisition separation based on employee mobility patterns, we find that acquirers are more likely to separate more structurally complex target startups. Additional analyses are consistent with the view that, as the target firm’s structural complexity can increase the disruptive costs of integrating the two firms but reduce coordination costs when kept separate, post-acquisition separation becomes more suitable. Our results shed light on an important organizational antecedent that shapes the relative appeal of integration versus separation in startup acquisitions.

Span of Control as a Dynamic Strategic Lever for Early-stage Firm Growth

Choi, J. & Lee, S.

Despite its centrality in classical management theory, the span of control has become a “forgotten pillar” in modern organizational design. We revitalize this construct by developing a computational model that formalizes how the span of control reconciles two fundamental, yet previously isolated, information-processing trade-offs: the tension between individual and organizational bottlenecks (driven by attentional constraints) and the tension between commission and omission errors (driven by evaluative constraints). We demonstrate that while either constraint in isolation favors the widest possible span of control, their concurrent interaction compels an intermediate optimum. Furthermore, we challenge established structural contingencies by revealing that a narrower span of control may prove superior not only in highly uncertain environments but also in highly munificent environments. Finally, we uncover a dynamic structural imperative: contrary to the prevailing view that the span of control should widen with size, we show that the optimum systematically decreases as firms grow. This progressive narrowing reflects a transition in the binding constraint on growth: from a scarcity of ideas in nascent ventures to a scarcity of time in mature firms. Together, these findings reestablish the span of control not as a static parameter, but as a strategic lever for navigating the evolving information-processing trade-offs.

When Autonomy Backfires: Adverse Selection in Startup Recruitment

Lee, S. & Hurst, R.

To compete with established firms for talent, startups often highlight their distinctive organizational attributes in recruitment efforts. Using a dataset of 228 million job postings, we first document that autonomy is a particularly prominent attribute: startups increasingly emphasize this organizational design feature, especially when targeting more educated candidates. To assess the efficacy of this strategy, we conduct a pre-registered field experiment in partnership with an actual startup. The results reveal a counterintuitive effect: explicitly emphasizing autonomy attracts less educated candidates while deterring their more educated counterparts. A follow-up survey experiment suggests that this adverse selection likely arises from education-contingent differences in interpretation. Specifically, more educated job seekers are more likely to construe this emphasis as (1) a marker of generalist work misaligned with their specialized human capital, (2) a sign of organizational dysfunction stemming from excessive autonomy, (3) a compensating differential substituting for inadequate pecuniary benefits, or (4) simply “cheap talk” that undermines employer credibility. Together, our findings demonstrate that, although autonomy is a core attribute that distinguishes entrepreneurial ventures from established firms, its explicit emphasis can paradoxically backfire—undermining efforts to attract the high-ability human capital critical to entrepreneurial success.

Sociopolitical Stance-taking and Labor Market Sorting: Evidence from Experiments

Hurst, R. & Lee, S.

We investigate whether corporate sociopolitical stance-taking drives political sorting. In a pre-registered field experiment with an actual employer, we surprisingly find that a progressive stance does not cause widespread sorting among Democrats and Republicans. Instead, it elicits a concentrated, asymmetric response: a sharp negative effect limited to strong Republicans. A supplementary survey suggests that this distinctly repellent effect may arise from this group’s intense ideological misalignment and negative inferences about the political composition of the workplace. These findings suggest that, rather than reshaping the broad applicant pool, sociopolitical stances function as a “filter,” selectively deterring the subset of candidates most likely to bring their politics into the workplace. We discuss how this filtering effect may complement organizational objectives while potentially deepening societal polarization.

The Effect of Immigration Policy on Founding Location Choice: Evidence from Canada’s Start-up Visa Program

Lee, S. & Glennon, B.

Immigrant founders play a disproportionate role in high-growth entrepreneurship, yet their founding location choices are uniquely bounded by immigration institutions that vary significantly across countries. We examine whether entrepreneurship-focused immigration policies influence cross-border shifts in where immigrant entrepreneurs found new ventures by analyzing Canada’s Start-up Visa Program. Using longitudinal data on 1.3 million individuals residing in the U.S. from 2009 to 2020, we show that following the program’s introduction in 2013, U.S.-based immigrants became 145% more likely than U.S. natives to migrate and found companies in Canada. To explain these patterns, we develop an institutional push–pull framework that highlights how binding legal constraints in an immigrant’s current host country, combined with comparatively attractive institutional and social conditions in an alternative destination, can drive the cross-border reallocation of venture creation. Our findings demonstrate that immigration policy does not merely regulate entry but actively reshapes the international geography of venture formation.

Research: When Should Startups Scale?

Kim, J. D. & Lee, S. — Harvard Business Review, 2024

Silicon Valley often touts rapid scaling as the best strategy for achieving startup success. However, new research reveals that scaling early, particularly within the first 12 months, significantly raises the risk of startup failure, especially for two-sided platforms. The key takeaway for entrepreneurs: Be cautious with early scaling and prioritize a culture of experimentation. Taking a “slow-and-steady” approach to scaling may provide a more sustainable path to long-term success.

Research: Flat Hierarchies Can Discourage Women Applicants

Hurst, R., Lee, S., & Frake, J. — Harvard Business Review, 2024

Many firms struggle to attract a diverse talent pool, particularly women. To stand out, companies often tout flatter organizational structures in their recruitment materials, assuming these unique features will appeal to a more diverse group of prospective employees. However, new research uncovers a surprising finding: highlighting a flatter hierarchy may, instead, diminish women’s representation in the applicant pool. Additional findings suggest that women tend to perceive flatter organizations as more difficult to fit into, burdening them with a heavier workload, and offering fewer career advancement opportunities. This research indicates that companies should carefully consider how they present their organizational structure to job seekers to avoid unintentionally discouraging women from applying.

Prose: Personalizing Beauty at Scale

Lee, S. — 2025

Liquid Death: Murdering Thirst or Drowning in Competition?

Lee, S. — 2025

STX Offshore and Shipbuilding Co.: Cross-border M&A of Aker Yards

Song, J. & Lee, S. — SNU Business Case Series (35), 2013

Samsung Electronics: Innovation and Growth Strategy in Semiconductor Division

Song, J. & Lee, S. — SNU Business Case Series (32), 2013

Hyundai Motor, Driving Global

Song, J. & Lee, S. — SNU Business Case Series (29), 2013

Method Atlas

Method Atlas is an open-access, interactive resource for learning causal inference methods. It covers 27 methods, from the established workhorses (OLS, fixed effects, DiD) to frontier approaches (double/debiased ML, causal forests, synthetic DiD). Each method includes interactive lessons, real-data examples from management research, and hands-on coding labs in R, Stata, and Python. 8 research practices complement the methods: sensitivity analysis, pre-registration, power analysis, randomization inference, specification curves, multiple testing corrections, clustering and few-cluster inference, and Lee bounds for attrition. Together, they cover the full pipeline from research design to credible reporting.

Patent World

PatentWorld is an interactive data exploration platform that presents 50 years of United States patent activity through quantitative visualizations. The project aims to render the complex dynamics of the patent system accessible to students, researchers, policymakers, and the broader public through rigorous, data-driven analysis.