Mapping Startup Ambidexterity: An ADO-Guided Systematic Review and Research Agenda

Abstract

Research on startup ambidexterity has grown rapidly, yet the evidence remains difficult to integrate because similar constructs are examined at different levels of analysis and exploration–exploitation is represented in markedly different ways. We systematically review 38 peer-reviewed journal articles published between 2016 and May 2026 and organize the evidence using the Antecedents–Decisions–Outcomes (ADO) framework. Following PRISMA 2020, studies identified through Scopus were screened using explicit document and substantive eligibility criteria. The review identifies six antecedent themes, four recurring decision-related domains, and five groups of outcomes, while also showing that exploration and exploitation are represented through different forms of evidence, ranging from explicit entrepreneurial choices to behavioral processes and broader configurations. The literature spans founder, team, venture, and cross-level relationships, with environmental, team, network, knowledge, and venture-stage conditions qualifying several reported associations. Building on these patterns, we develop a research agenda focused on clearer construct roles, direct examination of entrepreneurial choices, cross-level relationships, boundary conditions, and differentiated outcomes. The review contributes by separating construct role from analytical level, distinguishing direct decision evidence from realized exploration–exploitation patterns, and clarifying how contextual conditions and downstream consequences should be incorporated into cumulative research. The synthesis also clarifies how emerging digital capabilities enter startup ambidexterity as antecedent conditions rather than substitutes for exploration–exploitation. Practically, these distinctions can help founders, investors, and venture-support organizations assess whether a given factor represents an enabling condition, an exploration–exploitation pattern, or a downstream outcome, thereby supporting more targeted diagnosis and intervention.