Abstract: Digital technologies have brought profound transformations to the entrepreneurial activities, particularly by enabling new forms of interaction between firms and users. Among these changes, interaction orientation—a firm's strategic tendency to interact with and learn from users—has been increasingly recognized as a key enabler of innovation, its effectiveness remains context-dependent and underexplored in digital entrepreneurial settings. Prior research often highlights its linear benefits while overlooking potential drawbacks and the mechanisms through which it operates. This gap is especially salient in the context of digital entrepreneurial, where speed, uncertainty, and limited resources make strategic decisions more complex. To fill this gap, this study draws upon the Lean Startup Theory and investigates the “double-edged sword” effect of interaction orientation on innovation performance in digital new ventures. Specifically, it examines experimentation as a key mediating mechanism and incorporates latent need identification capability as a critical boundary condition that shapes the impact of user interaction. Based on a multi-stage survey of digital new ventures in China, this study provides robust empirical evidence of an inverted U-shaped relationship between interaction orientation and innovation performance. Moderate interaction levels help firms capture relevant market feedback, foster agile development, and achieve better innovation performance. However, when interaction becomes excessive, it generates diminishing returns—leading to information overload, resource strain, and decision-making fatigue due to redundant or conflicting user input. In addition, this study confirms the mediating role between interaction orientation and innovation performance. In dynamic and uncertain digital markets, experimentation enables ventures to iteratively test assumptions, validate minimum viable products (MVPs), and optimize product-market fit. Interaction orientation supplies the raw input—user insights and behavioral signals—while experimentation transforms this input into actionable innovation outcomes. The findings confirm the mediating role of experimentation, thus unpacking the black box of how user interaction drives innovation success. Moreover, this study highlights the moderating effect of latent need identification capability, offering new insights into the conditions under which interaction orientation is most effective. Ventures with strong capabilities to detect unspoken, future-oriented user needs are more adept at filtering valuable signals from user interactions and translating them into focused experiments. The moderation analysis reveals that such ventures reach the optimal point of experimentation at lower levels of interaction orientation, suggesting that latent need identification sharpens the strategic value of firm-user interaction. Theoretically, this study contributes in three key ways. First, this study clarifies the "double-edged sword" effect and underlying mechanism of interaction orientation on innovation performance in digital entrepreneurship contexts, offering a contextualized explanation to reconcile conflicting findings in existing literature. The inconsistent conclusions regarding the relationship between interaction orientation and innovation may stem from insufficient attention to contextual factors. By focusing on the characteristics of digital entrepreneurship, this study provides both theoretical justification and empirical evidence for the nonlinear (inverted U-shaped) effect of interaction orientation on the innovation performance of digital new ventures. It also opens the black box of this relationship by identifying experimentation as a key mediating pathway. Second, this study uncovers the inverted U-shaped relationship between interaction orientation and experimentation, addressing the research gap on the antecedents of experimentation from the demand-side perspective. While prior studies have primarily explored the drivers of entrepreneurial experimentation from the supply side, the highly dynamic nature of user needs in digital contexts necessitates timely sensing and response to guide effective experimentation. This study demonstrates the positive effect of interaction orientation on experimentation in digital new ventures, thereby enriching the antecedent research of entrepreneurial experimentation in digital settings. Third, this study highlights the critical contingency role of latent need identification capability in digital entrepreneurship, expanding the boundary condition literature on interaction orientation. Enabled by digital technologies, digital new vetures are increasingly capable of capturing and interpreting user needs. In this context, the capability to identify latent needs serves as a crucial catalyst for realizing the benefits of interaction orientation, yet it remains underexplored in existing research. Practically, this study highlights key innovation management strategies for digital new ventures. They should leverage digital technologies to interact deeply with users, personalize products and services, and predict latent needs through data analysis. They must remain agile by adjusting innovation directions based on market changes and iterating MVPs rapidly. At the same time, these new ventures should avoid excessive interaction that leads to information overload and resource dispersion. Prioritizing relevant user feedback, integrating information effectively, and allocating resources strategically are crucial for enhancing innovation performance in fast-changing digital environments.