This study examines retention strategies for Generation Z employees within the high-pressure startup ecosystem in Indonesia. The rapid growth of startups has created highly dynamic, competitive, and demanding work environments, making employee retention a critical issue, particularly for younger workers who tend to value meaningful work, flexibility, supportive leadership, and opportunities for growth. This research aims to explore how Generation Z employees interpret retention strategies implemented by startups and how such strategies shape their willingness to remain in the organization. Using a qualitative approach with a case study design, this study collected data through in-depth interviews, observation, and documentation involving Generation Z employees working in Indonesian startups. The data were analyzed through data condensation, data display, and conclusion drawing/verification. The findings reveal that retention strategies in high-pressure startup settings are not primarily determined by financial incentives alone, but by the quality of employees’ daily work experiences. Open communication, supervisor support, work flexibility, recognition, and learning opportunities emerged as the most influential factors shaping employees’ decisions to stay. The study further shows that high work pressure does not necessarily lead to turnover intention when the organization provides meaningful support and creates a psychologically safe and developmental work environment. Theoretically, this study contributes to the literature on employee retention by emphasizing a relational and experience-based perspective, especially in the context of Generation Z and startup organizations. Practically, the study suggests that startup managers should design retention strategies that integrate performance demands with employee well-being, supportive leadership, and career development. This study is limited to the context of Indonesian startups and Generation Z employees; therefore, future studies are encouraged to examine other sectors, compare different generations, and employ mixed-method approaches for broader generalization.