This thesis explores how early-stage startups navigate uncertainty through structured decision-making and mentorship. Building on research on how entrepreneurs learn and how their decisions are shaped by cognitive limits and unpredictability, it investigates whether mentorship amplifies or alters the impact of a scientific, hypothesis-driven approach to startup development. Through a randomized controlled trial involving 308 aspiring entrepreneurs, participants were trained either in a structured decision method or not, receiving mentorship support. Results show that the scientific approach improves short-term decision quality, leading to more frequent termination of non-viable projects, more iterative pivots, and higher early revenues. Mentorship moderates these effects by supporting reflective thinking and emotional resilience: it leads to fewer terminations, but a higher number of incremental pivots and greater early-stage revenue.