Randomized controlled trials (RCTs) have become a dominant method for policy evaluation, promising credible causal inference and evidence-based governance. However, their value depends on institutional and political conditions that are often unstable or absent. This paper examines the limits of RCTs and argues for an alternative framework rooted in democratic experimentalism. Rather than treating policy interventions as fixed treatments applied to passive populations, democratic experimentalism emphasizes iterative learning, institutional feedback, and participation by affected actors. Drawing on insights from political economy, institutional theory, and pragmatist philosophy, we explore how experimentation can remain informative even when generalizability is limited and environments are changing. The argument connects methodological debates about external validity to broader questions of governance, accountability, and learning under uncertainty. By situating experiments within institutional contexts, the paper clarifies when experimental evidence can support prediction and when it should instead inform adaptive policy design.