To make a comprehensive policy change, actors often turn to the gradual path where they introduce small-scale changes hoping that their accumulation will meet their goal over time. Nonetheless, they often stop the transformative process before meeting their original goal. This paper argues that this can be explained by policy learning. When actors learn from reliable information that the accumulation of the small-scale changes does not meet their expectations, they stop the transformative process. At the same time, the policy is not illuminated due to feedback effects and beliefs by the majority of actors that the small-scale changes are beneficial.