Research shows that innovations diffuse gradually rather than rapidly, because their usefulness is not immediately apparent to some adopters. These adopters need to learn about the value of the innovation prior to adoption. Although the importance of learning is recognized in most analytical diffusion models, few studies have actually empirically examined the role of learning in the diffusion process. In this paper, we fill this void in the literature by empirically examining the initial adoption and subsequent diffusion process of a performance-based incentive system within a cooperative bank. Based on both qualitative and quantitative field data, we find that agency theory variables explain early adoption. When we examine a longer horizon, we find that although not all local banks adopted immediately, within 3 years almost all local banks had done so. We analyze this process and find that extensive learning-by-doing and strong subsequent learning spillovers led to this relatively rapid diffusion of the system. We further find that the probability with which an initial non-adopter adopts the new incentive system in the next period depends on the strength of that local bank’s social network.