I study financial product innovation in a model with two classes of agents: “sophisticated” and “unsophisticated.” Unsophisticated agents are hit with frictions that lower the return to a conventional asset they hold. Sophisticated agents construct financial innovations that are perfect substitutes for the conventional asset, but are not subject to the friction. In the absence of complete information, unsophisticated agents learn about innovations through a contagion process, as they encounter competitors who have already adopted them. The model yields two equilibria: in one, the innovation persists; in the other, it disappears. Only one equilibrium is stable, and this is determined by the strength of the contagion and by early strategic interactions between sophisticated agents. The model suggests mechanisms for several empirical regularities in the financial innovation literature. Additionally, two applications demonstrate how to estimate the contagion parameter with a short time series of data, and how to use it to predict whether a financial innovation will spread.