This paper studies the welfare effects of behavioural policies that change a consumer's default option or their cost of optimizing. I find that such policies, though increasingly popular, lead not just to changes in the welfare of optimizers or default‐takers in the population—the payoff effect—but also to the membership of those groups—the composition effect. This can lead to costly and potentially counterintuitive effects: improving defaults may actually lower welfare, unlike decision simplification, which is unambiguously positive. Such approaches present a useful policy tool but are not always appropriate, and considerable knowledge of preferences is necessary for effective implementation.