The near-rationality hypothesis (Akerlof and Yellen, 1985) holds that the presence of even very small costs of optimization may lead many households to act irrationally. We embed this idea in a standard model of consumption-savings decisions: households pursue simple quick fix consumption policies unless they pay a cost to re-optimize. We design a novel survey to test this theory by eliciting hypothetical consumption responses to a large number of unanticipated income shocks, allowing us to estimate within-subject consumption policies. Consistent with the theory, 68% of households follow one of four simple quick-fix consumption policies that either fully consume or fully save out of small negative and positive shocks before abruptly switching to similar consumption policies for large shocks. Households’ quick-fixing types are essentially unpredictable given rich demographic and economic information but nevertheless explain 49% of the variance in MPCs across households. Quantitatively, an incomplete-markets model calibrated to our survey findings generates almost twice as much size-dependence in the aggregate consumption response to government transfer shocks as the nested rational model. This large difference in behavior arises while households experience consumption-equivalent welfare costs of irrationality of at most $65 per quarter.