We evaluate the impact of “foreign”fees paid by consumers on their cash withdrawals at automatic teller machines (henceforth ATMs). These so called foreign fees are paid when consumers withdraw cash from ATMs which are not owned by their home bank. We take advantage of a natural experiment whereby (non linear) foreign ATM fees in one specific bank became suddenly applicable at one point in time. We also use this experiment to evaluate the substitutions between foreign ATM withdrawals and other home ATM or desk withdrawals as well as with payments by card. Using panel data on accounts, we first estimate average treatment effects on the treated before carrying on with the estimation of a simple structural model. The latter procedure allows us to compute the counterfactual impacts of changing the schedule of foreign fees. Impacts are sizeable on bank profits and consumer welfare.