Abstract We study optimal monetary policy design in a simple model that deviates from the linear-quadratic paradigm and provides a rationale for the practice of inflation zone targeting. We show that the presence of either zone-quadratic preferences or a zone-linear relationship between inflation and economic activity provides strong incentives to deviate from conventional linear policies. We calibrate the model based on parameters for the United States and the euro area and employ a numerical dynamic programming algorithm to derive the optimal policies. With this algorithm, we examine the role of uncertainty, model structure and relative preference towards economic stability in determining the width of the implied targeted inflation zone.