Low Emission Zones (LEZs) -restricted areas for vehicles that fail to meet certain environmental standards- are becoming an increasingly popular policy tool to encourage the adoption of low emission vehicles. Despite this, little is known about their optimal design, particularly regarding which vehicles to restrict and when, and their geographic extension. Unlike price-based instruments (e.g., taxes), LEZs generate convex welfare functions, leading to «bang-bang» solutions: (i) the LEZ should be either as large as politically feasible or not implemented at all if it is too small; (ii) restrictions to the LEZ should be announced in advance, requiring commitment; and (iii) less-polluting vehicles (e.g., hybrids) should be treated as transition technologies -equally restricted but at a later stage. Combining reduced-form and structural analysis, we assess Madrid’s LEZ through the lens of this bang-bang logic.