Golosov and Lucas recently argued that a menu-cost model, when made consistent with salient features of the microdata, predicts approximate monetary neutrality. I argue here that their model misses, in fact, two important features of the data. First, the distribution of the size of price changes in the data is very dispersed. Second, in the data many price changes are temporary. I study an extension of the simple menu-cost model to a multiproduct setting in which firms face economies of scope in adjusting posted and regular prices. The model, because of its ability to replicate this additional set of microeconomic facts, predicts real effects of monetary policy shocks that are much greater than those in Golosov and Lucas and nearly as large as those in the Calvo model. Although episodes of sales account for roughly 40% of all goods sold in retail stores, the model predicts that these episodes do not contribute much to the flexibility of the aggregate price level.