The service level provided by a retailer influences demand and safes i
n several ways. In the short run, sales can fall short of demand when
customers experience stockouts and choose not to backorder. In the lon
g run, demand itself may decline as customers who experience excessive
stockouts shift permanently to more reliable sources. The traditional
approach to determining inventory policies assumes a fixed stockout c
ost attributable to each stockout occasion. In practice, stockout cost
s are difficult to determine, at least in part because they reflect th
e long-term loss of demand due to inadequate service. In this paper we
model the response of long-run demand to the service level of the ret
ailer, and determine optimal order-up-to inventory policies in the pre
sence of service-sensitive demand. We show under what conditions inven
tory policy is sensitive to both the mean and the variance of long-run
demand.