When IPO shares are sold sequentially, later potential investors can learn from the purchasing decisions of earlier investors. This can lead rapidly to cascades in which subsequent investors optimally ignore their private information and imitate earlier investors. Although rationing in this situation gives rise to a winner's curse, it is irrelevant. The model predicts that: (1) Offerings succeed or fail rapidly. (2) Demand can be so elastic that even risk-neutral issuers underprice to completely avoid failure. (3) Issuers with good inside information can price their shares so high that they sometimes fail. (4) An underwriter may want to reduce the communication among investors by spreading the selling effort over a more segmented market. CONSIDER A SCENARIO IN which an issuer is selling a new security of uncertain value, for example, an IPO (initial public offering) of stock or high-yield debt, through an underwriter. The S.E.C. has banned variable -price sales. While the value of this new security is highly uncertain to individual market participants, investors hold perfectly accurate information when aggregated. Moreover, there are many (potential) investors, and a small number of these investors can jointly determine the value of the firm (or its project) with high precision. It would seem that in this scenario underpriced offerings would succeed and overpriced offerings would fail. Nevertheless, this paper shows that, if the distribution channels of invest ment banks are limited, underpriced offerings can fail and overpriced offer ings can succeed. With limited distribution channels, it takes the under writer 'time to approach interested investors. Therefore, later investors can observe how well an offering has sold to date -- or at least how successful it has sold relative to offerings previously undertaken by this underwriter.