This paper estimates demand for Internet portals using a clickstream data panel of 2654 users. It shows that familiar econometric methodologies used to study grocery store scanner data can be applied to analyze advertising-supported Internet markets using clickstream data. In particular, it applies the methodology of Guadagni and Little (1983) to better understand households' Internet portal choices. The methodology has reasonable out of sample predictive power and can be used to simulate changes in company strategy.