Using Clickstream Data to Predict WWW Usage

The purpose of this study is to consider the determinants of World Wide Web (WWW) usage using clickstream data. Clickstream data is a natural byproduct of a user accessing WWW pages, and refers to the sequence of pages visited and the time these pages were viewed. A key component of this study is a large scale empirical analysis of clickstream data from a representative sample of PC owning households in the U.S. This sample comes from Media Metrix's PC Meter using a nationwide panel of over 5,000 households during 1997 and 1998. A primary finding of this research is that demographics are not a primary driver of web usage. While it is true that access to the WWW is influenced by demographic characteristics, the best way to predict current usage is to use information about past usage, and not use demographics. Additionally, we show there is a great deal of diversity in the usage of individual sites. These findings imply that if managers really want to learn how to fulfill the promise of interactivity on the WWW they need to learn how to capture and use their customers' clickstreams.