Date of Original Version



Working Paper

Rights Management

All Rights Reserved

Abstract or Description

Are there any trends in web browsing behavior by individual users? On the popularity of web sites? On the frequency of revisitation? These are questions that we address in this work. Our study uses an extensive clickstream data from U.S. households from July 1997 to December 1999. It is unique, for several reasons: (1) it uses an extensive clickstream data from U.S. households for a long interval (30 months), and (2) it uses web viewing patterns captured from the browser window, as opposed to relying upon hits recorded in server logs, and (3) these users comprise a representative sample of U.S. web users. We present some surprising empirical regularities that seem to hold consistently for the time-span of our dataset. Specifically, we found that (a) the growth of the internet usage seems exponential over time, with a growth rate of about 2.4% per month (b) the growth is mainly due to the new users, while the per-user behavior seems to remain stable over time (c) the popularity of web sites (visits per month per user) seems to follow a power law, with slope about -1.2 (d) the ‘stickiness’, that is, revisitation patterns (inter-arrival times) follows another power law, with slope -.6.