Web developers have been concerned with the issues of Web latency and Web data consistency for many years. These issues have become more important in our days since the accurate and imminent dissemination of information is vital to businesses and individuals that rely on the Web. In this paper, we evaluate different run-time management policies against real Web site data. We first define the meaning of data intensive Web sites and categorize them according to their hit patterns. Our research relies on real world Web data collected from various popular Web sites and proxy log files. We propose a Web site run-time management policy that may apply to various real Web site hit patterns and Web data update frequencies.