BrowseRank starts with processing the user behavior logs and uses a new data structure named user browsing graph to represent the data. Since there is staying time information contained in the new graph, one can no longer employ the discrete time markov chain used by PageRank for the modeling. We instead use a continuous time markov process for this purpose. We use the stationary distribution of this stochastic process as the page importance.