With the development of multimedia, the form of online advertising has become more diversified. In the increasingly competitive environment between e-commerce enterprises, the proportion of online advertising investment in total advertising investment continues to grow. However, the operating costs of enterprises are limited, in order to maximize the effectiveness of operating costs, evaluating the effect of various forms of online advertising effectively to determine the scientific online advertising budget allocation is crucial. Based on the enterprise’s Web log data, we extract the characteristic variables that represent consumer's behavior including click behavior, visit behavior and purchase behavior. On the one hand, using multivariate linear regression method to assess the short-term effect of three forms of online advertising in an overall level. On the other hand, we use logistic regression models to investigate the impact of various consumer behaviors on the purchase conversion of consumers who enter the website from different advertising channels.