Business to Business (B2B) marketing aims at meeting the needs of other businesses instead of individual consumers. In B2B markets, the buying processes usually involve series of different marketing campaigns providing necessary information to multiple decision makers with different interests and motivations. The dynamic and complex nature of these processes imposes significant challenges to analyze the process logs for improving the B2B marketing practice. Indeed, most of the existing studies only focus on the individual consumers in the markets, such as movie/product recommender systems. In this paper, we exploit the temporal behavior patterns in the buying processes of the business customers and develop a B2B marketing campaign recommender system. Specifically, we first propose the temporal graph as the temporal knowledge representation of the buying process of each business customer. The key idea is to extract and integrate the campaign order preferences of the customer using the temporal graph. We then develop the low-rank graph reconstruction framework to identify the common graph patterns and predict the missing edges in the temporal graphs. We show that the prediction of the missing edges is effective to recommend the marketing campaigns to the business customers during their buying processes. Moreover, we also exploit the community relationships of the business customers to improve the performances of the graph edge predictions and the marketing campaign recommendations. Finally, we have performed extensive empirical studies on real-world B2B marketing data sets and the results show that the proposed method can effectively improve the quality of the campaign recommendations for challenging B2B marketing tasks.