Nowadays, retailers use mutliple on- and offsite channels to sell their products. This approach is named multi-channel, and the typical channels are social media, e-commerce, e-mail, stationary. On the one side, using different channels promises higher sales, but also the chance to gain more valuable data about the customers. On the other side, the required infrastructure to run a multi-channel business is enormously, and the usage of the data is often unclear. To overcome both challenges, the recommender system TaxoRecommend is presented in this article. The system is based on widely used techniques: the Structured Query Language (SQL), as well as the database and metadata techniques provided by Microsoft Dynamics CRM. This allows the system to aggregate customer knowledge across multiple channels and predict preferences on differently scalable object and audience levels.