In today's grocery the products are getting more homogenous and attributes are getting more complex. Consumers should be supported in their decisions during the shopping process. The main objective of this paper is to look at the conditions needed to transfer the concepts of online recommender systems to conventional brick-and-mortar grocery stores. RFID and plan recognition is considered to observe and forecast the consumer's behavior. The basic architecture consists thereby of a RFID infrastructure, a personal shopping assistant and a back-end system integrated into the retail information system. Based on the proposed architecture the second objective is to build a concept explaining how to judge the economic feasibility. Therefore, a framework to rate single applications and the overall system will be developed. The sum of the implemented applications needs to satisfy both, the seller and the consumer through net benefits. The variables determining these net benefits are identified and explained