Recommender systems have been widely adopted by electronic commerce and entertainment industries for individualized prediction and recommendation, which benefit consumers and improve business intelligence. In this talk, we propose an innovative method for tensor recommender systems, which utilizes the structure of a tensor response to integrate information from multiple modes, and creates an additional layer of nested latent factors to accommodate between-subjects dependency. One major advantage is that the proposed method is able to address the "cold-start" issue in the absence of information from new customers, products or contexts. Specifically, it provides more effective recommendations through subgroup information. To achieve scalable computation, we develop a new algorithm, which incorporates maximum block improvement into the cyclic blockwise-coordinate-descent algorithm. In theory, we investigate algorithmic properties, along with the asymptotic consistency of estimated parameters. The proposed method is applied to IRI marketing data with 116 million observations of product sales. Numerical studies demonstrate the superior performance.