In order to deeply explor the characteristics of power customers and accurately grasp the customers’service needs, refined marketing strategies for power customers are studied and formulated to improve customers’satisfaction. By combing the precision marketing theories, big data mining technology is adopted to analyze the basic behavior characteristics of customers, and a customer attribute index system that meets customer characteristics and marketing strategies is built. With the multi-attribute matching algorithm of tensor decomposition, the marketing strategy matching model of multi-attribute power customers based on tensor decomposition is constructed. An example analysis is carried out with 1 096 commercial customers of a power grid company. The results have important reference value for the power grid companies to carry out refined management work and differentiated services.