Recommender system, which helps consumers to filter through large information and enhance consumers’ shopping experience and loyalty, is a critical application of big data analytics in retail industry. Many recommender systems are proposed and applied to real-word applications. However, existing recommender systems mainly concentrate on predicting the preference scores of unknown/unpurchased items to target users but ignore the consideration of consumers’ repeated consumption behaviors. Such assumption is not reasonable in some applications, such as department stores or grocery stores where repeated consumptions are common. Accordingly, this study concentrates on addressing this limitation by incorporating item repurchased probabilities into recommendation generation. The proposed repurchase-based collaborative filtering (RP-CF) technique has the ability to recommend both unpurchased and repurchased items. We collect a real-world dataset from a major chain department store in Taiwan for empirical evaluation. The findings are promising and shown that the RP-CF technique indeed improves the performance of personalized recommendation.