A personalized recommendation system is an effective and consistent marketing method that caters to consumer demand for goods. It solves the problem of consumer shopping choices on e-commerce websites and is currently a hot spot in the field of network information services. In this paper, from the perspective of the user’s shopping process, we summarize the four factors that can maximally reflect the consumer’s interest and preference, i.e., product browsing, collection, adding to cart, and purchasing behavior. To achieve static user interest weights, we quantify each index factor and set the corresponding rules. Considering the change in user interest, the decaying interest value over time is designed to compensate for the deficiencies of static system recommendations. We use implicit user feedback data to identify content that truly interests users and then construct a personalized recommendation algorithm for social media marketing content using multi-source data. The test results show that UIPR has an MRR value of more than 0.8 and an NDCG value of more than 0.55 in both the MovieLens and Taobao datasets, compared to the baseline model’s best value. This proves that UIPR consistently performs better and makes better recommendations.