To achieve a lean operation, fresh e-commerce enterprises need to grasp users’demands timely and accurately. This research aims to find consumers’heterogeneous preferences and implicit demands from reviews of fresh products on e-commerce platforms. Considering the social influence of implicit feedback information from reviews on customers’ purchasing decisions, this study designs a Mixed Logit model incorporating the impact of reviews on customers’ utility to simulate the different purchase behaviors and capture customers’ implicit requirements and heterogeneity effectively. Firstly, this research collects the online reviews of fresh products and then adopts natural language processing (NLP) technology to extract the customers’ implicit requirements from the reviews. Secondly, the implicit requirements are combined with customer attributes and fresh product attributes to construct the utility function and establish the customers’ choice behavior model. Finally, an example analysis is conducted based on the actual data from a fresh product e-commerce platform. The results show that the explicit attributes of fresh products cannot fully capture customers’ demands, and more invisible demands are hidden in other customers’ comments, which would affect the purchase decisions of potential customers. By incorporating the implicit feedback information in online reviews, the discrete choice model which considers the social impact of reviews is well interpreted and suitable for purchasing scenarios of fresh product e-commerce platform. This model shows a higher degree of goodness of fit compared with the model without considering the social influence of reviews. Thus, the implicit feedback in online reviews has been demonstrated to present more practical value for influencing customers’ choice behavior. The proposed method could provide valuable insights for fresh product e-commerce enterprises on precision marketing, business strategy adjustment, and product optimization.