In an age of explosive information growth, information has the attributes of timeliness, sharability, and relativity of value. Personalized recommendations can help resolve the difficulty in making choices for users who are confronted with information overload. The information aging problem is to be expected during the development of information networks, and incorporating the time factor into recommendation algorithms is becoming a trend. Most studies on recommendation algorithms considering the time factor have either focused on changes in users’ interests or the current popularity of items. However, the timevarying comprehensive popularity of items (ComPI) has attracted little attention. In this study, a realtime recommendation algorithm based on ComPI was proposed. The main contributions of the present study were as follows: (1) the development stage of item in product life cycle was considered in the recommendation; (2) the time factor was incorporated into the recommendation algorithm, and an exponential timeweight decay model was built for the quantification of time weights; (3) the total probability model was combined with the timeweight decay model to establish the ComPIbased recommendation algorithm, which effectively solved the problem of “marketing myopia” in current algorithms; (4) the algorithm was tested on real data sets. The results showed that our realtime recommendation algorithm had better accuracy, diversity, and novelty; (5) we also researched the growth characteristics of data and its impact on data partitioning and algorithm performance evaluation, and proposed that we should pay more attention to the data with exponential growth when studying recommendation algorithms. In brief, this algorithm can effectively improve the user experience by using the time factor in recommendation systems. In addition, the ComPI model can also help to study the evolution process of network topology, and can be applied to network related fields.