Recommendation systems have gained popularity in recent years. Among them, the best known are those that select products in stores, movies, videos, music, books, among others. The companies, and in particular, the banking entities are the most interested in implementing these types of techniques to maximize the purchases of potential clients. For this, it is necessary to process a large amount of historical data of the users and convert them into useful information that allows predicting the products of interest for the user and the company. In this article, we analyze two essential problems when using systems, one of which is to suggest products of one commerce to those who have never visited that place, and the second is how to prioritize the order in which users buy certain products or services. To confront these drawbacks, we propose a process that combines two models: latent factor and demographic similarity. To test our proposal, we have used a database with approximately 65 million banking transactions. We validate our methodology, achieving an increase in the average consumption in the selected sample.