Recommender Systems are smart systems that bring to the users a set of personalized suggestions from an specific type of items(objects). In order to do this, several techniques are used to collect the user’s characteristics for, using data processing, to find a subset of items that could be relevant to him. The improvement of the recommendation’s accuracy is crucial because offering relevant content (based on needs or likes) to the visitors of web sites, mainly commercial ones, is trending. This article shows a comparative analysis among different similarity and evaluation metrics proposed for collaborative-filtering based recommender systems; doing tests on commonly used datasets to determine its efficiency on production time.