Recommender systems are mechanisms that filter information in order to predict the preference of a user for an item drawn from a finite collection. Prime examples include recommendation platforms for movies, games, travel destinations, and books. Such systems rely on identifying users with similar preferences, as indicated by an appropriately selected metric, to yield reliable preference estimations. Recently, distributed implementations of recommender systems have improved scalability and can be applied by Internet-based companies to obtain insight for their respective products through latent patterns, typically of higher order nature, with having considerable impact on their profits. One of the potent personalization technologies currently powering the adaptive Web is collaborative filtering. The latter brings together the opinions of large interconnected communities on the Web, supporting filtering of substantial quantities of data. An implementation over Apache Spark of a typical recommender system is presented here. The evaluation of its prediction correctness, based on the Web movies dataset from Amazon as a benchmark, is in terms of the mean absolute error, the mean square error, and the root mean square error. Based on the results, suggestions for improved ratings are given, mainly by increasing user engagement and immersion.