We present the results of an initial analysis conducted on a real-life setting to quantify the effect of shilling attacks on recommender systems. We focus on both algorithm performance as well as the types of users who are most affected by these attacks.