After building a classifier with modern tools of machine learning we typically have a black box at\nhand that is able to predict well for unseen data. Thus, we get an answer to the question what is the\nmost likely label of a given unseen data point. However, most methods will provide no answer why\nthe model predicted a particular label for a single instance and what features were most influential\nfor that particular instance. The only method that is currently able to provide such explanations are\ndecision trees. This paper proposes a procedure which (based on a set of assumptions) allows to\nexplain the decisions of any classification method.