In this paper, different facial alignment techniques are revised in terms of their effects on machine learning algorithms. This paper, investigates techniques that are widely accepted in literature and measures their effect on gender classification task. There is no special reason on selecting gender classification task, any other task could have been chosen. In audience measurement systems, many important demographics, i.e. gender, age, facial expression, can be measured by using machine learning algorithms. Moreover; in such a system, these demographics are so substantial since their discriminative features on customers' habits. Due to this importance, any performance enhancement on any machine learning algorithm becomes important. A carefully chosen alignment method can boost performance of machine learning algorithms used within system.