An Empirical Analysis of eXtreme Programming Practices and its Impact on Software Quality Metrics Roberto Pepato Mellado 1 , Gabriel de Souza Pereira Moreira 1 , Robson Luiz Monteiro Júnior 1 , Adilson Marques da Cunha 1 , Luiz Alberto Vieira Dias 1 1 Brazilian Aeronautics Institute of Technology (ITA) – São José dos Campos – SP - Brazil {rpepato,gspmoreira}@gmail.com, robson@kimmay.com.br,{cunha,vdias}@ita.br Abstract: This work presents an investigation of three different industrial projects of software development by a Brazilian enterprise. During projects’ execution, the company has changed its approach on software processes from RUP based process to agile like processes. To assess software product quality metrics evolution, an investigation of product metrics history was conducted in those three projects. This paper characterizes the use of eXtreme Programming practices within the analyzed projects and the observed measures of quality metrics in the developed software products. Keywords: Agile software development, extreme programming, object- oriented software, software quality metrics 1. Introduction On the last decade, a growing interest on agile methods for software construction from both industry and academy [Dyba et al. 2008] [Hanssen et al. 2010] has arisen. Agile methodologies rely on a set of best practices that are considered to increase quality assurance and control. These procedures are integrated across the entire life-cycle development. Agile methods state that quality should be built into the product through a combination of best practices. Studies support advantages of agile practices towards product quality improvement [Sfetsos 2010]. Extreme Programming (XP) [Beck 1999], one of established agile methods, has defined sets of values, principles, and practices. Quality is one of the principles that must be followed and tracked when applying XP. Beck [Beck 1999] states that pushing quality higher often results in faster delivery, while lowering quality standards results in later and less predictable delivery. One approach to monitor software quality over time is the usage of metrics. Some studies have proposed sets of metrics to measure systems designed and constructed by Object Oriented (OO) principles [Chidamber et al. 1994] [Martin 1994]. Lehman [Lehman et al. 1985] defines the law of increasing entropy. He explains that, by a number of factors, a programming system undergoes never-ending maintenance. As a result of these changes, the system becomes more complex and unmanageable, requiring specific work to maintain or reduce the increase of its entropy.