Defect Data Analysis as Input for Software Process Improvement

摘要

In this paper, we present the results of defect data analysis done with three software companies' defect databases. 11879 software defects were classified and analyzed in order to find out what the real world defect distributions are like and what are the most common defect types. The most common defects in every company were functional defects (65.5%), i.e. defects in computation and/or functional logic. The defect types that were most uncommon were defects due to misunderstood or poorly written requirements (0.2%) or documentation (0.4%). The results of the analysis offer practical data to be used to support Software Process Improvement (SPI).