This paper discusses how software projects following different processes, which display disparate characteristics, have been categorized and analyzed using statistical techniques to measure and improve process capability. The analysis shows that language type, size of the application and complexity (represented by ratio of Total Project Effort to Build Phase Effort) are the predominant discriminating factors affecting quality and process parameters. Categorization done based on this stratification scheme results in minimum variation within the group concerned. Such logical grouping also facilitates meaningful comparison of a project performance with almost similar historical data.