Performance is an important property for a software system. Performance requirements should be specified and validated before detail design starts. However, some factors that can affect the system performance may not be available during early phases of software development. This dissertation proposes Performance Refinement and Evolution Model (PREM), an evolutionary model for performance requirements specification. A development team may use PREM to specify simple PRs early in the software development process, and add more details when the team gains more knowledge of the system performance. Two performance requirements improvement approaches were designed based based on PREM. The first approach, called Performance Requirement Improvement from Failure Reports (PRIFF), uses the field failure reported by the customers to improve performance requirements. PRIFF was applied on the requirements and field failure reports for a commercial distributed software system. The results demonstrate that the information in the field failure reports was integrated into the requirements that could be used for the next release. The resulting performance requirements are more complete and more specific than the original ones. The second approach, called DeNaP, improves the PRs with defect reports that are designated as not a problem (NaP). If a defect report is designated as NaP, the development team does not take any action on the defect report. A NaP defect report wastes the time of the development team and other key stakeholders since resources are spent on analyzing the problem but, in the end, the quality of the software is not improved. Reducing the NaP occurrence rate improves the efficiency of the development team. DeNaP was applied on a firmware development project of an embedded control module from ABB Inc and a file processing system from EMC Corporation. After applying DeNaP, we were able to create new performance requirements and refine the original ones from the NaP defect reports. A survey was conducted to examine the development teams’ reaction to the resulting performance requirements. The results show that more than half of the defect reports could have been avoided given the resulting performance requirements from DeNaP.