Measurement for improving accuracy of estimates: the case study of a small software organisation

Abstract This paper describes the case study of a small Canadian software development organisation that has established and sustained a measurement program for its software activities, which includes functional size measurement using the COSMIC-FFP method. This company has been in operation for over 20 years, and has 11 employees, all directly involved in software projects. Their “not to exceed” estimate business model, guarantees that fixing all defects found by their customer are free of charge. Quality is absolutely not an issue since they deliver a new software release to their major customer every other week with less than one defect per release on average, which is usually fixed within three hours. For that reason, they do not require a defect management system. Thus, the motivation for a measurement program came from other issues such as the inaccuracy of initial estimates, commitment to quality and productivity by applying best practices and continuous improvement, and the desire to improve productivity due to the loss of potential contracts to offshore organisations. Many challenges were encountered and resolved while implementing the measurement program such as dealing with company growth, tuning of estimation models to improve accuracy between initial estimates and project actual performance data, and applying the required rigour to sustain measurement activities. This paper also describes a simple measurement plan. Measurement results presented were used to improve the accuracy of estimation models, in which the step-by-step approach is described. With more accurate estimates, several sound business decisions were made regarding future projects.