A comparative study on measuring software functional size to support effort estimation in agile

Abstract

Software effort estimation models based on functional size allow software organizations to plan their development projects. A large number of organizations have adopted agile processes, but there is little evidence on the adoption of functional sizing methods to support software effort estimation in agile contexts. In this study, we compare four functional size estimation methods as the basis for effort estimation in the context of a startup company that develops mobile applications using an agile methodology. Measurements of software size, expressed in User Story Points (USP), Use Case Points (UCP), IFPUG Function Points (UFP), and COSMIC Function Points (CFP), were taken for a set of requirements from one project in the company. Effort estimation models were then derived from these measurements, using regression, and their accuracy was determined by the Mean Magnitude of Relative Error (MMRE) and Mean Balanced Relative Error (MBRE). We obtained the following MMRE results for each functional sizing method: 0,86 for UCP, 0,36 for USP, 0,36 for UFP and 0,22 for CFP, and the following MBRE results: 0,98 for UCP, 0,45 for USP, 0,53 for UFP and 0,35 for CFP. The effort estimation model based on COSMIC function points turned out to be the most accurate in the context of the software organization under study. Additionally, convertibility models between sizing measurements were generated to allow the organization to convert its historical measurements into any other software size measure, without having to perform the counting process of the target method.