Early Web Size Measures and Effort Prediction for Web Costimation

Size measures for Web costimation proposed in the literature are invariably related to implemented Web applications. Even when targeted at measuring functionality based on function point analysis, researchers only considered the final Web application, rather than requirements documentation generated using any existing Web development methods. This makes their usefulness as early effort predictors questionable. In addition, it is believed that company-specific data provide a better basis for accurate estimates. Many software engineering researchers have compared the accuracy of company-specific data with multiorganisation databases. However the datasets employed were comprised of data from conventional applications. To date no similar comparison has been adopted for Web project datasets. As a result, this paper has two objectives: The first is to present a survey where early size measures for Web costimation were identified using data collected from 133 Web companies worldwide. All companies included in the survey used Web forms to give quotes on Web development projects, based on gathered size measures. The second is to compare the prediction accuracy of a Web company-specific data with data from a multiorganisation database. Both datasets were obtained via Web forms, used as part of a research project called Tukutuku. Our results show that best predictions were obtained for company-specific dataset, for the two estimation techniques employed.