University of Twente Student Theses


The red queen’s race: from delivery stories to effort estimation: an empirical study

Vermolen, R.M. (2012) The red queen’s race: from delivery stories to effort estimation: an empirical study.

[img] PDF
Abstract:Accurate software effort estimation is a key factor for software development project success. Both the need for and the complexity of correct estimates have increased due to the growing trend for complex software and distributed projects. The introduction of agile software development has added another challenge to accurate effort estimation, this is partly due to the way requirements are engineered in agile projects. This master thesis is concerned with the state-of-the-practice of effort estimation in a large-scale, outsourced and agile project, in order to understand current effort estimation practices and identify possible causes for inaccuracies in those estimates. The first part of the thesis presents a literature review on large, distributed and agile software development projects. It reports on the specific challenges related to: (i) agile projects, scaling those projects and dealing with distribution of such projects; and (ii) agile RE in large-scale environments and effort estimation in distributed and agile projects. The state-of-the-practice in the case study project was analysed using the knowledge that was gathered during semi-structured interviews with software professionals. The vendor has organized this agile project based on a planning phase, which included a pilot scrum. The experiences from this planning phase helped to plan the different releases. A remarkable aspect of the application of agile in this project, is the use of the delivery story (DS) artefact. The DS is a design document which incorporates all information necessary for development of a particular piece of functionality. As a consequence, delivery stories also play a pivotal role in effort estimation. All participants in this study are satisfied with the current usage of the DS. The key findings from this research effort include effort estimation as a recurring task that appears in four different levels of the project. Effort estimation takes place on: project level, DS specification level, DS development level and functional testing level. Estimations usually go through 3 phases, starting with a ballpark estimation. The ballpark estimation evolves into a baseline estimation and the baseline is updated once more detailed estimations are available. Effort estimation seems to work well as it is happening right now. All different occurrences of effort estimation could have added up to a big deviation in total project effort, but our findings indicate this is not the case. Many review- and approval-cycles are combined with a vast amount of software development experience, which leads to fairly good results when it comes to effort estimation. It can be concluded that in this project there is no urge to improve effort estimation, the current practices apparently work well enough. This research makes two types of contributions: from a theoretical perspective, it contributes to the emerging research on effort estimation in agile, distributed, large-scale projects. In large, agile projects, delivery stories support effort estimation practices and reduce the need for communication. From a practical perspective, this research offers practitioners knowledge about the use of agile artefacts in effort estimation and the value of learning-oriented estimation.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Computer Science MSc (60300)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page