Effort estimation in agile software development: Case study and improvement framework

Abstract

Effort estimation is more challenging in an agile context, as instead of exerting strict control over changes in requirements, dynamism is embraced. Current practice relies on expert judgment, where the accuracy of estimates is sensitive to the expertise of practitioners and prone to bias. To improve the effectiveness of the effort estimation process, the goal of this research is to investigate and understand the estimation process with respect to its accuracy in the context of agile software development from the perspective of agile development teams. Using case study research, 2 observations and eleven interviews were conducted with 3 agile development teams at SAP SE, a German multinational software corporation. The results reveal that factors such as the developer's knowledge, experience, and the complexity and impact of changes on the underlying system affect the magnitude as well as estimation accuracy. Furthermore, there is a need for a tool that incorporates expert knowledge, enables explicit consideration of cost drivers by experts and visualizes this information to improve the effectiveness of the effort estimation. On the basis of the findings of the case study, a framework, inspired by the quality improvement paradigm is proposed to improve effort estimation in agile development

Investigating the use of duration‐based windows and estimation by analogy for COCOMO

“…Reliable software effort, cost, and time estimates are essential inputs for making decisions about investment, staffing, scheduling, and other planning and controlling of software projects . Thus, the software engineering research has introduced and investigated many approaches to software estimation, which is evidenced via systematic literature reviews, surveys, and model assessment studies (eg,). Model‐based estimation approaches like COCOMO have been widely investigated in the software engineering literature and continue to be used in practice with commercial models such as SEER‐SIM, Price‐S, and SLIM despite the increasing application of agile methods in software development.…”

An evaluation of effort estimation supported by change impact analysis in agile software development

“…The selection and integration of impact analysis and effort estimation techniques were done by using our proposed framework and guidelines to instantiate it. The results of the state‐of‐the‐art analysis of effort estimation in agile development, indicate the expert‐based estimation methods are most often used, with “Planning Poker” being the most frequently employed method .…”

“…The previous preliminary assessment at SAP SE and this evaluation at Insiders Technologies GmBH indicated positive results for HyEEASe to improve the accuracy of effort estimation. Therefore, we wanted to introduce the method to software engineering students along with Planning Poker (the most frequently used method in the industry).…”

“…HyEEASe was developed incrementally. Initially, a conceptual framework was developed in close collaboration with SAPcompany. The framework was instantiated, and the potential tool support was materialized using mock‐ups with SAP.…”

“…7,8 Change impact analysis (IA) techniques identify software life cycle objects that are likely to be affected by a given change request. 9In our previous work, 8,10,11 to support expert-based estimation methods with change impact analysis, we have synthesized research from two areas: (1) change impact analysis 9 and (2) effort estimation (both expert and model based). We used case study and survey research to understand the needs of industry practitioners and their challenges and development context.…”