Identification of Inaccurate Effort Estimates in Agile Software Development

Abstract

Like in every process model, agile processes (e.g. Scrum, eXtreme Programming) depend on accurate estimations to enable meaningful prioritization, iteration- and release planning. The emphasis of this paper is on "Planning Poker", a widely used estimation technique in agile context. The goal is to identify inaccurate effort estimates to enable more precise project scheduling and release planning. In a first step, basic terms (e.g. development effort, functional effort) of traditional effort estimation were declared in the agile context. Afterwards-within the main part of the paper-a set of metrics is proposed, which can be used to evaluate accuracy of estimates. These metrics are based on the estimated efforts and interim results of the estimation process. Since the usage of these metrics needs to be seamlessly integrated into the Planning Poker process, we have conceptualized a computer-aided tool to collect and evaluate necessary data. To get a first proof of concept and preliminary feedback, we developed a corresponding prototype and applied it to a students' project. The prototype is based on a stationary multi-touch device and offers an intuitive user interface to perform the Planning Poker process. While estimation takes place, it collects required data and identifies inaccurate estimates with the help of the proposed metrics. The second part of the paper summarizes assets and drawbacks concerning usability and the user interface of the prototype.