Predicting and Analyzing Feature Value when R&D is an Experiment System
Examensarbete för masterexamen
Software engineering and technology (MPSOF), MSc
Þorvaldsson, Ivar Daði
Software development has experienced major changes in the last decades, from traditional waterfall development to the agile way of working. In the last years, companies have been moving beyond agile practices, towards both continuous integration (CI) and continuous deployment (CD), before advancing to continuous experimentation. When continuously experimenting, R&D can be viewed as an experiment system, with the aim of rapidly testing and validating software based on customer and user feedback. However, it has proven problematic to gather the relevant data from customers, resulting in an ‘open loop’ between customers and product management. As a consequence, decisions are often based on ‘gut feeling’ rather than actual data. Several process models have been created to analyze feature value throughout the software development process. Nevertheless, no concrete procedure for predicting and analyzing feature value has been developed. This study presents the DVOCE model to fill that gap. DVOCE is a detailed and extended version of the previously published highlevel HYPEX model and covers the pre-development and development phases. DVOCE provides a detailed procedure in how to model the feature to enable the prediction. In addition, it includes a sub-process for selecting the appropriate customer feedback and data collection techniques so the value can be tracked before analyzing whether a feature lives up to its expectations. In the study, design science research (DSR) is used as a research methodology. To help with validating the process model, a prototype was created based on the core aspects of the model. The process model was then validated in eight validation sessions at four companies, with a total of ten participants. The results suggest that the DVOCE process model can be used to model, predict and analyze feature value, as well as to track the realized feature value using the appropriate customer feedback and data collection techniques. Further work is needed to validate the model with a larger audience.
Informations- och kommunikationsteknik , Data- och informationsvetenskap , Information & Communication Technology , Computer and Information Science