To main content

Agile Retrospectives: An Empirical Study of Characteristics and Organizational Learning


Agile retrospectives could help teams learn from the past and identify improvement opportunities. In this thesis we investigate the characteristics of the present retrospective practice and the organizational learning it provides for teams practicing them, through a multiple case-study.
The research is based on a depth-study of five years of retrospective prac- tice from one team, and a breadth-study of seven interviews from other ret- rospective practicing teams.

Our results show that teams today are mostly satisfied with their retro- spective practice and are able to identify improvement opportunities and im- plement them, which contradicts previous research. However we identify one barrier related to team commitment dependent on enthusiasm, and previous implementation of improvement opportunities that results in a feedback-loop that could both help implement future improvements and hinder them. We investigate the learning happening through today’s retrospectives and find that the practice is approximating a learning system where teams are able to test their current work practices, learn from them and improve from them. Where most of the governing values and behavioral consequences for such a system are already present in today’s practice. However we find that the primary learning type remains single-loop, even though double-loop should be expected. We identify a barrier that hinders double-loop learning, consisting of several factors and propose a method based on our findings. The method aims to facilitate triple-loop learning and adaption of the retrospective prac- tice, which can lower the learning barrier.

Finally we conclude that today’s retrospective practices provide agile development teams the ability to adapt their current work-practices and enable them to learn from past development iterations and thus provide the means for identifying improvement opportunities and improve from them.


Masters thesis


  • Research Council of Norway (RCN) / 235359
  • Research Council of Norway (RCN) / 236759




  • Torgeir Dingsøyr
  • Nils Brede Moe
  • Alf Magnus Stålesen
  • Bjørn Dølvik


  • Norwegian University of Science and Technology
  • SINTEF Digital / Software Engineering, Safety and Security
  • Unknown




Institutt for datateknikk og informasjonsvitenskap, NTNU

View this publication at Cristin