Team Effectiveness in Software Development Human and Cooperative Aspects in Team Effectiveness Models and Priorities for Future Studies
Torgeir Dingsøyr
Tore Dybå
SINTEF NO-7465 Trondheim, Norway
[email protected]
SINTEF NO-7465 Trondheim, Norway
[email protected]
Members of such self-managed teams are usually responsible for managing and monitoring their own processes and executing tasks. Typically, they also share decision authority, leadership tasks, and responsibilities [4]. Although the majority of studies on self-managing teams report that using such teams has positive effects, some studies offer a more mixed assessment. In addition, research on team performance indicates that the effects of autonomous work groups are highly situational dependent. Further, autonomy on the individual level may conflict with autonomy on the group level. Barker [5], for example, pointed out that self-managing groups may end up controlling group members more rigidly than they do under traditional management styles, while Markham and Markham [6] suggested that it may be difficult to incorporate both individual autonomy and group autonomy in the same work group. Given these characteristics of software development, and focusing on human and cooperative aspects in software development teams, we would like to discuss the use of team effectiveness models for better theoretical understanding and priorities for future studies of teamwork. The question is: Do we need our own team effectiveness models in software development? In the remainder of this article, we first describe three models of team effectiveness, which have been developed in other research fields. These are models that we either have used or are planning to use in future studies of software development teams. We describe pros and cons of these models, and finally discuss what we see as priorities for future studies of team effectiveness in software development.
Abstract—Software development is most often done in teams, where human and cooperative aspects are vital for team effectiveness. This has been the topic of study in several disciplines, and in this article we describe three team effectiveness models from other fields. We discuss priorities for future studies on software teams, and ask: Do we need our own effectiveness model for software teams? Keywords-Team effectiveness models; software development; agile development; empirical studies; sociotechnical system
I. INTRODUCTION Software development can be seen as a sociotechnical system comprised of human and technical entities that, when functioning as an integrated, coordinated unit, can address a wide range of problems that are too complex to be addressed by individuals or technologies working alone. However, the design and implementation of complex software systems tends to place primary emphasis on technological innovation without equal consideration for the social component – the teams and groups of software developers – that uses that technology to create the software systems. Three key issues characterize complex sociotechnical systems. First, such systems adapt to environmental uncertainty through self-managed (viz. “self-regulated”; [1]) processes without a central executive or blueprint guiding that adaptation. Complex systems that are robust to environmental uncertainty, therefore, are adaptive by maintaining a level of autonomy in how they adapt to uncertainty. Second, technological interventions do not increase sociotechnical system effectiveness if they are not supported by the social (team and group) components of the system. Early sociotechnical research, for example, revealed that isolated “tinkering” with either the technical or social components of the system may not result in the desired improvement of overall system effectiveness [2]. Finally, a third key issue in complex sociotechnical settings is that adaptation emerges from coordinated and collaborative interactions at a team or group level. Such team interactions are of vital importance in software development, and the recent focus on agile software development [3] has, therefore, led to an increased interest in how to effectively organize small self-managing teams.
c 2012 IEEE 978-1-4673-1824-2/12/$31.00
II. MODELS OF TEAM EFFECTIVENESS There are a number of models of team effectiveness, originating from several disciplines of science [7]. In particular, teamwork has been a much researched topic in management science and in psychology. In the following, we present three models of teamwork that we have used or will be using in future studies on software teams. The two first are taken from psychology, while the last one is taken from management science. These models all focus on team effectiveness, and mainly on internal aspects of the team. There are other models that take into account characteristics of individual team members and the external environment.
27
CHASE 2012, Zurich, Switzerland
2
14
N.B. Moe and T. Dingsøyr
The Core
Team Leadership
Mutual Performance Monitoring
Team Orientation
Back-up Behavior Share Mental Models
impact work process work process on the collaborative on the collaborative model focuses model focuses (i.e., (i.e., More r Hauptm cross-fu the Success of Innovative of Innovative and the Success 1 Teamwork Teamwork Quality and Quality Figure 1 Figure way co Projects Projects ence on Performance Performance Team Team pirical ?* Effectiveness ?* Effectiveness (Quality) (Quality) Teamwork Quality Teamwork Quality * * and(Schedule +'.^ +'.^ manag Budget) and Budget) Efficiency (Schedule Efficiency * Communication * Communication * Coordination * Coordination to be a * Balance of Member * Balance of MemberContributions Contributions and Le * MutualSupport * MutualSupport * Effort * Effort * Cohesion * Cohesion
+
-
Personal Success Personal Success * +Work * WorkSatisfaction - Satisfaction * Learning(Knowledge * Learningand Skills) and Skills) (Knowledge
Figure 3: Team effectiveness according to Hoegl [12].
Mutual Trust
Coor ent fun
!
2001 2001 No. characteristics No. 4, July-August ORGANIZATION ORGANIZATION SCIENCE/Vol. 4,12,July-August 12,other back andSCIENCE/Vol. coordination are than what is Adaptability used in the Salas model. In management science, Hoegl [12] identified the model described in Figure 3 based on a literature review, which we are currently using for a survey on teamwork in agile Team development. Teamwork quality is described as consisting of Effectiveness communication, coordination, balance of member contributions, mutual support, effort and cohesion. The three models were developed for different purposes. Closed loop communication The Salas model was developed for the US Army to summarize team research in a practical model inspired by the Fig.1:1.Team The proposed modelasofseen “BigbyFive” "big five" personality factors in psychology. The Dickinson Figure effectiveness Salasinetteamwork al. [9]. by Salas et al. [16] and McIntyre model was similarly directed at practical use, In the case, because of highly specialized skills and a corresponding division of but focused on self-managing teams. On the other hand, the In a study which examined how the agile method Scrum work, it was a problem to develop shared mental models both on a project level and Hoegl model was developed for a survey study on the effect facilitates teamwork [8], Moe and Dingsøyr used thedeveloped, model planon a task level. Without a clear understanding of the system being of teamwork quality on team performance in innovative by Salas et al. [9] (Figure 1). This model defines three ning was difficult. The planning meetings turned out to be discussions between the projects. coordinating mechanisms are said exist in toalldoeffective Scrum-master, product owner and thethat developer thattowas going the work. Although the models use different terminology, there are teamwork: Shared mental models, mutual trust and closedDefinition [16]: “An organizing knowledge structure of the relationships among a number of similarities in the mechanisms included. loopthe communication. Given these, five other the task team is engaged in and how the team members will factors interact”. are hypothesized to “Anticipating influence and team effectiveness: TeamIdentifyCommunication is central in all models; team orientation in Behavioural Makers [16]: predicting each other’s needs. leadership, mutual performance monitoring, changes in the team, task, or team-mates and implicitly adjusting strategies team as needed”.the first two models is related to effort (defined as "Do team Scrum: The team isback-up supposedbehavior to focus onand the adaptability. high level goal setting in the up frontmembers exert all efforts to the team's tasks?"). Back-up orientation, planning.AThe productofowner shouldmodel provideisa that visionit to help this process. strength the Salas originates from a Every(behavior) can be found again in mutual support. Hoegl puts stakeholder is involvedreview. in this planning which is as a co-located solid literature A weakness is conducted that the model is not meet-more emphasis on the importance of motivation in work ing. The sprint and product backlogs describe what is to be developed. effectiveness through the focus on team cohesion. In the particularly tailored to self-managing teams, which is Case-study: The product owner was responsible for communicating the features other models motivation is embedded in team leadership, expressed as an important goal for agile development teams. this system was going to provide. But he lived in another city, was very busy, and In a further study of the same case by Moe et al. [10] we became sick in a critical phase. There was a lot of communication between thewhich involves "motivating team members" in Salas, and explained theproduct findings withbutanother model, is tailored developers and the owner according to the which developers it should have"establishing a positive atmosphere". The Salas model is the selftomanaging teams: The Dickinson and McIntyre beento more, help them better understand how the system was going model to be used. Aonly one to explicitly focus on trust and shared mental proper vision also in missing. models as requirements for effective teamwork. Learning [11], as was shown Figure 2. and feedback is a characteristic of the Dickinson and This model uses many of the same mechanisms as Salas McIntyre model. et al: Team leadership, team orientation, back-up behavior and communication. Monitoring is similar to what Salas et III. PRIORITIES FOR FUTURE STUDIES OF SOFTWARE TEAMS et al. / Information and Software Technology 52 (2010) 480–491 al. describe N.B. as Moe mutual performance monitoring, while feedAlthough there are several models of team effectiveness originating from other disciplines, there are many challenges and opportunities for research on software teams, teamwork, and team effectiveness. Building on Salas et al. [13], we see five issues that are fundamental to further our understanding of software team effectiveness, which should be prioritized in future studies.
Fig. 1. The Dickinson and McIntyre teamwork model [13].
Figure 2: Team effectiveness according to Dickinson and McIntyre [11]. elements that are found in software teams, such as coordination of work [25]. It specifies what teamwork skills should be observed, in that the model is presented with a conceptual framework for developing measures of teamwork performance that can ensure effective individual and team performance [13], pp. 22. It considers the teamwork process as a learning loop in which teams are characterized as adaptable and dynamically changing over time. Continuous self-management requires a capacity for double-loop learning that allows operating norms and rules to change along with transformation in the wider environment
A. Better Measurement Although there have been advances in the measurement of team behavior, teamwork quality and team cognition in other disciplines, we still need more robust and valid diagnostic measurement approaches of teamwork and team to make decisions,
team members should share the authority rather than having: (a) a centralized decision structure in which one person (e.g. the team leader) makes all the decisions or (b) a decentralized decision structure in which all team members make decisions regarding their work individually and independently of other team members [20]. So, while the traditional perspective of a single leader suggests that the leadership function is a specialized role that cannot be shared without jeopardizing group effectiveness, when leadership is shared, group effectiveness is achieved by empowering the members of the team to share the tasks and responsibilities of leadership [22].
28
!
performance in software development that can be practically implemented in the field.
software development teams should connect better to this pool of research. With respect to empirical studies of software teams, there is a need to test, extend and update theories from other fields. With the current state-of-the-art in studies of software teams, we cannot say whether we need our own models of team effectiveness in software research, but we need to increase the understanding of the genuine characteristics of software development to know what applies.
B. More Rigorous Industrial Case Studies Software teams are embedded in organizations and in broader sociotechnical systems. The connections between the team and the other parts of the system definitely affect teamwork and team effectiveness; however, there are few rigorous studies of industrial teams in their fully situated context. Such case studies can enable researchers and practitioners to provide higher-quality context-specific guidance that complement existing theoretical models.
IV. REFERENCES [1] T. G. Cummings, “Self-Regulating Work Groups: A SocioTechnical Synthesis,” The Academy of Management Review, vol. 3, no. 3, p. 625, Jul. 1978. [2] E. L. Trist and B. K.W., “Some Social and Psychological Consequences of the Longwall Method of Coal-Getting,” Human Relations, vol. 4, no. 1, pp. 3-38, Feb. 1951. [3] T. Dybå and T. Dingsøyr, “Empirical Studies of Agile Software Development: A Systematic Review,” Information and Software Technology, vol. 50, pp. 833-859, 2008. [4] N. B. Moe, T. Dingsøyr, and T. Dybå, “Overcoming Barriers to Self-Management in Software Teams,” IEEE Software, vol. 26, no. 6, pp. 20-26, 2009. [5] J. R. Barker, “Tightening the iron cage – concertive control in self-managing teams,” Administrative Science Quarterly, vol. 38, no. 3, pp. 408–437, 1993. [6] S. E. Markham and I. S. Markham, “Self-management and self-leadership reexamined: a levels-of-analysis perspective,” The Leadership Quarterly, vol. 6, no. 3, pp. 343-359, 1995. [7] E. Salas, K. C. Stagl, C. S. Burke, and G. F. Goodwin, “Fostering team effectiveness in organizations: toward an integrative theoretical framework.,” Nebraska Symposium on Motivation. Nebraska Symposium on Motivation, vol. 52, pp. 185-243, Jan. 2007. [8] N. B. Moe and T. Dingsøyr, “Scrum and Team Effectiveness: Theory and Practice,” in XP2008, 2008, vol. 9, pp. 11-20. [9] E. Salas, “Is there a ‘Big Five’ in Teamwork?,” Small Group Research, vol. 36, no. 5, pp. 555-599, Oct. 2005. [10] N. B. Moe, T. Dingsøyr, and T. Dybå, “A teamwork model for understanding an agile team: A case study of a Scrum project,” Information and Software Technology, vol. 52, pp. 480–491, 2010. [11] T. L. Dickinson and R. M. McIntyre, “A conceptual framework of teamwork measurement,” in Team Performance Assessment and Measurement: Theory, Methods, and Applications, M. T. Brannick, E. Salas, and C. Prince, Eds. Psychology Press, 1997, pp. 19-43. [12] M. Hoegl and H. Georg, “Teamwork Quality and the Success of Innovative Projects,” Organization Science, vol. 12, no. 4, pp. 435-449, 2001. [13] E. Salas, C. Florida, N. J. Cooke, and M. A. Rosen, “On Teams, Teamwork , and Team Performance: Discoveries and Developments,” Human Factors, vol. 50, no. 3, pp. 540-547, 2008. [14] L. Mathiassen, “Dynamic capabilities in small software firms: A sense-and-respond approach,” IEEE Engineering Management, vol. 54, no. 3, pp. 522-538, 2007.
C. Better Understanding of Dynamic Configurations Increasingly, software organizations require dynamic capabilities to build, integrate and reconfigure resource to maintain responsiveness to highly complex and turbulent environments (see e.g., [14]). This entails organizational configurations enabling spatially and temporally distributed resources to be assembled dynamically to meet the changing needs of the software organization. However, merely connecting people with collaborative technology is not sufficient to guarantee effective team performance. Work is needed to better understand the dynamics of distributed or virtual software teams. D. Increased Emphasis on Team Cognition Although much is generally known about rule-based and automated task performance, far less is known about the complex cognitive tasks typically performed by software teams. Better theories and models of software teamwork and software team effectiveness are therefore needed to better represent and capture this type of collaborative cognition. E. Better Understanding of Multicultural Contexts As software development globalization progresses, there is an increasing need to better understand the role of culture in software team performance. Currently, the majority of team performance research involves US or Western populations. However, this research raises the question of whether the existing models are representative and useful for software teams with a heterogeneous cultural composition. The degree to which the existing models and frameworks of team effectiveness apply to these multicultural contexts must be assessed, and the models must evolve to include the role of culture in team performance. In the introduction we asked the question: Do we need our own team effectiveness models in software development? As with other areas of software development research, we also need more empirical studies and better theoretical grounding in studies of software teams and software team effectiveness. In this article, we have shown that that there exists a large pool of relevant theory in other disciplines, including a number of models of team effectiveness. Research on
29
