Swigger et al 2011
Objective
They want to detect temporal communication patterns in global software student teams.
Approach
They got transcripts from student group’s communications and classify them according Curtis and Lawson’s code scheme.
They split communications in project’s 10%-periods and compare those values against its adjacent (10% vs 20% period, 20% vs 30%, and so on).
They compare communication type and performance.
Results
They detected 2 points of significant difference between 10%-periods: 40% and 70%. They related these peaks in data with software development cycle’s phases of design (40%) and coding (70%).
When comparing communication type and performance, they found that high-performance teams have a high planning phase early, early social interaction and high values of contributing and seeking input after “planning phase”.
Low- and high-performance groups were correlated in contributing and seeking input behaviors (but they were not the same amplitude), and they were not correlated in social, planning and reflection communications.
They also found that total number of communications is not a good predictor for performance.
What should this paper be cited for?
The finding that team’s good performance is likely when the “rhythm” of the software development cycle is follow, i.e. planning and social at an early stage, followed by high number of contributing and seeking input communications (coding phase).
The distribution of the number of communication by 10%-period varies during the project; besides, this is significant in the 40%- and 70% of the project (using data from both low- and high-performance groups).
Important, clever or liked
It was clever to split project in 10 periods (10% each one) and then compare with an statistic tool whether adjacent periods have a difference.
It is important that they related those changes (40% and 70%) with the design and coding phase of software development cycle.
It is interesting the way they analyzed “when” a specific type of communication should occur in each project for getting a good performance.
How could it be improved?
When analyzing communication type and performance, it seems that chart values are the sum of communications:
g1’s social communication + g2’s social communication + ….. + gn’s social communication
so, some “peaks” could be more related to a specific project rather than a overall behavior. So, instead of using sum of communications, we could use an average (with variance) and see whether these behaviors (peaks in specific periods) stay. At this moment, we can assume that each team has the same behavior in the set (low- or high performance groups).
Analyze the distribution on each team of these communications: How is the communications distribution within the team? In high-performance group is there someone (leader) who produce more communications? what kind of communications?
What is the rhythm within a group? Analyze communication behavior within a group. So, we can notice its interaction cycles. What kind of interaction cycles low- or high-performance group have? It is possible that almost a consistent cycle could be in high-performance group, assuming this means that some people are interacting frequently.
Image source: reinholdbehringer
This work is licensed under a Creative Commons Attribution 3.0 Unported License.
Learn, code & play 