Conference PaperPDF Available

Sprint Performance Forecasts in Agile Software Development - The Effect of Futurespectives on Team-Driven Dynamics

Sprint Performance Forecasts in
Agile Software Development
The Effect of Futurespectives on Team-Driven Dynamics
Fabian Kortum, Jil Kl¨
under, Wasja Brunotte, and Kurt Schneider
Software Engineering Group
Leibniz University Hannover, Germany
Email:{fabian.kortum, jil.kluender, wasja.brunotte, kurt.schneider}
Abstract—In agile software development, the sprint perfor-
mances and dynamics of teams often imply tendencies for the
success of a project. Post mortem strategies, e.g., retrospectives
help the team to report and share individually gained experiences
(positives and negatives) from previous sprints, and enable them
to use these experiences for future sprint planning. The interpre-
tation of effects on sprint performance is often subjective, espe-
cially with concern to social-driven factors in teams. Involving
strategies from predictive analytics in sprint retrospectives could
reduce potential interpretation gaps of dynamics, and enhance the
pre-knowledge, also awareness situation when preparing for the
next sprint. In a case study involving 15 software projects with a
total of 130 involved undergraduate students, we investigated the
post-effects on team performances and behavioral-driven factors
when providing predictive analytics in retrospectives. Besides
measures for productivity, we consider human factors, e.g., team
structures, communication, meetings and mood affects in teams
as well as project success metrics. We developed a unique JIRA
plugin called ProDynamics that collects performance information
from projects and derives trend-insights for next sprints. The
ProDynamics plugin enables the use of a times series and neural
network model within a JIRA system to interpret factorial
dependencies and behavioral pattern, thus to show the next sprint
course of a team.
Index Terms—team dynamics, human factors, data analytics,
futurespectives, sprint performances, agile
In agile software development, accurate sprint estimations
and development performances by teams are essential and
can cause the difference whether project goals will be suf-
ficiently completed in time, or neglected [1] [2]. However,
studies with the focus on software process improvement are
often motivated by increasing the overall success of projects,
with the help of using modern methods or technologies that
enable additional feedback [3] [4]. For organizations, sprint
feedback is important and valuable because it allows the teams
to receive both subjective and objective information from
different perspectives, e.g., productivity monitoring systems
like JIRA, customer responses, and team perceptions also
experiences. The feedback mentioned above strongly relies
on human factors [5]. While tool-oriented technologies and
development frameworks are continuously improved, do the
social-driven factors of the team present particular difficulty
DOI reference number: 10.18293/SEKE2019-224
in the development process chain. Accordingly, the need for
understanding the effects of human factors has continuously
grown in relevance for the software engineering discipline, in
particular for the improvement of the development process [6]
[7]. Towards this, team feedback with focus on past sprints
conditions can provide a substantial insight towards earlier
experiences, dysfunctional habits or positive performance in-
fluences. For example, feedback through retrospective sprint
characterization and visualizations benefits of post-mortem
summaries, similarly to sprint reports [8]. However, depen-
dency implications or trend highlighting concerning team
behavior or performances changes over time are barely con-
sidered or hard to grasp. This leads to possible knowledge and
interpretation gaps when preparing follow-up sprints because
previous effects may not be fully identified or adequately
encountered. It is highly desirable to give teams a more
sustainable feedback opportunity that involves sprint feedback
involving both, retrospective and future characterization. Sub-
sequently, we derive the following two research questions from
the above-reported context.
We addressed and investigated RQ1 and RQ2 within a
case study involving 130 undergraduate students working in
15 software projects, all founded from industrial, govern-
ment or public institutional partners. The teams followed a
Scrum-oriented development process with support through the
project management software JIRA. The case study involves
weekly self-assessments resolved in JIRA. The question set
covers team behavioral driven features, to gain the situational
dynamics in a project with effect for the performances in
sprints — satisfactory reflections of customer became ad-
ditionally elicited at the end of every sprint. Productivity
information, e.g., velocity during sprints, estimation gaps or
sprint interventions become tracked and accessed directly by
JIRA. Half of the projects (seven) were granted to obtain a
JIRA plugin called ProDynamics [8]. The plugin is based on
earlier studies, enabling teams to give and receive additional
retrospective feedback for a better knowledge transfer at the
end of sprints [8] [9]. In this study, we extended the primarily
retrospective feedback in ProDynamics with two predictive
analytics features. Teams with access to ProDynamics can
additionally perform time series analyzes for short-term fore-
casts, also sprint performance tendencies with the help of a
neural network encoder-decoder model [10].
This paper is structured as follows. In Section II, we discuss
related work on team behavioral effects, feedback and data
analytics adaptions. Section III provides a brief overview of
the study context. In Section IV, we address the methodology
about self-assessments in sprints, also time series and neural
network forecasts. In Section V, we describe and interpret
results. Section VI concludes our research and future work.
This study builds on previous results and related work with
the focus on human-centered software engineering, and the
relevance of fast feedback in an agile context.
Human-centered software engineering takes an import role
in modern software development [5]. The focus on team
communication, self-organization, and well-working relation-
ships within teams reflect the need for better understanding
of behavior-driven factors [9]. When planning sprints in ag-
ile software development, pre-knowledge on people having
different personalities, skills, and ambitions within the orga-
nizational structure is crucial [7]. By means, understanding
such human interdependencies enrich project improvements
throughout better team performances due to a reduction of
estimation gaps [11]. Besides, e.g., the meeting and communi-
cation manner or atmosphere changes over time can grant valu-
able project directions towards potential follow-up conflicts
or misleading habits that endanger the sprint performances
[12] [13]. Nevertheless, human-centered software engineering
strongly depends on the contribution and self-reflection of
teams, in sharing experiences from previous sprints [14].
Whenever targeting sprint performance improvements in
agile development, team feedback, and change adaptions
should be sincerely considered [15] [4]. In the early 2000s,
research results by Rising et al. [16] subsequently revealed
that team feedback during iterative development phases helps
the teams most when also involving customer feedback and
reflections. The customer perception forms a substantial base
in the improvement process by associating group subjective
sprint perceptions with the team and development expectation
of stakeholder. Vetr`
o et al. [3] also focused on the effect of
fast feedback cycles in software development. The authors
observed the impact of different feedback mechanism when
gathering information from software development teams di-
rected affects for the quality and transferability of experiences
and knowledge with changes in following sprints.
Retrospectives are commonly applied in Scrum and most
other agile processes to share and interpret team experi-
ences on performance effectiveness collected during the last
sprint [5] [11]. Knowledge gained this way is then taken
into account to estimate the next sprints more effectively
according to previous outcomes. However, the interpretation of
team-behavioral pattern often remains subjectively. However,
computer-supported interpretation of sprint performance be-
comes increasingly important and visible - study results, e.g.,
reported by Vetr `
o et al. [15] show the improvement potentials
in combining data analytics with traditional team feedback
appraisals to improve future sprint estimations [17] [15]. The
authors applied multiple-regression analyzes to characterize
behavioral pattern on, e.g., meeting and development manner,
to grant teams better insights on factorial influences over
time, especially for efficiency hazards occurred during sprints.
Our case study combines feedback mechanisms and predictive
analytics about human-centered sprint performance factors.
This approach focuses on the effects of feedback on de-
velopment performance by teams when providing additional
feedback that involves forecasts and sprint tendencies. The
futurespectives considered in this study base on computer-
supported analyzes that relates to team and customer reflec-
tions. Predictive analytics is also integrated to characterize
interdependencies of behavior pattern. We analyzed RQ1 and
RQ2 using a comparative study design [18] by observing 130
undergraduate students working in fifteen project teams (eight
to nine student per team). Each project was pre-estimated with
an approximated development complexity of 2,000 working
hours, equally distributed over 15 weeks. Seven of the fifteen
teams used the ProDynamics JIRA plugin, enabling them to
characterize the previous team and development performances,
and to esteem next sprints with the support of times series
forecasts and neural network prediction models.
Fig. 1. Comparative Study: Projects, Sprints and Self-Assessments
Figure 1 shows a process overview of the comparative case
study design. Participating in the software projects is eligible
for students in the fifth semester of their computer science
undergraduate studies. The main focus is to collect practical
team experience following an iterative development process
including agile methods and practices, such as weekly scrums
and storyboards. Students do not receive gradings for par-
ticipating or performances during the projects. Nevertheless,
the course is compulsory within the syllabus. As mentioned
before, Scrum was mandatory in each project. All teams had to
self-organize inner structures, meetings, and communication,
and to manage sprint tasks on a Scrum-board using the project
management software JIRA.
For a primary information flow and status exchange, each
team was requested to meet face-to-face for at least once a
week. For product progress updates, a subsequent meeting
with the customer was regularly scheduled once a week.
During the sprints, change request could occur, which led to
adjusted or discarded issues. At the end of each sprint, teams
used retrospectives to highlight and reflect on positive and
negative sprint situations. This has mostly been experiences
that help the team to better estimate the next sprint tasks
and organizational structures. During the projects, additional
customer-reflections and satisfaction feedback were assessed
to monitor group and development performances also from
customer perspectives.
This methodology section starts with (A) the ideology of our
JIRA plugin for advanced sprint retrospective and future trend
support. The chapter covers the role and realization of (B) the
self-assessments on teams and customer satisfaction that were
collected as part of the case study. Furthermore, we show how
the elicited team and sprint information is later used within
(C) the time series and neural network prediction models. In
the last subsection, we describe (D) the monitoring process
for the sprint performances with a focus on RQ1 and RQ2.
A. ProDynamics –Retrospectives & Futurespective Support:
The project management system JIRA is worldwide known
for its team-oriented sprint planning and issue tracking sup-
port, commonly used in agile software development. Its usage
is scalable from large industrial projects to small entrepreneur
solutions. The standard features of JIRA provide substantial
help for teams to monitor not yet completed sprints and
to derive performance reports for past sprints (development
velocity). Lessons learned or experiences gained during a
sprint are often reflected by the teams through post-mortem
retrospectives [19]. This way, dysfunctional or beneficial sprint
characteristics, as well as other individually gained insights,
become shared and discussed within the team, enabling them
to plan the next sprint in addition to pre-experienced situations
or manners. However, reasons for performance affects are
not always easy to explain, in particular since the standard
JIRA system only characterizes productivity statistics without
further implication. We addressed this problem in enabling
teams to access fast feedback through a JIRA plugin called
ProDynamics [8]. In Fig. 2, we give an overview on the sprint
analysis features currently provided in ProDynamics.
The red box highlights the two newly integrated sprint
forecasting features of this study. The plugin addresses the
ideology to support and increase the team’s understanding
and awareness for factorial effects on team-driven sprint
behavior and development performance. Previous study results
have shown that after suitable preparation of retrospective
data, teams can be supported by retrospective computer-aided
feedback [8] [15]. However, we believe that the support for
such teams in their sprint estimations can further enrich the
organizational and development performances with the help of
integrated data analytics solutions.
Fig. 2. Behavior-Driven Feedback Support on Sprint Futurespectives in JIRA
B. Self-Assessments for Teams, and Customer-Satisfaction:
Interpretations of previous sprint dynamics or estimations of
follow-up sprints concerning the social nature of agile teams
can be challenging [7]. Various human factors, such as mood,
communication or meeting manner have a direct effect for the
ongoing project, while dysfunctional behavior often remains
undetected or hard to grasp until problems enlarge [9] [20].
Our approach is designed for teams with an open mentality
for self-reflection in exchange for sustainable feedback that
enables opportunities for change-driven improvements of or-
ganizational and development structures [14] [4]. Using the
ProDynamics plugin, we enable an integrated self-assessment
solution for a systematic elicitation of team dynamics in
ongoing sprints. During our comparative case study with n
= 15 projects, three different assessments were applied to
grasp the maximal descriptive team characteristics over time.
The assessment designs and question features are based on
previous studies, also related work [21] [9], and continuously
refined to reach the currently applied versions. The question
set is self-adapting, e.g., the information elicitation about
communication or meeting behavior only occur for members
with active information exchange. With this, the interviewees’
effort to complete a survey could be minimized to a length
of 1-2 minutes. All assessments are realized through 5-points
Likert scales, determining his or her level of agreement on
a symmetric agree-disagree scale with predefined sprint or
team behavioral statements. In the following, we explain the
three assessment types in detail. A summary of the variants,
in particular, the intervals, self-assessment sprint information
by categories, and interviewees are shown in Fig. 3.
Fig. 3. Self-Assessment Intervals, Interviewees and Features
1) Self-Assessment at the End of each Week: During the
15 weeks of this case study, seven out of the fifteen student
developer teams voluntarily participated in the study.
Besides the access to the ProDynamics retrospective and
futurespective sprint features, the participation included a
weekly self-assessment for each team member. The assess-
ments capture social- and organization features, such as:
who-to-whom communication and media channel use
meeting quantity and average duration
the atmosphere in the group, personal mood
satisfaction about the last weeks’ performances
We derived other team and sprint metrics, e.g., meeting
participation, maverick trends, and centralized communication
structures from each team member’s response. A summary of
all assessment information is shown in Fig. 3. Subsequently,
all category features that are currently considered by the
time series and neural network model, including productivity
measures from JIRA, become listed in Fig. 5.
2) Self-Assessment at the End of each Sprint: The second
self-assessment question set on the satisfaction with the team
and development performances for all 15 software projects
was also answered by the customer, and the scrum master
at the end of each sprint (except the first). We activated the
performance assessments with the second sprint, because the
first three weeks of the project were mainly for exploration, to
create and fill the backlog, form team structures, get to know
the customer and reach a steady state before the next sprint.
However, the survey covers in total ten questions, four about
the team organization, four on the development performance
as well as two items focusing on the overall satisfaction with
the team and product. The question structure became realized
through 5-points Likert scales, determining the customers and
scrum master’s level of agreement on a symmetric agree-
disagree scale, similar to the other two self-assessments. The
question set was used as one reference indicator between
team-driven dynamics during the sprints and potential effects
for the customers’ satisfaction. The scrum masters’ responses
become utilized to determine possible offsets between external
views and the team inside knowledge. For the comparison
of satisfaction changes between the teams with access to
ProDynamics and those that did not, the customer and scrum
master of all fifteen projects were invited to complete this
self-assessment form.
3) Self-Assessment at the End of each Project: The third
self-assessment took place only once at the end of each
project and became applied to all 130-student developer. The
survey includes questions on the personal perception of the
developers, e.g., whether there were moments with a need
for additional feedback during the sprints, and if there were
recognizable effects (positive or negative) within the own team
in case of provided feedback. The following four assessment
features became only ones elicit at the end of each project for
the validation of every student’s perception of their team and
development performances during the 15 weeks:
importance of organizational feedback for the team
importance of product feedback for the development
perceived effect in the team because of team feedback
perceived effect in the team because of product feedback
Fig. 4. Time Series Sprint Forecasts derived in ProDynamics
The responses were compared with the customer and scrum
master satisfaction after each sprint. This survey also involves
questions about the need and usefulness, e.g., of a centralized
feedback solution in JIRA. Also, whether JIRA was an ade-
quate solution to manage and organize sprints during the case
study project. Those are by-product information, in particular
with no relevance to answer RQ1 and RQ2.
C. Sprint-Trends through Time Series and Neural Networks
With this case study, we investigated whether teams can
gain a more sustainable use of feedback when considering
both, past retrospective records and future sprint performance
predictions. Besides this, we incorporated retrospective sprint
characteristics (e.g., organization structures, communication-
and meeting behavior, productivities, motivation) with two
predictive models to grant teams an additional trend char-
acterization on behavior-driven factors when esteeming the
next sprints. We chose both predictive methods based on
their functional properties in supporting inference-statistical
analysis, e.g., sprint series. The analysis helps to estimate
sprint and team measurements in the future based on past
team-behavior in sprint sequences.
The times series forecast used for the sprint-trend esteems
based on an open source java library published by Workday1.
The library provides time series analyzes, involving ARIMA-,
Mean-, Na¨
ıve- and Drift-forecasts. The forecasting model in
this study become fitted by the measurements listed in Fig. 4.
The ARIMA-model characterizes seasonal inferences from
past and forecasts future points in the series [10]. The Mean-
method derives the arithmetic mean from the sequence of
sprint data to esteem follow-up values within fitted parameter
ranges. The seasonal Na¨
ıve-method uses sprint metrics from
the second sprint week on to predict the third sprint parame-
ters. The prediction of the fourth sprint metrics is derived with
the values from the third sprint, and so on. The Drift-method
obtains a straight line between the first and last data point to
characterize the sprint metrics tendency drift. Seasonal patterns
are not taken into account with this Drift-method. Figure 4
shows a time series forecast for the communication metrics
Media Channel Usage and Perceived Intensity.
The time series viewer enables the teams to choose from
one to four supported forecasting methods. The interactive
chart allows a user to select different past sprints and the
underlying metrics. With this, the chosen sprint metric(s)
become analyzed and forecast with the help of the four
forecasting methods. The prediction interval can reach a
maximal length smaller than the number of yet completed
sprints. The colored lines within the gray background area
shown in Fig. 4 present the real data points for two selected
communication parameters. The colored areas on the right half
of the chart mark the 95% confidence interval of the prediction
of each forecast. For example, the maximally available forecast
horizon in Fig. 4 is two sprints, because the time series model
derives its prediction based on three yet completed sprints.
Sprint forecasts are labeled on the time axis through a counter
and completed sprints name tags.
The ProDynamics – Neural Network viewer focuses on
the second prediction approach for estimating social-driven
team measurements based on retrospective sprint and team
records. The neural network is implemented in using the open
source library Deeplearning4j. The Deeplearning4j-library
covers cross-platform algorithms on machine learning and
artificial intelligence is implemented in Java and runs in a
JVM such as used by the JIRA system. Similar to our time
series forecast solution the neural network viewer provides
past sprint metric plots based on the user selections.
Also, the neural network model allows the team to re-
view past sprint conditions covering organizational structures,
social-driven team behavior, customer satisfaction, productiv-
ity measures as well as problem and conflict appearances. An
overview of all covered factors is listed in Fig. 5, which is
a result of the assessed team and customer responses as well
of development performances that is natively tracked in JIRA.
The model is trained with the listed data features considering
the availability of already completed sprint records.
Fig. 5. Neural Network Sprint Predictions derived in ProDynamics
Training the model requires enlarged computation effort
for the data encoding process. Therefore, model updates are
performed automatically, but only once during the weekend.
The neural network viewer enables the teams to choose
individual real data plots of past sprints, or with an active
future horizon. The maximal size of future-horizon is limited
to the number of yet completed sprints minus one, similar
to the time series model. The interactive chart allows a user
to select between the sprint metrics and plot trends for the
considered team-driven factors. The colored lines in front of
the dash border line present the real data points for the two
chosen communication parameter Media Channel Usage and
Perceived Intensity. The colored lines on the left side of the
dashed border present the computed prediction according to
the feature records encoding of three yet completed sprints.
D. Sprint Performance Monitoring
In this study, the sprint performances of teams’ base on the
development (velocity of tasks) as well on the organizational
performances during each sprint. The velocities in all teams
are comparable productivity measures tracked within JIRA,
thus did not require to be separately assessed. This enabled
us a direct performance comparison was between a particular
group and the average of all other teams during a sprint.
Besides the sole productivity measures of teams, we used
the customer and scrum master satisfaction feedback after
each sprint to determine whether the development outcome
also fulfilled product expectations according to quality and
functional requirements.
However, the team performances over time with concern on
social-driven changes due to futurespective feedback became
solely traced through the customer, and scrum master feedback
elicit at the end of every sprint. We are reasoning this
processing with the fact, that only half of all teams could
access the ProDynamics futurespectives, and also frequently
completed the self-assessments on social-driven behaviors. In
considering the customer and scrum master perceived team
performances during each sprint, we could compare the or-
ganization performance changes of all teams. Subsequently,
the effects and trends could become characterized due to the
comparative study subjects with different sprint estimation
and planning support. Of course, at this appraisal level, sole
factorial influences become not closer taken into account.
However, it allowed us first interpretations about whether
teams adopt the ProDynamics usage, also whether groups with
access get used to derive better sprint estimations with constant
or even positively improving customer satisfaction outcome,
on organizational and product aspects.
In the following two subsections, we statistically interpret
and discuss the effects of ProDynamics futurespectives on the
sprint performances, also emphasize related threats to validity.
A. Interpretation of statistical measures
In Section IV, we described the sprint performances as
a result of sprint productivity (velocity), also the customer
and scrum master satisfaction responses on the team and
product performances after every sprint. With the help of Pear-
son correlation analysis and 2-paired t-Tests, we determined
sprint performance differences between the seven groups that
actively used the ProDynamics prediction features and the
other eight teams without access. We found out that the
teams with access to the provided sprint forecasts showed
fewer estimation errors, therefore with more optimal velocity
distribution at 98% as the comparison of the orange boxes
(1) in Fig. 6 and Fig. 7 reveal. The overall sprint estimation
error for those teams was ±9 %. In particular, do the groups
Fig. 6. Team Projects without ProDynamics-Futurespective access
without access to ProDynamics showed strong tendencies for
over-estimating their sprint tasks during the first two sprints,
followed by underestimations that caused strong deviations
between the number of scheduled tasks and completed ones.
Due to this, an overall sprint estimation error of ±19 %
was identified. However, in comparing the yellow markings
(2) in both figures, the team’s organizational performances
revealed no benefits given due to the additionally provided
sprint forecasts. Moreover, the outcome is on a constant level,
while both, customer and scrum master reflected a sustainable
organization and communication structure in most teams.
The third factor of the sprint performances involves the
software product, in particular, the quality, and requirements
fulfillment after each sprint. The futurespective in ProDy-
namics enabled a few teams to preview customer satisfaction
according to previous performances. However, the forecasts
only highlight chances for adjustment, while the teams decide
whether to use the available information to improve the previ-
ous situation. By comparing the blue boxes (3) in Fig. 6 and
Fig. 7, an affect for teams with ProDynamics usages becomes
reflected throughout an increasing product satisfaction by the
customer and scrum master. The satisfaction increase can be of
course also because of excitement about the product majority.
Nevertheless, a significant rise in teams with customer
feedback knowledge could be measured, while comparable
projects remained on a constant level. For redundancy, we
also considered the product satisfaction of scrum master,
which significantly correlates with our interpretation. Beside
results of this case study showed strong accordance between
customer and scrum master perceived team and development
performances. While the scrum master usually tends to have
more team internal information and critical knowledge about
accomplishments, does the deviation with external customer
ratings present a low perception gap.
B. Threats to validity
Construct validity: We looked at the social-driven team
affects only through statistical and artificial methods. The
sprint information features obtained in ProDynamics are cho-
sen based on previous studies [18] [13]. However, the ac-
Fig. 7. Team Projects with ProDynamics-Futurespective access
curacy of the predictions strongly depends on the quality
and completeness of the self-assessed data. The effects on
performances depend on whether the teams considered the
sprint forecasts in follow-up esteems. Customer satisfaction
responses as one success indicator might have involved offsets,
because of unjustified expectation or lack of experience.
Internal validity: The interpretations and results of this
study rely on the self-assessed team and customer feedback.
Only voluntary groups completed surveys, in exchange for
accessing additional prediction support in JIRA. Therefore,
team responses can be assumed to be accurate and unbiased
[16]. Participating teams accepted the privacy limitation, e.g.,
communications and performances were viewable, but exclu-
sively by the assigned team.
External validity: Since we examine student teams, the
results should not be overgeneralized. However, the software
projects are founded by a real customer from industry, public
institutions or governments. Hence, data collected from other
company or university projects could lead to different results.
Due to further limitations like the involvement of social-driven
factors and unknown domain influences, our interpretations
may not embrace all possibly existing project scenarios.
Conclusion validity: All interpretations are plausible and
statistically valid. However, there may be different self-
assessment responses when repeating the project with the
same participants: team-behavioral factors, emotions, skills
or unknown influences could have changed in the meantime.
Subsequently, the accuracy of predictions could vary due to
differently completed surveys, and project progresses. How-
ever, the methodology can be generalized and applied to
various agile projects that allow assessments on team and
sprint information.
This study focuses on the effects of social-driven dynam-
ics in agile software developments when providing teams
additional feedback on sprint tendencies. To determine the
impact of futurespective feedback, we realized a JIRA plugin
named ProDynamics that simplifies the elicitation of team-
driven factors as well as performance measures within JIRA.
The feedback becomes resolved through sprint series forecasts
and neural network predictions in an integrated JIRA plugin
solution. The computer-based sprint analyzes use team and
customer reflections that were frequently assessed, analyzed
and characterized for factorial interdependencies on develop-
ment performances and team-driven behavioral pattern.
With the help of a comparative case study involving fifteen
software projects with 130 students, throughout 15 weeks, we
gathered weekly information about communication, meeting
and emotional metrics from half of the projects. The elicit data
became frequently used to train time series and neural network
models, enabled the 7 out of 15 groups to gain additional
insights about previous sprint and team performances, also
derive trend-forecasts for follow-up sprints. Measuring the
performance differences between the groups with pro-active
feedback and those without involved customer and scrum mas-
ter feedback from all 15 projects, that became repeatedly elicit
at the end of every sprint. The feedback covered past sprint
perceiving on both, team and development performances.
Pearson correlation statistics helped us to interpret the
effects on sprint performance, in particular, the team and
development performance in each project. We found statistical
evidence towards that the groups with access to the addi-
tionally provided ProDynamics forecasts showed a definite
decrease for sprint estimation gaps by 10%, while the groups
without ProDynamics access tend to have more volatile veloc-
ity performances. We could show, that the additional use of
forecasting methods supports the groups to interpret customer
satisfaction better, thus improve the product outcome at the
end of sprint. The study also revealed, that teams not neces-
sarily adjust internal organization structures due to predictive
information. Most of the groups showed an almost steady level
in their weekly communication and meeting behavior, towards
no significant affects could be determined.
We can conclude that the ProDynamics futurespectives
enabled a sustainable team organizational and development
performance improvement for the groups with access to the
plugin. The team performances dynamics during the sprint
sequences showed strong stabilizing characteristics, due to
more accurate sprint esteems compared with the comparison
groups that only used general sprint information in JIRA,
e.g., burndown- and velocity charts. Besides, the ProDynamics
plugin realized a simplified data elicitation for social-driven
team factors, while some group had could reach a positive
effect for follow-up sprint executions.
We are currently working on a newer version for the
ProDynamics plugin, that does extend the retrospective, and
futurespective sprint analyzes, by a planning-oriented simu-
lation feature. With this, we believe that teams can poten-
tially improve sprint estimations because of training effects.
Various scenarios could be explored before an official sprint
start, by incorporating past performances with a generalized
simulation model for agile development processes, e.g., system
dynamics. A simulation-based approach could grant the team a
better insight about appearing behavioral dynamics over time,
also help to discover new characteristics, that would remain
undetermined otherwise. Generally spoken, simulations could
gain further knowledge and train the sprint estimation skills
of teams and project manager.
This work was funded by the German Research Society
(DFG) under the project name Team Dynamics (2018-2020).
Grant number 263807701.
[1] N. Agarwal and U. Rathod, “Defining ‘success’ for software projects:
An exploratory revelation,International journal of project management,
vol. 24, no. 4, pp. 358–370, 2006.
[2] P. Mohagheghi and M. Jørgensen, “What contributes to the success of
it projects? an empirical study of it projects in the norwegian public
sector.JSW, vol. 12, no. 9, pp. 751–758, 2017.
[3] A. Vetr`
o, S. Ognawala, D. M. Fern´
andez, and S. Wagner, “Fast feedback
cycles in empirical software engineering research,” in Proceedings of
the 37th International Conference on Software Engineering-Volume 2.
IEEE Press, 2015, pp. 583–586.
[4] L. Williams and A. Cockburn, “Agile software development: it’s about
feedback and change,” IEEE Computer, vol. 36, no. 6, pp. 39–43, 2003.
[5] A. Cockburn and J. Highsmith, “Agile software development: The people
factor,Computer, no. 11, pp. 131–133, 2001.
[6] F. Kortum, J. Kl¨
under, and K. Schneider, “Characterizing relationships
for system dynamics models supported by exploratory data analysis,” in
29th International Conference on Software Engineering and Knowledge
Engineering. KSI Research Inc, vol. 15, 2017, pp. 39–43.
[7] E. Whitworth and R. Biddle, “The social nature of agile teams,” in Agile
2007 (AGILE 2007). IEEE, 2007, pp. 26–36.
[8] F. Kortum, J. Kl¨
under, and K. Schneider, “Behavior-driven dynamics in
agile development: The effect of fast feedback on teams,” in Proceedings
of the 2019 International Conference on Software and System Process.
ACM, 2019.
[9] K. Schneider, O. Liskin, H. Paulsen, and S. Kauffeld, “Media, mood, and
meetings: Related to project success?” ACM Transactions on Computing
Education (TOCE), vol. 15, no. 4, p. 21, 2015.
[10] G. P. Zhang, “Time series forecasting using a hybrid arima and neural
network model,” Neurocomputing, vol. 50, pp. 159–175, 2003.
[11] N. B. Moe and T. Dingsøyr, “Scrum and team effectiveness: Theory and
practice,” in International Conference on Agile Processes and Extreme
Programming in Software Engineering. Springer, 2008, pp. 11–20.
[12] S. Kauffeld and N. Lehmann-Willenbrock, “Meetings matter: Effects
of team meetings on team and organizational success,” Small Group
Research, vol. 43, no. 2, pp. 130–158, 2012.
[13] J. Kl ¨
under, K. Schneider, F. Kortum, J. Straube, L. Handke, and S. Kauf-
feld, “Communication in teams-an expression of social conflicts,” in
Human-Centered and Error-Resilient Systems Development. Springer,
2016, pp. 111–129.
[14] C. J. Stettina and W. Heijstek, “Five agile factors: Helping self-
management to self-reflect,” in European Conference on Software Pro-
cess Improvement. Springer, 2011, pp. 84–96.
[15] A. Vetro, R. D¨
urre, M. Conoscenti, D. M. Fern´
andez, and M. Jørgensen,
“Combining data analytics with team feedback to improve the estimation
process in agile software development,Foundations of Computing and
Decision Sciences, vol. 43, no. 4, pp. 305–334, 2018.
[16] L. Rising and N. S. Janoff, “The scrum software development process
for small teams,” IEEE software, vol. 17, no. 4, pp. 26–32, 2000.
[17] F. A. Batarseh and A. J. Gonzalez, “Predicting failures in agile software
development through data analytics,Software Quality Journal, vol. 26,
no. 1, pp. 49–66, 2018.
[18] T. Dyb˚
a and T. Dingsøyr, “Empirical studies of agile software devel-
opment: A systematic review,” Information and software technology,
vol. 50, no. 9-10, pp. 833–859, 2008.
[19] J. Highsmith and A. Cockburn, “Agile software development: The
business of innovation,Computer, vol. 34, no. 9, pp. 120–127, 2001.
[20] K. Schneider, J. Kl¨
under, F. Kortum, L. Handke, J. Straube, and
S. Kauffeld, “Positive affect through interactions in meetings: The role of
proactive and supportive statements,Journal of Systems and Software,
vol. 143, pp. 59–70, 2018.
[21] J. A. Ross, “The reliability, validity, and utility of self-assessment,” 2006.
... We strive towards investigating sustainable interdependencies with long-term relevance for individual teams and across projects. This approach is part of a series of previous publications on team feedback described by Kortum et al. [8,10]. The authors focus on pro-active feedback using retrospective sprint visualizations and future trend highlighting. ...
... We developed a JIRA plugin called ProDynamics. Based on this plugin, we collected sprint data of six software projects with a total of 53 undergraduate students using controlled self-assessments on the social-driven team factors and productivity measures [8,10]. ...
... In agile software development, adequate summaries, visualization of rapidly changing sprint condition, and progress information are essential [4,13]. Retrospective reports and futurespective implications enable overall conclusive insights about sprint performances and functional manners, thus support teams in the next sprint planning [8,10]. This part of the research extends the existing visual concepts in ProDynamics with force-directed network graphs for enhanced transparency and understanding of team-driven dependencies during sprints. ...
Full-text available
In agile software development, proper team structures and sprint estimations are crucial aspects to reach high-performance outcomes. Performance can vary due to the influence of social-driven team factors. Resulting in team dynamics with the focus on human factors are usually difficult to capture and thus often not monitored. However, their impact can impede the planning and fulfillment of sprints. Data on team behavior should be simplified to track, analyze, and interpret as sprint influences are important to understand. We provide a centralized solution that extends JIRA functionally and continuously captures sprint characteristics in the daily working environment of teams. In this paper, we describe a JIRA plugin that enables the assessment of team behavior in combination with exploratory analyses. The tool became approached with six software projects and a total of 53 undergraduate students. Characterizations made with the plugin can reveal sprint and team dynamics over time, involving development performance and team-related measures. The feature comes with a feedback mechanism for teams that visualize and implicates the sprint dependencies. The approach reveals a set of team-related sprint dynamics, its systematically capturing, and characterization. With the achieved solution, team leader and developer can be supported to understand the ongoing sprint and team-driven dynamics better. Thus, they can keep track of their habits for future sprint planning and team adjustment impacts.
... The following two questions were added in the survey to receive some first feedback on the acceptance and thoughts on the JIRA solution. In the last section of this survey, we introduced the functional features of our operationalized system-aided feedback solution for JIRA [13], [14], [28]. An animated overview of the information support was provided in the survey. ...
Full-text available
Fast feedback promotes agile teams to improve their work during the software process, making it crucial for team success. Information systems accelerate the availability of information that result in compact knowledge sources. In practice, feedback in Sprints is often limited to sole progress and performance measures, e.g., burndown charts or velocity diagrams. Sprint insights related to team dynamics are rarely considered, even though they frequently cause project failures, e.g., lack of social interaction. In this paper, we describe a survey study conducted with international members of the software engineering community to reveal which information helps agile teams the most and provides practical support in Sprints. We describe results in an experience report highlighting the frequent information problems and needs of agile teams, considering the perspective of 90 researchers and practitioners. The responses were quantitatively interpreted. The report promotes understanding about how or what kind of information would be useful for agile development teams. Moreover, it reveals what information problems were perceived as crucial for project success and avoidable, considering proper team feedback. The study endorses practical needs for system-aided feedback that supplies knowledge on the human factors in Sprints. The findings are relevant for practitioners and researchers that struggle on improving team feedback based on information needs.
Full-text available
The complexity of software projects and inherent customer demands is becoming increasingly challenging for developers and managers. Human factors in the development process are growing in importance. Consequently, understanding team dynamics is a central aspect of steady development planning and execution. Despite the many available management systems and development tools that are being continuously improved to support teams and managers with practical process information, the equally crucial sociological aspects have typically been addressed insufficiently or not at all. In people-focused agile software processes, a first socio-technical understanding can also be promoted by sharing positive and negative development experiences during specific team meetings (e.g., sprint Retrospectives). Nevertheless, there is still a lack of systematically recorded and processed socio-technical information in software projects, making it difficult for subsequent reviews by teams and managers to characterize and understand the sometimes volatile and complex team dynamics during the process. This thesis strives to support teams and managers in understanding and improving awareness of the team dynamics that occur in their agile software projects by introducing computer-aided sprint feedback. The concept builds on four information assets: (1) socio-technical data monitoring, (2) descriptive sprint feedback, (3) predictive sprint feedback, and (4) exploratory sprint planning. These assets unify interdisciplinary fundamentals, practical methods from software engineering, data science, organizational and social psychology. Using a design science research process for information systems, observations in several conducted studies (32 in academic project environments and three in industry) resulted in the foundations and methods for a practical feedback concept on the socio-technical aspects in sprint, prototypically realized for Jira. A practical evaluation involved two industry projects in an action research methodology that helped improve the concept’s usability and utility through practitioner reflections. The collaboration between industry and research resolved practical issues that did not arise during the design science process. Several beneficial outcomes based on the provided sprint feedback are reported and described in this study (e.g., the effect of team structures on development performance). Moreover, the reflections underscored the practical relevance of systematic feedback and the need to better understand human factors in the software development process.
Conference Paper
Full-text available
Agile software development teams strive for fast and continuous feedback. Both the quality of the resulting software and the performance of the team require feedback. The performance of the team developments is often addressed in retrospectives which are not only part of the SCRUM framework, but also in general. Reflecting on incidents during the last sprint helps the team to increase performances, expressed by, e.g., efficiency and productivity. However, it is not only essential to identify volatile sprint performances, but also to characterize the primary cause to solve them. Main reasons for low performance are often not visible, primarily when they are related to socialdriven team behavior, such as communication structures, mood, or satisfaction. In this paper, we analyze whether automated team feedback about retrospective sprint-behavior can help the team to increase performances due to additional awareness about the dynamic effects over time. In a comparative case study with 15 software projects and a total of 130 undergraduate students, we investigated the sustainable impact of feedback on human aspects. Our results indicate that automated feedback positively affects team performances – and customer satisfaction.
Full-text available
We apply a mixed research method to improve the user stories estimation process in a German company following agile software development. We combine software project data analytics with elicitation of teams’ feedback, identify root causes for wrong estimates and propose an improved version of the estimation process. Three major changes are adopted in the new process: a shorter non numerical scale for story points, an analogy-based estimation process, and retrospectives analyses on the accuracy of previous sprints estimates. The new estimation process is applied on a new project, and an improvement of estimates accuracy from 10% to 45% is observed.
Full-text available
Artificial intelligence-driven software development paradigms have been attracting much attention in academia, industry and the government. More specifically, within the last 5 years, a wave of data analytics is affecting businesses from all domains, influencing engineering management practices in many industries and making a difference in academic research. Several major software vendors have been adopting a form of “intelligent” development in one or more phases of their software development processes. Agile for example, is a well-known example of a lifecycle used to build intelligent and analytical systems. The agile process consists of multiple sprints; in each sprint a specific software feature is developed, tested, refined and documented. However, because agile development depends on the context of the project, testing is performed differently in every sprint. This paper introduces a method to predict software failures in the subsequent agile sprints. That is achieved by utilizing analytical and statistical methods (such as using Mean Time between Failures and modelling regression). The novel method is called: analytics-driven testing (ADT). ADT predicts errors and their locations (with a certain statistical confidence level). That is done by continuously measuring MTBF for software components, and using a forecasting regression model for estimating where and what types of software system failures are likely to occur. ADT is presented and evaluated.
Conference Paper
Full-text available
Background/Context: Gathering empirical knowledge is a time consuming task and the results from empirical studies often are soon outdated by new technological solutions. As a result, the impact of empirical results on software engineering practice is often not guaranteed. Objective/Aim: In this paper, we summarize the ongoing discussion on "Empirical Software Engineering 2.0" as a way to improve the impact of empirical results on indus- trial practices. We propose a way to combine data mining and analysis with domain knowledge to enable fast feedback cycles between researchers and practitioners. Method: We identify the key concepts on gathering fast feedback in empirical software engineering by following an experience-based line of reasoning by argument. Based on the identified key concepts, we design and execute a small proof of concept with a company, to demonstrate potential benefits of the approach. Results: In our example we observed that a simple double feedback mechanism notably increased the precision of the data analysis and improved the quality of the knowledge gathered. Conclusion: Our results serve as a basis to foster discus- sion and collaboration within the research community for a development of the idea
Full-text available
This study follows the idea that the key to understanding team meeting effectiveness lies in uncovering the microlevel interaction processes throughout the meeting. Ninety-two regular team meetings were videotaped. Interaction data were coded and evaluated with the act4teams coding scheme and INTERACT software. Team and organizational success variables were gathered via questionnaires and telephone interviews. The results support the central function of interaction processes as posited in the traditional input-process-output model. Teams that showed more functional interaction, such as problem-solving interaction and action planning, were significantly more satisfied with their meetings. Better meetings were associated with higher team productivity. Moreover, constructive meeting interaction processes were related to organizational success 2.5 years after the meeting. Dysfunctional communication, such as criticizing others or complaining, showed significant negative relationships with these outcomes. These negative effects were even more pronounced than the positive effects of functional team meeting interaction. The results suggest that team meeting processes shape both team and organizational outcomes. The critical meeting behaviors identified here provide hints for group researchers and practitioners alike who aim to improve meeting success.
Full-text available
Agile software development represents a major departure from traditional, plan-based approaches to software engineering. A systematic review of empirical studies of agile software development up to and including 2005 was conducted. The search strategy identified 1996 studies, of which 36 were identified as empirical studies. The studies were grouped into four themes: introduction and adoption, human and social factors, perceptions on agile methods, and comparative studies. The review investigates what is currently known about the benefits and limitations of, and the strength of evidence for, agile methods. Implications for research and practice are presented. The main implication for research is a need for more and better empirical studies of agile software development within a common research agenda. For the industrial readership, the review provides a map of findings, according to topic, that can be compared for relevance to their own settings and situations.
Software projects are dominated by meetings. For participants, not all meetings are useful and enjoyable. However, interaction within a meeting has an impact on individual and group affects. Group affect influences team performance and project success. Despite frequent yet vague dissatisfaction with some meetings, many software engineers are not aware of the crucial importance of their behavior in those meetings. This can set the tone for the entire project. By influencing group affect, meeting interaction influences success without participants even noticing. Due to this lack of awareness, it depends on good or bad luck whether software teams will adopt a promising meeting style. In a study of 32 student projects with 155 participants, we coded fine-grained interaction elements during the first internal meeting of each team. The analysis of resulting codes showed that constructive remarks had a positive impact on positive group affect tone (PGAT). However, this effect was only observed when constructive remarks were followed by supportive utterances. We were able to show a complete mediation of this statistically significant effect. Seemingly subtle behavior patterns influence group affect. Software projects could significantly benefit from supportive meeting behavior. We propose practical interventions to improve meeting quality.
Despite widespread use of self-assessment, teachers have doubts about the value and accuracy of the technique. This article reviews research evidence on student self-assessment, finding that (1) self-assessment produces consistent results across items, tasks, and short time periods; (2) self-assessment provides information about student achievement that corresponds only in part to the information generated by teacher assessments; (3) self-assessment contributes to higher student achievement and improved behavior. The central finding of this review is that (4) the strengths of self-assessment can be enhanced through training students how to assess their work and each of the weaknesses of the approach (including inflation of grades) can be reduced through teacher action.
Success is found relatively rare in the world of software projects. One possible reason may be the difference in the perception of the meaning of ‘success’ in the minds of people who evaluate the project performance. Usually, stakeholders external to the project organization use target cost and time for inferring ‘project success’ while those internal to project agree that the attainment of ‘scope’ of development decides the ‘project success’. Hence, project success criteria, as believed by different groups of stakeholders, do not match. In this study, we examine the views of one such internal set of stakeholders: Programmer/Developers, Project Managers and Customer Account Managers. We conducted an exploratory survey to determine their view of a successful project. We found surprising uniformity in different constituents of this particular group of stakeholders and all of them overwhelmingly considered meeting the ‘scope’ of software projects, which comprises the functionality and quality of the project outcome, as the highest determinant of success. We believe that this view of software project success in the eyes of software professionals should be studied further to build a comprehensive project evaluation framework and should be utilized effectively to achieve success in terms of external objectives like customer satisfaction and customer happiness.