Data-driven process improvement in Agile software development – An industrial multiple-case study
Thesis event information
Date and time of the thesis defence
Place of the thesis defence
TS101, Linnanmaa campus
Topic of the dissertation
Data-driven process improvement in Agile software development – An industrial multiple-case study
Doctoral candidate
Master of Science Prabhat Ram
Faculty and unit
University of Oulu Graduate School, Faculty of Information Technology and Electrical Engineering, Empirical Software Engineering in Software, Systems and Services (M3S)
Subject of study
Information Processing Science
Opponent
Professor Martin Höst, Lund University
Custos
Professor Markku Oivo, Empirical Software Engineering in Software, Systems and Services (M3S)
Data-driven process improvement in Agile software development – An industrial multiple-case study
Practitioners have been trying to capitalise on the software development data produced as a result
of the use of modern development approaches like Agile Software Development (ASD).
Structured software development data have been utilised in software metrics programme for
undertaking process improvement, but what metrics practitioners prefer and why need further
clarity, especially in large industrial contexts. Success factors for a metrics programme in ASD are
also not as well understood as they are in traditional software development. Lastly, there is little
knowledge on how practitioners can capitalise on their unstructured data, which are generated in
larger volume than structured data.
In the context of the European Union Horizon 2020 project Q-Rapids, a multiple-case study
was conducted with four software-intensive Agile companies to address the above research gaps
in two phases. In the first phase, knowledge about the state of the practice and the practitioners'
perspective influencing the definition, operationalisation, and use of metrics programme was
gathered. In the second phase, empirical evidence for how practitioners utilised their structured
data in a metrics programme for process improvement was elicited. Lastly, empirical evidence was
sought on how practitioners can capitalise on their unstructured data.
To utilise their data to increase awareness and exercise control, practitioners prefer metrics for
measuring planning, implementation, and testing processes. Contextual factors like company size
and project characteristics determine if metrics will be a trigger or the main driver for process
improvement. The prerequisites that facilitate such use of metrics programmes concern data,
process, and metrics actionability. For unstructured data, text-mining techniques such as Latent
Dirichlet Allocation can help derive development-related insights. These results highlight the
utility of two distinct approaches that practitioners can use to capitalise on their software
development data, even in large industrial contexts.
of the use of modern development approaches like Agile Software Development (ASD).
Structured software development data have been utilised in software metrics programme for
undertaking process improvement, but what metrics practitioners prefer and why need further
clarity, especially in large industrial contexts. Success factors for a metrics programme in ASD are
also not as well understood as they are in traditional software development. Lastly, there is little
knowledge on how practitioners can capitalise on their unstructured data, which are generated in
larger volume than structured data.
In the context of the European Union Horizon 2020 project Q-Rapids, a multiple-case study
was conducted with four software-intensive Agile companies to address the above research gaps
in two phases. In the first phase, knowledge about the state of the practice and the practitioners'
perspective influencing the definition, operationalisation, and use of metrics programme was
gathered. In the second phase, empirical evidence for how practitioners utilised their structured
data in a metrics programme for process improvement was elicited. Lastly, empirical evidence was
sought on how practitioners can capitalise on their unstructured data.
To utilise their data to increase awareness and exercise control, practitioners prefer metrics for
measuring planning, implementation, and testing processes. Contextual factors like company size
and project characteristics determine if metrics will be a trigger or the main driver for process
improvement. The prerequisites that facilitate such use of metrics programmes concern data,
process, and metrics actionability. For unstructured data, text-mining techniques such as Latent
Dirichlet Allocation can help derive development-related insights. These results highlight the
utility of two distinct approaches that practitioners can use to capitalise on their software
development data, even in large industrial contexts.
Last updated: 23.1.2024