Ward Cunningham in his experience report presented at the OOPSLA’92 conference introduced the metaphor of technical debt. This metaphor is related to immature, incomplete or inadequate artifacts in the software development cycle that cause higher costs and lower quality.
A strategy for technical debt management is still a challenge because its definition is not yet part of the software development process.
Carolyn Seaman and Yuepu Guo proposed a technical debt management framework based on three stages:
- Identify and list debts are identified and listed.
- Measure debts by their payment efforts
- Select debts to be considered in the software development cycle.
This study evaluates the application of this framework in the real context of software projects adopting Scrum.
Action research is conducted in two companies where their projects have significant technical debt. We performed three action research cycles based on the three stages of the framework for both companies.
The main contribution of this paper is to provide real experiences and improvements for projects using Scrum and that may adopt the technical debt management framework proposed by Seaman and Guo.
Both teams recognized that the proposed approach is feasible for being considered in the software development process after some modifications.
Because of project time constraints and ease of use, we reduced the use of the proposed metrics to two:
- Principal
- Current Amount of Interest.
In consequence, decision-making was benefitted by the early consideration of the debts that really need to be paid. Instead of using probabilities to find the interest, these are registered every time the technical debt occurs. During the first phase, the debt identification was improved when all Scrum roles participated, while measurement and decision-making were improved when the team was responsible for these phases. The Product Owner role in both companies understood the importance of Technical Debt monitoring and prioritization during a development cycle. With these changes, the two teams mentioned they would remain using the resulting approach.
Frederico Oliveira
Institute of Research Technology
São Paulo, Brazil
[email protected]
Alfredo Goldman
Institute of Mathematics and Statistics,
University of São Paulo
São Paulo, Brazil
[email protected]
Viviane Santos
Federal University of Pará
Belém, Brazil
[email protected]
