What is it about?
In finance, taking on debt can help you move faster in the short term, but it comes with interest, and if that interest isn't paid off, it accumulates into a serious burden. Software development has a similar problem, known as "technical debt": developers often make quick, suboptimal design decisions to ship faster, and while this pays off in the short term, it makes future changes to the code progressively more expensive. One of the most damaging forms of this debt comes from poor architectural decisions: the way a system's components are structured and connected. When left unaddressed, these decisions manifest as "architectural smells". A common concern is that these smells make two files more likely to need updating together, even without an obvious reason for them to be linked. This phenomenon is called co-change. To study this at scale, we built an automated data collection pipeline that continuously scans thousands of open-source GitHub projects with a specialized analysis tool (Arcan), producing a large, constantly updated dataset of architectural smells. The results confirm the concern: files affected by architectural smells are significantly more likely to change together than clean files, especially as file size increases. Poorly designed code doesn't just look messy, it measurably creates extra maintenance work. Practically, this gives developers a way to prioritize: large files that keep changing together with a smelly counterpart are the strongest candidates for refactoring first.
Featured Image
Photo by Pankaj Patel on Unsplash
Why is it important?
Our study offers two important contributions. First, most research on architectural smells relies on small and static datasets. Our data collection pipeline addresses this by continuously and automatically collecting data from thousands of GitHub projects, offering other researchers a reusable resource for large-scale studies. Second, our empirical findings give developers something actionable: smelly files really do create measurably more maintenance work through co-changes, especially in large files.
Perspectives
This paper marks a real milestone for me: it's my very first publication in absolute, and on top of that, my first in a journal, as corresponding author. Beyond the result itself, this project taught me how to actually do research and how to structure a study design that's as rigorous as possible and as free of bias as possible.
Dr. Matteo Bochicchio
Universita degli Studi di Milano-Bicocca
Read the Original
This page is a summary of: An empirical study on architectural smells through a pipeline for continuous technical debt assessment, Information and Software Technology, September 2025, Elsevier,
DOI: 10.1016/j.infsof.2025.107783.
You can read the full text:
Resources
Contributors
The following have contributed to this page
