Towards Measuring the Impact of Technical Debt on Lead Time: An Industrial Case Study
Bhuwan Paudel, Javier Gonzalez-Huerta, Ehsan Zabardast, Eriks Klotins
TL;DR
The paper investigates whether technical debt density ($TDD$) measured by SonarQube influences lead time for resolving Jira issues in six industrial components within a FinTech context. Using monthly, component-level data, the authors compare linear, quadratic, and cubic models and find mixed results, with cubic fits explaining up to about $0.41$ of the lead-time variance in some components. The study demonstrates that TD alone does not fully account for lead-time variation and highlights potential confounders such as change size, complexity, and team ownership. The findings motivate broader, multi-variable investigations and finer-grained analyses (e.g., file-level) to better understand the TD–lead-time relationship in industry settings.
Abstract
Background: Software companies must balance fast value delivery with quality, a trade-off that can introduce technical debt and potentially waste developers' time. As software systems evolve, technical debt tends to increase. However, estimating its impact on lead time still requires more empirical and experimental evidence. Objective: We conduct an empirical study investigating whether technical debt impacts lead time in resolving Jira issues. Furthermore, our aim is to measure the extent to which variance in lead time is explainable by the technical debt. Method: We conducted an industrial case study to examine the relationship in six components, each of which was analyzed individually. Technical debt was measured using SonarQube and normalized with the component's size, while lead time to resolve Jira issues was collected directly from Jira. Results: We found a set of mixed results. Technical debt had a moderate positive impact on lead time in two components, while we did not see a meaningful impact on two others. A moderate negative impact was found in the remaining two components. Conclusion: The findings show that technical debt alone can not explain all the variance in lead time, which ranges from 5% up to 41% across components. So, there should be some other variables (e.g., size of the changes made, complexity, number of teams involved, component ownership) impacting lead time, or it might have a residual effect that might manifest later on. Further investigation into those confounding variables is essential.