Sustainably Establishing Technical Debt Management in Practice

Abstract

Context: Software systems are subject to continuous change due to evolving requirements and contextual shifts, particularly in agile environments, where changing requirements are explicitly welcome. Implementing a maintainable software system often requires more effort than delivering a quick solution. As a consequence, suboptimal technical solutions are introduced that are beneficial in the short term but hinder long-term development and maintenance. The Technical Debt (TD) metaphor was introduced to communicate such trade-offs, comparing difficult-to-maintain implementations to financial debt. Following this metaphor, interest in the form of additional effort to make changes must be paid if the debt is not repaid, i.e., if the suboptimal solution is not improved. Over time, research has identified various TD types, causes and consequences of TD, and TD Management (TDM) activities (i.e., identification, documentation, measurement, prioritization, repayment, monitoring, and prevention).

Despite this extensive body of research, dedicated TDM processes remain uncommon in practice. Research on holistic approaches covering multiple TD types and activities is rare. Socio-technical aspects such as communication and awareness are often neglected, and research findings are insufficiently transferred to industrial contexts.

Objectives: The research goal of this thesis is to understand the reasons behind the poor adoption of TDM processes in practice and to identify ways to make research findings on TDM more applicable. Existing approaches often focus on single TD activities or specific TD types, predominantly code debt, and rarely address the socio-technical challenges of communication, cultural change, or awareness of TD in decision-making situations. Furthermore, concrete, practical guidance on establishing and integrating a TDM process into agile work environments is limited. To this end, this thesis focuses on establishing a holistic and feasible TDM process, concentrating on its socio-technical aspects, i.e., TD communication and TD awareness, as well as TD prevention. In the end, the thesis shall provide best practices as a low-level starting point for a general TDM best practice library.

Research Design and Methods: The research of this thesis is structured into three steps. Step I analyzes the research area and identifies research gaps through literature reviews, case studies, interviews with IT managers and software architects, surveys, and exploratory analyses. These studies focus on communication structures, changes in awareness due to a TDM process, prevention strategies, and existing TDM practices.

Based on these insights, Step II develops artifacts to address identified research gaps, including a game to foster discussions about TD and a five-step workshop concept to establish a TDM process. The idea behind the workshops is to present the practitioners with a few TD activities per workshop, introduce approaches identified by research for these activities, and let the practitioners decide which approaches to adopt. Additionally, practitioners were encouraged to develop new, useful approaches for their specific situations. During this step, the TD-SAGAT (TD-Situation Awareness Global Assessment Technique) method was developed. TD-SAGAT is an adaptation of a method from psychology for the TD context, designed to measure changes in TD awareness during TDM establishment.

Step III evaluates the five-step workshop concept through action research studies in three companies. The measures determined during the workshops in the action-planning phase are applied between workshops in the action-taking phase. The evaluation and learning phases during the workshops allow for changes to the process, following the agile "inspect&adapt" approach. Data collection comprises surveys, meeting observations, TD-SAGAT, and backlog analyses. In this study, the adopted approaches and the observed challenges are further outcomes of the research, i.e., part of the collected data. Comparing results across cases enables the identification of best practices and recurring challenges.

Results & Contributions: The results show that socio-technical aspects are critical to successfully establishing a TDM process, and a self-empowered approach is superior to pre-defined processes for agile working teams.

Communication about TD improves when causes and consequences are determined, discussed, and visualized, for example, using the V4CTD model (Visibility, Cycles & Chains of Causes & Consequences of TD). Communication structures need to be ensured and should account for psychological effects, e.g., factors that hinder or facilitate communication. The TD Game allows for discussion of this critical topic in an emulated, and thus psychologically safe, environment.

The TD-SAGAT method defines the term "TD awareness" and enables its measurement during daily work. The action research studies indicate that establishing a TDM process sustainably increases both subjective and observable indicators of TD awareness when attributes are added to the backlog issue types that remind practitioners to discuss certain aspects of a decision.

TD prevention is closely linked to TD communication and TD awareness. Debiasing architectural decision-making by explicitly considering alternatives, drawbacks, and risks of potential solutions contributes to reducing unintentional TD incurrence. These factors are among the attributes that should be added as reminders to all backlog issues. Recording intentionally incurred TD items and repaying them as part of a project might prevent TD by creating a feedback loop.

The five-step workshop concept provides a feasible, transferable approach to establishing a holistic TDM process. Across three companies, teams integrated all TD activities into their existing agile workflows. The cross-case comparison reveals common elements that can serve as \textit{best practices} and are outlined in the TDM Guide, a whitepaper for practitioners.

Limitations: The findings are limited to teams that are dedicated to changing their processes, willing to invest time in this, and supported by their management. The workshop concept is specifically designed for agile teams who are authorized to adapt their processes. A key constraint seems to be the guidance from an expert familiar with the TDM topic during the workshops.

Conclusions and Future Work: The findings indicate that TDM can be sustainably established when socio-technical aspects are explicitly addressed and when the process is tailored to the team’s context. The workshop-based establishment method proved feasible across multiple companies and led to improved communication, increased awareness, and more systematic handling of TD.

Future work includes further evaluating the TDM guide, particularly the need for external guidance, streamlining the five-step workshop concept, scaling TDM to the company level, and developing online course formats to support broader adoption.