A 2017 survey of 258 software practitioners (Besker, Martini, and Bosch, presented at the IEEE International Conference on Software Maintenance and Evolution, ICSME) estimated that, on average, 36% of development time is wasted paying the 'interest' on technical debt (TD) — the recurring extra effort caused by accumulated internal-quality shortcuts — and reported perceived role differences and a non-monotonic relation between system age and wasted time. All of these figures are practitioner estimates, not measurements. The open problem is to quantify TD interest through instrumented or observational measurement (e.g., version-control, issue-tracker, or integrated-development-environment (IDE) telemetry, or controlled observation), to determine whether perception-based estimates — 36% in the original cross-sectional survey, revised to 23% in the authors' longitudinal follow-up — are validated, adjusted, or refuted by objective data, and to establish what actually drives the measured interest (TD types, development activities, system age, roles). No published study has performed this validation; the few objective studies of related observables give mixed signals, so the perception-reality correspondence for TD cost remains open.

Background

Technical debt (TD) is a metaphor, coined by Ward Cunningham in 1992, for the future cost that a software organization accepts when it ships a quick, internally imperfect solution instead of a cleaner one: as with a financial loan, the 'principal' is the cost that would be paid today to remediate the shortcut (e.g., refactoring a tangled module), while the 'interest' is the recurring extra effort that the accumulated imperfections impose on later work — extra comprehension, testing, debugging, and rework that clean code would not require. TD interest is the quantity that matters for capacity planning and for deciding when remediation pays off, because it is the ongoing tax on every future change.

Empirically pinning down that tax has proven hard. The quantitative anchor of this problem is the 2017 study by Besker, Martini, and Bosch ('The Pricey Bill of Technical Debt: When and by Whom will it be Paid?', presented at the IEEE International Conference on Software Maintenance and Evolution (ICSME), DOI 10.1109/icsme.2017.42), which surveyed 258 practitioners — recruited mainly from partner companies in Scandinavia and from LinkedIn software-engineering groups, with follow-up group interviews of 32 practitioners from seven companies — about how much development time TD wastes and where that time goes. Practitioners estimated on average 36% (standard deviation 17.8%) of overall development time is wasted paying TD interest; Software Architects and Experts reported the highest averages (about 41%) and Developers lower (about 35%), though the differences across roles were not statistically significant; the estimated wasted time varied across system-age intervals (about 33.8% for systems younger than 2 years to about 40.5% for systems older than 20 years) with no significant linear correlation between age and wasted time but statistically non-independent age and wasted-time distributions; and the activity practitioners reported spending the most extra TD-driven time on was 'understanding and/or measuring the issues'. A companion longitudinal study by the same authors (presented at the International Conference on Technical Debt (TechDebt) 2018 and extended in the Journal of Systems and Software in 2019) surveyed 43 developers repeatedly and reported a lower average, 23% of working time wasted due to TD, with additional testing as the most time-consuming extra activity.

All of these numbers are self-reported perceptions. The authors themselves flag the core limitation: the results rest on practitioner estimates rather than instrumented measurements, so it is unknown whether perceived TD cost corresponds to objectively wasted effort, whether the perceived role and age patterns reflect real cost structure or perception bias, and whether the magnitude is stable across designs (36% cross-sectionally versus 23% longitudinally) and populations (the sample was Scandinavia-heavy). Objective measurement of TD interest is nontrivial: prior work had proposed monitoring developer activity related to code comprehension as an interest proxy (Singh, Snipes, and Kraft, 2014), and static-analysis tools express 'debt' in remediation-time units, but none of these has been validated against actual wasted effort or against practitioner perceptions. Meanwhile the first objective studies of adjacent observables disagree — a mining study of 39 proprietary codebases found issue resolution in low-quality code takes 124% more development time, while an industrial case study found technical debt explained little of the variation in Jira lead times — underscoring that the perception-reality correspondence for TD cost is an open empirical question rather than a settled fact.

Problem Statement

Quantify the interest on technical debt — the portion of software development time actually consumed by the consequences of accumulated TD (extra comprehension, testing, debugging, tracking, refactoring, and rework that would not occur in the absence of the debt) — by instrumented or observational measurement rather than practitioner self-report, and thereby determine whether the perception-based estimates are validated, adjusted, or refuted.

Concretely, an answer must establish, on objective data (for example version-control, issue-tracker, or integrated-development-environment (IDE) telemetry, or controlled observation of development work):

  1. The measured magnitude of TD interest, expressed as the fraction of development time it consumes (to be compared with the perceived 36% of the 2017 survey by Besker, Martini, and Bosch at the International Conference on Software Maintenance and Evolution (ICSME) and the 23% of the same authors' longitudinal follow-up), together with an account of how TD-induced extra time is attributed and separated from task-inherent complexity and other confounders. The magnitude is estimated from recorded development activity rather than self-report, with TD-induced extra time isolated from task-inherent complexity, and measured for a defined population of development work.
  2. The composition of the measured interest: which TD types and which development activities actually consume the objectively wasted time, to be compared with the perceived rankings (e.g., 'understanding and/or measuring the issues' perceived as dominant).
  3. Whether the perceived patterns hold objectively: whether the measured interest differs across practitioner roles and across system ages in the way practitioners perceive (including whether the perceived non-monotonic age–wasted-time relation is real or a perceptual artifact).

The authors of the 2017 survey state that their data are 'not based on measured or observed data but on estimations made by the respondents', that their findings are correlational rather than causal, and that they plan future work 'to include physical measurements and observations'. No restriction to particular TD types, system classes, development contexts, or measurement instruments is intended.

References

  1. Ward Cunningham, 'The WyCash Portfolio Management System', ACM SIGPLAN OOPS Messenger, 1992. DOI: 10.1145/157710.157715. https://doi.org/10.1145/157710.157715
  2. Terese Besker, Antonio Martini, Jan Bosch, 'The Pricey Bill of Technical Debt: When and by Whom will it be Paid?', 2017 IEEE International Conference on Software Maintenance and Evolution (ICSME), 2017, pp. 13–23. DOI: 10.1109/ICSME.2017.42. https://doi.org/10.1109/icsme.2017.42
  3. Terese Besker, Antonio Martini, Jan Bosch, 'Technical Debt Cripples Software Developer Productivity: A Longitudinal Study on Developers' Daily Software Development Work', Proceedings of the 2018 International Conference on Technical Debt, 2018. DOI: 10.1145/3194164.3194178. https://doi.org/10.1145/3194164.3194178
  4. Terese Besker, Antonio Martini, Jan Bosch, 'Software developer productivity loss due to technical debt — A replication and extension study examining developers' development work', Journal of Systems and Software, 2019. DOI: 10.1016/j.jss.2019.06.004. https://doi.org/10.1016/j.jss.2019.06.004
  5. Vallary Singh, Will Snipes, Nicholas A. Kraft, 'A Framework for Estimating Interest on Technical Debt by Monitoring Developer Activity Related to Code Comprehension', 2014 Sixth International Workshop on Managing Technical Debt, 2014. DOI: 10.1109/MTD.2014.16. https://doi.org/10.1109/mtd.2014.16
  6. Nicolli Rios, Rodrigo Oliveira Spínola, Manoel Mendonça, Carolyn Seaman, 'The practitioners' point of view on the concept of technical debt and its causes and consequences: a design for a global family of industrial surveys and its first results from Brazil', Empirical Software Engineering, 2020. DOI: 10.1007/s10664-020-09832-9. https://doi.org/10.1007/s10664-020-09832-9
  7. Robert Ramač, Vladimir Mandić, Nebojša Taušan, Nicolli Rios, Sávio Freire, Boris Pérez, Camilo Castellanos, Darío Correal, Alexia Pacheco, Gustavo Lopez, Clemente Izurieta, Carolyn Seaman, Rodrigo Spinola, 'Prevalence, common causes and effects of technical debt: Results from a family of surveys with the IT industry', Journal of Systems and Software, 2022. DOI: 10.1016/j.jss.2021.111114. https://doi.org/10.1016/j.jss.2021.111114
  8. Abdullah Aldaeej, Mohammad Alshayeb, 'Familiarity, Common Causes and Effects of Technical Debt: A Replicated Study in the Saudi Software Industry', Arabian Journal for Science and Engineering, 2024. DOI: 10.1007/s13369-023-08596-w. https://doi.org/10.1007/s13369-023-08596-w
  9. Areti Ampatzoglou, Nikolaos Mittas, Angeliki-Agathi Tsintzira, Apostolos Ampatzoglou, Elvira-Maria Arvanitou, Alexander Chatzigeorgiou, Paris Avgeriou, Lefteris Angelis, 'Exploring the Relation between Technical Debt Principal and Interest: An Empirical Approach', Information and Software Technology, 2020. DOI: 10.1016/j.infsof.2020.106391. https://doi.org/10.1016/j.infsof.2020.106391
  10. Saulo S. de Toledo, Antonio Martini, Dag I.K. Sjøberg, 'Identifying architectural technical debt, principal, and interest in microservices: A multiple-case study', Journal of Systems and Software, 2021. DOI: 10.1016/j.jss.2021.110968. https://doi.org/10.1016/j.jss.2021.110968
  11. Lan Cheng, Emerson Murphy-Hill, Mark Canning, Ciera Jaspan, Collin Green, Andrea Knight, Nan Zhang, Elizabeth Kammer, 'What improves developer productivity at Google? Code quality', Proceedings of the 30th ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering, 2022, pp. 1302–1313. DOI: 10.1145/3540250.3558940. https://doi.org/10.1145/3540250.3558940
  12. Adam Tornhill, Markus Borg, 'Code Red: The Business Impact of Code Quality — A Quantitative Study of 39 Proprietary Production Codebases', Proceedings of the International Conference on Technical Debt, 2022, pp. 11–20. DOI: 10.1145/3524843.3528091. https://doi.org/10.1145/3524843.3528091
  13. Bhuwan Paudel, Javier Gonzalez-Huerta, Ehsan Zabardast, Eriks Klotins, 'Exploring the Relationship Between Technical Debt and Lead Time: An Industrial Case Study', 2025 IEEE International Conference on Software Analysis, Evolution and Reengineering (SANER), 2025. DOI: 10.1109/SANER64311.2025.00071. https://doi.org/10.1109/SANER64311.2025.00071
  14. Judith Perera, Ewan Tempero, Yu-Cheng Tu, Kelly Blincoe, 'A Systematic Mapping Study Exploring Quantification Approaches to Code, Design, and Architecture Technical Debt', ACM Transactions on Software Engineering and Methodology, 2024. DOI: 10.1145/3675393. https://doi.org/10.1145/3675393
  15. Terese Besker, 'Technical Debt: An empirical investigation of its harmfulness and on management strategies in industry', Doctoral thesis, Chalmers University of Technology, 2020 (open access; Chapter 11 reprints the ICSME 2017 paper in full). https://research.chalmers.se/en/publication/518318