Technical Debt Management
Types of technical debt, making debt visible with registers and metrics, prioritizing by interest paid, funding models and communicating debt to business stakeholders.
Types of technical debt
| Type | Example |
|---|---|
| Code | Duplication, complexity, missing tests |
| Architecture | Shared database between services, wrong boundaries, tight coupling |
| Infrastructure | Manual servers, unsupported OS versions |
| Dependency | Outdated libraries with vulnerabilities |
| Data | Inconsistent master data, missing constraints |
| Process | Manual deployments, no observability |
Martin Fowler's technical debt quadrant: deliberate vs inadvertent × reckless vs prudent. Prudent deliberate debt ("ship now, refactor next sprint") can be rational; reckless debt is not.
Making debt visible
- Debt register: item, impact (incidents, slowness, risk), "interest" (time lost per month), remediation estimate, owner, due date.
- Metrics: lead time and change failure rate trends, incident causes, dependency age (libyears), vulnerability backlog, test coverage on critical code, hotspots (files with high churn × complexity).
quadrantChart title Prioritize debt by interest vs cost to fix x-axis Low cost to fix --> High cost to fix y-axis Low interest --> High interest quadrant-1 Plan as project quadrant-2 Fix now quadrant-3 Fix opportunistically quadrant-4 Accept / monitor Shared DB between services: [0.8, 0.9] Flaky test suite: [0.3, 0.8] Old logging library: [0.2, 0.3] Monolith module split: [0.85, 0.4]
Funding models
- Fixed capacity allocation (e.g. 15–20% of each team's capacity).
- Debt items attached to feature work in the same area ("boy scout rule").
- Dedicated modernization initiatives for large architectural debt with business cases.
- Error-budget policy: when SLOs are missed, reliability work takes priority.
Communicating with the business
Translate debt into business terms: slower delivery of features (lead time), outage risk (incidents and revenue impact), security exposure, cost. Show trends and the outcome of investments.
Interview questions and answers
How do you convince leadership to fund technical debt reduction?
Quantify the interest: e.g. "every release takes 3 extra days due to manual testing; 4 incidents last quarter cost X; this library has a critical CVE". Propose a bounded investment with measurable outcomes (lead time from 3 weeks to 1, incidents halved) and report progress.