How to measure tech debt: four methods in five minutes
You cannot fix what you cannot measure. These four methods range from "ask the team" (one afternoon) to "instrument the codebase" (one quarter). Pick one to start; do not try all four at once.
Methods at a glance
Method 01
Developer survey
Method 02
Static analysis
Method 03
Cycle time analysis
Method 04
Incident post-mortems
Method 01 / The fastest
Developer survey
Five questions, anonymous, 10 minutes. Your engineers know exactly what slows them down. Run quarterly to track trend.
The Five Questions
- What % of your time last sprint went to maintenance, debt, or workarounds?
- Name one part of the codebase you fear changing.
- How many minutes does our slowest CI build take?
- How many incidents in the last quarter trace to known bad code?
- If you could clean up one thing this quarter, what would it be?
Interpreting Q1
| Healthy | 0-15% |
| Manageable | 15-30% |
| Painful | 30-50% |
| Critical | 50%+ |
Method 02 / Code-level
Static analysis
Tools score your codebase against a quality model and produce a remediation cost in person-hours. Multiply by your fully-loaded hourly rate to get dollars.
List prices verified June 2026 (Sonar, Codacy publish per-developer rates; Code Climate Quality is per-committer, ~$50-80/mo list before volume discounts; Stepsize and CAST quote on request). Vendor-neutral list. Pick one to pilot, not all of them.
Conversion formula: remediation_hours x fully_loaded_hourly_rate = remediation cost. A SonarQube score of 4,200 hours at $90/hour fully-loaded is $378K.
Method 03 / Velocity proxy
Cycle time analysis
Measure delivery friction directly. The DORA four metrics correlate strongly with debt level. Pull from Jira, Linear, or GitHub.
Cycle time
Time from first commit to production. Elite < 1 day, low > 1 month.
Source: GitHub PRs, Jira
Deployment frequency
Elite multiple per day, low less than once per month.
Source: CI/CD logs
Change failure rate
% of changes causing incidents. Elite 0-15%, low 46-60%.
Source: Incident tooling
MTTR
Mean time to restore. Elite < 1 hour, low > 1 week.
Source: Incident tooling
Method 04 / Forensic
Incident post-mortems
Tag every incident root cause against a small taxonomy. Architectural, dependency, test coverage, configuration, human error. After 90 days you will have the share of incidents attributable to debt.
Conversion formula: incidents/month x avg_incident_cost x debt_attributable_share = annual debt-driven incident cost. 4 incidents/month x $25K avg x 50% (an example share, not a benchmark; measure your own from the 90-day tagging above) = $600K/year.
On the quote
Where does “you cannot fix what you cannot measure” come from?
It is a management adage with no single verified author, and the people most often credited either never said it or argued the opposite. Treat it as a useful rule of thumb, not a sourced citation.
- Peter Drucker is the usual attribution (often as “what gets measured gets managed”), but there is no record of him writing or saying it.
- W. Edwards Deming is also credited, yet he called the idea a costly myth: “It is wrong to suppose that if you can’t measure it, you can’t manage it — a costly myth” (The New Economics, 1993). His point: much of what matters most is unmeasurable, and you must manage it anyway.
- The verifiable root of the measurement-as-knowledge idea is Lord Kelvin (William Thomson), 1883: “When you can measure what you are speaking about, and express it in numbers, you know something about it; but when you cannot measure it… your knowledge is of a meagre and unsatisfactory kind.”
For tech debt: you do not need a perfect number to act. A rough, repeatable measure — survey % or cycle time tracked quarterly — beats a vibe, and improves as you refine it.
Recommended starting point
Pick two, run quarterly, plug into the calculator
For most teams: survey + cycle time. The survey gives you the human signal in an afternoon. Cycle time gives you the operational signal from data you already have. Run both quarterly, plug the drag % into techdebtcalculator.com for dollar figures, present the trend in your QBR.