Guide
Managing technical debt without stopping the roadmap
Treat technical debt as a cost line, not a moral failing: quantify what each item costs you per month in slowed delivery, incidents or risk, then pay down the expensive ones continuously. A steady 10–20% of capacity beats an annual cleanup sprint that never gets scheduled.
6 min read · Updated 2026-08-06
What is technical debt, really?
Technical debt is the ongoing interest you pay for a past decision that no longer fits — deliberate shortcuts, decisions overtaken by scale, or code that outlived its assumptions. Not all of it is worth repaying; debt in a part of the system nobody touches costs nothing.
How do you quantify technical debt?
Convert each item into a monthly cost: engineer-hours lost, incidents caused, or risk exposure. Once debt has a number attached, prioritising it stops being an argument between engineering and product.
| Debt item | Interest paid monthly | Repay? |
|---|---|---|
| Flaky test suite | ~20 engineer-hours of reruns and mistrust | Yes, immediately |
| Manual deploy process | ~10 hours + release-day risk | Yes |
| Legacy admin tool nobody touches | 0 | No — leave it |
| Unversioned public API | Blocks every change customers can see | Yes, before more customers |
| Duplicated onboarding logic | Two places to fix each bug | When it next causes a bug |
How much capacity should go to paying down debt?
Ten to twenty percent of engineering capacity, continuously, attached to the features that touch the same code. Debt paid down alongside feature work is cheap; debt paid down in a dedicated project competes with revenue and loses.
How do you explain technical debt to a CEO or board?
Speak in delivery speed and risk, never in code quality. 'This change costs us three extra weeks per quarter and caused two of our last four incidents' gets funded; 'the code is messy' does not.
- Tie each item to a business consequence: slower launches, incidents, security exposure, or hiring difficulty.
- Offer options with costs, not a single demand.
- Show the trend after you fix something — earned credibility funds the next one.
When is a rewrite the right answer?
Rarely, and almost never all at once. Prefer strangling the old system: put a stable interface in front of it and replace it component by component while it keeps serving customers.
- Define the interface the rest of the system will depend on.
- Route traffic through it while the old implementation still serves.
- Replace one component at a time, behind feature flags.
- Delete the old path as soon as it is unused — an unremoved old path is new debt.
Frequently asked questions
- Is technical debt always bad?
- No. Deliberate, documented shortcuts that buy you speed to a validated product are good business. The problem is undocumented debt that nobody chose and nobody is tracking.
- Should you schedule dedicated tech debt sprints?
- Occasionally, for large, indivisible items. As a default policy they get cancelled the first time a customer commitment slips, which is why continuous allocation works better.
- Who decides what technical debt gets paid down?
- Engineering proposes with costs attached; the CTO or VP of Engineering prioritises alongside the roadmap. It should be a portfolio decision, not an individual engineer's side project.
From the podcast
Founders who lived this
Code Story episodes where founders work through the same problem in their own words.
Season 12
S12 Bonus: Rickard's Deterministic Return: Converting Scattered Data into Autonomous, Production-Grade Apps with Rickard Hansson, Founder & CEO of Gainable
Using data and moving AI to the most optimal spot in the SDLC
Aug 6, 2026 · 24 min
Season 12 · Episode 30
S12 E30: Wastewater Guardians: Automating Biology to Protect Clean Water with Virginia Szepietowski, Co-Founder of Nyad AI
Building delightful AI for the Guardians of our Water System
Aug 4, 2026 · 28 min
Season 12
S12 Bonus: The App-Aware Illusion: Why Infinite Compute Fails Without Underlying Infrastructure Accountability and the Case for "Boring" IT with Richard Luna, President & Founder of Protected Harbor
Infrastructure that is optimized for your specific SaaS application
Jul 30, 2026 · 32 min
Season 12 · Episode 28
S12 E28: The AI Throughput Illusion: Why Splurging on Expensive Models Fails to Ship Code and How to Measure Real Engineering Output with Emilie Schario, Co-Founder & Head of Product & Engineering at Kilo Code
The Future of AI is multi-model via the best coding harness
Jul 21, 2026 · 24 min
Keep reading
Related guides
Engineering metrics that matter (and the ones that don't)
Which engineering metrics to track at a startup: DORA metrics, quality and reliability signals, what they tell you, and why measuring individual output backfires.
How to choose a tech stack for a startup
A decision framework for choosing a startup tech stack: what actually matters, which choices are reversible, and how to avoid the rewrite that eats your Series A.
Technical due diligence checklist for startups
A technical due diligence checklist for founders and investors: what reviewers examine across architecture, security, team and process, and how to prepare before a raise.