The Hidden Cost of Technical Debt Nobody Puts in the Backlog
Technical debt has a reputation problem: it sounds like a line item you can schedule, when most of the time it's a tax that's already being paid, quietly, by whoever touches that part of the codebase next.
Debt compounds in places nobody's watching
The modules that accumulate the most debt are rarely the ones anyone's actively afraid of. They're the ones that "just work," so nobody's opened a profiler or a test file in there for a year. Debt in a module everyone fears gets paid down eventually, out of self-preservation. Debt in a module everyone trusts just keeps compounding.
The tax shows up as velocity, not bugs
This is the part that makes debt hard to sell to a non-engineering stakeholder: a debt-heavy module doesn't usually produce more incidents. It produces a slower team. A feature that should take two days takes six, because half the work is figuring out which of three half-finished abstractions is safe to extend. Nobody files a ticket for "this took three days longer than it should have" — it just disappears into the estimate.
A quick way to spot it
Ask an engineer to estimate a small change in a given file. If the estimate is dominated by "figuring out what's safe to touch" rather than "writing the change," that file is carrying debt whether or not it's ever caused an outage.
How we triage debt without stalling the roadmap
- Tag debt at the moment you notice it, in the code, not in a separate backlog nobody re-reads
- Pay it down opportunistically, inside the feature work that already touches that file — never as a stand-alone "refactor sprint" that competes with real deadlines for buy-in
- Reserve dedicated refactor time only for debt that's actively blocking a specific upcoming feature, not debt that's merely unpleasant
A rule of thumb that's served us well
If a piece of debt hasn't slowed anyone down in six months, it's not debt — it's just code you don't like the shape of. Leave it. The debt worth paying down is the kind that's already costing someone time this week.
