Win Support to Pay Down Technical Debt in Software Roadmaps

When deadlines are tight, I separate technical debt by blast radius, not by how ugly the code looks. If the debt can break a release, slow the team inside the same deadline, create security or data risk, or make a customer-facing workflow unstable, we handle it now. If it is mainly architectural neatness, naming, duplication in a quiet part of the system, or a better abstraction we simply wish we had, we defer it.

The important thing is to make this discussion concrete. "We need refactoring" is easy for stakeholders to reject. "If we skip this migration script cleanup, QA will have to retest every order state manually before release" is a business conversation. We describe debt in terms of delivery risk, support cost, onboarding cost, or change cost. That changes the tone from engineers asking for extra time to the team managing known risk.

The planning habit that helped us most is a debt register tied to product areas, not a separate graveyard of engineering wishes. Every debt item has three fields: what decision created it, what failure it may cause, and what event should trigger repayment. For example, "fix before we add the next payment flow" is much easier to accept than "someday we should rewrite payments."

Our funding rule is simple: if a feature creates debt to meet a deadline, the repayment task is added to the next planning conversation before new scope in that same area. We don't promise to fix everything immediately, but we do not let temporary decisions become invisible. Stakeholders accept debt paydown when they can see that it protects the next delivery, not just the codebase.

The debt I prioritize under deadline pressure is whatever is actively slowing delivery today, not whatever looks worst in the codebase. On a payments modernization project where a full platform rebuild would have taken 18-24 months the client didn't have, we used a strangler-fig pattern instead — building new compliance and settlement modules alongside the legacy monolith rather than replacing it outright, touching only the parts of the old system that were actually blocking the new work. Everything else stayed exactly as ugly as it was, because rewriting it wouldn't have moved the delivery date.

The planning habit that actually got stakeholders to accept debt paydown: we stopped presenting it as "cleanup" and started tying it to a specific, already-happened cost — an incident, a delayed release, a spike in support tickets. Debt that's caused a measurable problem goes into the next sprint automatically; debt that's just aesthetically bad goes on a backlog nobody's obligated to fund. Framing it around incidents rather than code quality is what got budget approved, because it's a business argument, not an engineering opinion.

For a long time, debt work only happened when something broke. We'd say we'd "get to it later," but later never came. There was always another deadline.

What fixed it was a simple habit, I stopped asking for separate time to clean up code. Instead, I attached the cleanup to the feature work that was already touching that same code.

The rule is easy. If we're already working in an area, and there's messy code right there making the job harder, we fix it as part of the feature. Not a separate ticket. Just part of doing the feature properly.

That's also what made stakeholders okay with it. Asking for a whole sprint "just for cleanup" always sounded like engineers polishing things while the deadline slipped. People pushed back, and fair enough. But saying "this feature will take a little longer because the code here is fragile, and fixing it now means the next few features ship faster", that's a tradeoff a product person can actually follow. It helps delivery instead of fighting it.

The other half is knowing what to leave alone. Not all debt is worth fixing. If some messy code is stable and nobody's going near it, I skip it. I only spend time on debt that's slowing down the work we're already doing.

The payoff was fewer surprises. The areas we cleaned as we went stopped causing last-minute fires, so the next deadlines got easier to hit. And that's the part stakeholders care about most.

I work in AI infrastructure, where deadlines are relentless because the field moves so fast. We're constantly changing the ground underneath us, so we pile up debt as we go, making point-in-time decisions we later realize weren't the right long-term call. A common example: we build a one-off implementation to unblock a specific product or customer, then realize we should have invested in a centralized, common platform that most solutions could share. So I don't treat all debt as one bucket. I weigh blast radius: debt that threatens reliability, security, or our ability to ship the next features gets fixed now, because deferring it only makes the next deadline worse. The harder call is one-off versus platform. You don't want to stop innovating or trying new technologies, but you have to be honest about how much energy you're spending on one-offs that become tomorrow's debt versus a common platform that's stable and reusable long term. For each item I ask: if we don't fix this now, does it get more expensive or more dangerous next quarter? If yes, it moves up; if not, it's deferred, on the record.

On the funding side: as a technical program manager, my job is to bring visibility into every decision, and that's how I get debt paid down. If the VP or leader approving a plan doesn't understand the cost of not paying down a piece of debt, they're deciding without the full picture. So I don't argue about debt abstractly. I put the complete problem statement in front of them, including the concrete consequences of not doing it, and let them and the steering committee make the call. Leadership is usually very aware of the top programs the company is investing in and the value they bring, but far less aware of the debt building underneath. Once the consequences are visible and it's their sign-off on the line, they start giving debt equal weight, because nobody wants to have approved ignoring a risk that later broke delivery. Alongside that, the funding rule that stuck was reserving a small, fixed slice of every cycle for debt paydown, treated as non-negotiable so it never competes head-to-head with features. Predictable and boring beats heroic cleanups.