How to Build the Business Case for Paying Down Technical Debt — Research Notes
Companion post for Tech Manager Lab — Part 5 of the vendor series. The four-part structure below is the meeting-ready version.
| Part | What you do | The failure it fixes |
|---|---|---|
| 1. Quantify | One business number per debt item: delivery delay, incident cost, onboarding time, rate of change | “The code is brittle” — an engineering claim in a business room |
| 2. Attach to a KPI | Map the debt to a number someone already owns: on-time release rate, MTTR, cost per seat, unpatched exposure | Asking leadership to care about your problem instead of showing theirs |
| 3. Shrink the ask | One sprint, two weeks, named owner, defined exit — not “three months to refactor” | A request with no end and no bound, read as a vacation from feature work |
| 4. Attach a deadline | Tie it to a moment that already has one: audit, migration, go-live, renewal | A long-term cost vs a short-term deadline — the cost always loses |
Worked example (INR, illustrative — your rates decide): a recurring integration failure at ~₹45,000/month in incident response and missed delivery is ₹5,40,000 a year. A feature slipping from two weeks to eight, at ~₹4,000/day loaded across four engineers, is ~₹1,00,000 lost capacity per cycle. One quarterly severity-one outage often runs ₹40,000–₹60,000 in loaded labour alone.
Why the usual argument fails: a budget meeting does not vote on code quality. It votes on revenue protected, cost avoided, risk reduced, deadline met. If your case does not map to one of those four, it is outvoted before you speak. Every debt item gets exactly one business number — the abstract version never changed a vote, the number did.
The one-page case, in order: the number, the KPI it hits, the bounded ask, the date it must land before. Four lines. That page is what survives the meeting.
Sources: cost figures are working assumptions (illustrative) stated in the episode — verify against your own numbers. Series: Part 4 — asset lifecycle, Part 3 — TCO, Part 2 — helpdesk metrics. — @techmanagerlab