Product.

The framework

1. 01

   Four corners, one word — and the word is doing the damage.

   Fowler's grid asks two questions of every shortcut: was it chosen or stumbled into, and was it sensible at the time. Prudent and deliberate is a real loan — you knew the cost, you took the release, and the repayment is scheduled. Reckless and deliberate is skipping design because a date got promised to someone. Reckless and inadvertent is a team that doesn't yet know what it doesn't know. Prudent and inadvertent is the honest kind: you couldn't have found the right design without building the wrong one first. Sort your known compromises into the four before you argue about any of them.

2. 02

   Where is the interest actually being charged?

   The size of the debt doesn't tell you much on its own; what matters is the tax it puts on the changes you actually make. Ugly code in a module nobody's opened in two years charges you nothing, and rewriting it is a hobby. The same ugliness in the file every feature touches is billing you on every ticket. So measure the interest rather than the principal: for each compromise, note how often the surrounding code gets changed and how much longer each change takes because of it. Repay in the order of the bill, not in the order of how much it offends you.

3. 03

   Three things, written down while the reason is still fresh.

   The debt that ruins teams is the debt nobody wrote down. When you take a shortcut deliberately, record what you did instead of the right thing, what it will cost to undo, and the trigger that forces the undoing — a date, the next time anyone touches that module, the day the customer count passes a number you name now. A compromise with an owner and a trigger is a loan. A compromise carried in somebody's head is a defect, waiting to be discovered by whoever inherits it.

4. 04

   Fund the repayment inside every cycle, never in a quarter set aside for it.

   The refactoring sprint is a promise the calendar never keeps: it gets scheduled, then sold off the moment a customer date moves. Repayment survives only as a standing share of ordinary delivery — a fixed slice of every cycle, spent on whatever is charging the most interest, defended like any other commitment. Keep it small enough that nobody needs to approve it, and then it isn't a negotiation. And take the same honesty into estimates: a feature landing in debt-heavy code costs more, and saying so in planning is how the interest finally becomes visible to the people authorising the borrowing.