Rule 37 of 38 · Chapter VI — Shipping Is Part of Engineering
A plan survives contact with production
Why this rule exists
Production is where all the assumptions in your plan meet reality: real traffic, real data shapes, real dependencies failing in ways your staging environment never showed. A plan is a hypothesis, not a guarantee, and the useful question is not whether it will need adjusting but how quickly you'll notice and how cheaply you can change course when it does.
In practice
Build plans that expect to be wrong. Ship behind flags and in small steps so course-correcting is cheap. Decide in advance what you'll watch and what reading means abort, and roll out gradually so problems surface at low volume. When production contradicts the plan, believe production. Keep a rollback ready and don't be too invested in the original design to abandon it.
When it doesn't apply
This isn't a license to skip planning or to wing risky changes. High-stakes, hard-to-reverse work still deserves careful design and rehearsal up front. The point is to hold the plan loosely once traffic hits, not to plan less.