Deferring the polish

The fastest way to miss a deadline is to gold-plate the parts nobody asked about. Especially as this gold-plating is most often accompanied with many hours of naval gazing about architecture and stuff like "what to name things"!

On this rebuild I treated scope discipline as a feature in its own right. The job was to move the site onto modern foundations without regressions, and anything that wasn't that got logged and left for later on purpose, not by accident.

That discipline left a paper trail, and the trail is the point:

• Around eight conscious steps away from like-for-like parity. Each one was a deliberate, recorded call I signed off: fixing an obvious typo rather than faithfully porting the bug, a clearer accessibility label, a couple of small content corrections.
• Twenty-plus improvements consciously deferred rather than done: an accessibility backlog, a content refresh, an analytics-consent layer, various speculative hardening. All real, all worth doing, none of them this project's job.
• Zero speculative work. Although, there were plenty of opportunities to tidy up and/or make small quality of life or feature improvements, nothing got built "while I was in there" on the off-chance it would be useful one day.

Every deferred item is evidence the build stayed in scope, not a corner cut. A "no, not now" written down is worth more than a "yes" that quietly blows the budget.

The temptation in a project like this is always to keep polishing: to harden the edge case that can't currently happen, to refactor the thing that already works. I find it more useful to name those urges, write them down, and protect the delivery instead. The deferred list isn't a list of failures. It's the record of a build that knew where its edges were.