PHASE THREE — LAUNCH.

//Before the switch#

Real-user monitoring goes in before launch, not after, so there's a baseline to compare against when someone claims the new site feels slower. Performance budgets go into CI as a failing check. Error tracking is wired up with enough context to be actionable, which mostly means capturing what the user was doing rather than only what the stack trace says.

//Handover#

A recorded session walking through the architecture, the decisions and the known rough edges, plus written documentation of the same. The most valuable section is always the list of things I chose not to do and why — that's the knowledge which otherwise evaporates the moment I stop answering messages, and it's what stops the next engineer from cheerfully reintroducing a problem I already solved.

I also hand over the assumption log from discovery, annotated with which assumptions held and which didn't. It's an uncomfortable document to produce and it's genuinely useful next time either of us estimates something similar.

//The two weeks after#

Included in the phase. Real traffic finds things no test plan does — the device nobody owned, the connection pattern nobody simulated, the content editor who does something entirely reasonable that I never considered. I'd rather be there for it than be re-engaged in a hurry a month later.