// PATCH №047 · PERFORMANCE
When a page is slow, the engineer on call gets the alert. The engineer on call is usually backend. The backend code is usually fine. The slowness is somewhere downstream — the analytics tag, the third-party widget, the marketing iframe — but the on-call engineer can’t fix any of those without permission they don’t have.
Trace the budget
A 1200ms time-to-interactive on a typical page breaks down roughly:
- Network + TLS: 80–150ms (geography-dependent)
- Server-side rendering: 50–250ms (your code)
- Critical CSS + initial JS: 200–400ms (your bundler’s fault)
- Third-party scripts: 300–600ms (someone else’s code)
- Render + paint: 80–150ms (browser cost)
If your team is being told to fix the 1200ms, but the third-party scripts alone account for 500ms, the budget is being negotiated in the wrong room.
What helps
Real-User Monitoring (RUM) data segmented by route + device + script-load attribution. Without that, every conversation is anecdote vs anecdote. With it, the conversation moves to the team that actually controls the third-party tag manager — which is usually marketing, who didn’t know they were spending the budget.