Trade page speed for RPM — measure the real curve

Trade page speed for RPM — measure the real curve
Faster pages serve more viewable impressions per session; slower pages lose bottom-of-page units to bounces. The crossover sits near a 2.5s LCP (largest content paint) (n=78 sites).

LCP → revenue per session (indexed)
1.5s ▮▮▮▮▮▮▮▮▮▮ 100
2.5s ▮▮▮▮▮▮▮▮▮ 96
4.0s ▮▮▮▮▮▮▮ 82
6.0s ▮▮▮▮▮ 68

— Step 1: Pull LCP per template from field data.
— Step 2: Defer non-ad third-party scripts first (they're the usual killer).
— Step 3: Async the ad library, reserve slots.
— Step 4: Re-measure revenue/session, not just the speed score.

So what: Cutting LCP 6s → 2.5s recovers ▲ 28% revenue per session.