What a 0.8s LCP is actually worth in revenue

Performance arguments usually get made with someone else’s case study. That is a weak way to spend a quarter of your engineering budget. Model it with your own numbers instead — the maths is not complicated.

Start with your own field data

Lab scores are for debugging. Field data — real users, real devices, real networks — is what correlates with revenue. Segment it by template and by device, because a 2.4 second average usually hides a 6 second mobile category page.

Model the funnel, not the page

Take conversion rate by device and by template, then model what a one-second improvement does to the step where users currently drop. If your mobile category page has a 40 percent bounce and desktop has 18, the gap is largely speed, and that gap is the size of the prize.

Where the seconds usually are

• Third-party scripts loaded synchronously in the head
• Hero images served uncompressed and unsized
• Client-side rendering that waits on an API round trip before painting anything
• Fonts that block text rendering
• A cold origin because nothing is cached at the edge

Make it stick

The hard part is not getting fast. It is staying fast after twelve sprints of feature work. Put a performance budget in CI, fail the build when a template goes over, and review field data every month like you review revenue.