Clients ask for “a 90 on PageSpeed.” I translate that into something they can feel: the hero appears quickly, buttons do not jump, and the form is usable on 4G. Lab scores help. Field experience is the brief.
Hosting and DNS come first. No amount of lazy-load will save a tired shared server or a DNS zone that still points at the old host after a migration. I handle cPanel, DNS, SSL, and caching in that order.
Theme and plugin diet is second. Elementor Pro is enough for most of my builds. If a plugin adds a marketing popup, an extra slider, and three tracking pixels, it is a candidate for removal. I look at the network panel, not the plugin marketing page.
Images are still the usual crime. I export hero images at the size they display, compress them, and set width and height so the layout does not shift. Decorative videos do not autoplay on mobile. Background images are not 4000 pixels wide “for quality.”
Fonts: two weights of Poppins is a choice. Twelve is a delay. I subset when I can and I never load italic 100 through 900 “just in case.”
Third-party scripts get a meeting. Pixel, Analytics, Search Console, a chat widget, a heat-map tool — each one has a cost. I load what the campaign needs and I delay the rest. Consent and performance are both part of the same conversation.
On Elementor specifically I avoid excess nested containers, unused widgets left on the canvas, and CSS print that ships on every URL. I regenerate CSS after large edits. I test the real templates: home, inner, product, landing — not only the homepage.
After launch I watch Search Console and Analytics for the pages ads actually send people to. Optimising a blog nobody visits while the landing page crawls is a hobby, not a service.
This is the work behind “performance optimization” on my CV. It is not a badge. It is a checklist I still run on every site.
