How do you make a website faster?
Measure first, then cut what the page has to download before it can show anything. In my experience asset size makes the biggest difference. Aim for a Largest Contentful Paint of 2.5 seconds or less for at least 75% of visits.
A good Largest Contentful Paint is 2.5 seconds or less, and over 4 seconds is poor, measured at the 75th percentile of page loads.
From my work
Turing.tn, a site that showcases IT training, started with very poor scores. We built a custom server-side rendering app in PHP, and every Lighthouse check went to 100. Across projects, the single fix that has mattered most is asset size.
Measure before you change anything
Look at field data first: what real Chrome users got over the last 28 days. That tells you which pages are slow and which metric fails. Then run Lighthouse on those pages to find the cause. Lab runs are for debugging and testing a fix, not for deciding what matters.
Page speed optimization, in the order I check
- Images. Serve them at the size they are shown, in a modern format, and lazy-load what is below the fold. Never lazy-load the hero image.
- JavaScript. Ship less of it, split it by page, and remove libraries nobody uses. Many projects carry a whole date library for one function.
- Fonts. Fewer families and weights, and only the characters you need. No visitor has ever asked for a fifth weight.
- Rendering. Send the main content as HTML from the server, so the page does not wait for JavaScript to show it.
- Caching. A CDN for static files, and long cache headers on files that never change.
- Third-party scripts. Chat widgets, trackers and tag managers add up. Remove the ones nobody reads.
When the server is the problem
Sometimes the page is light and the server is slow. At one client the app had timeouts on Heroku. My first view was that the dynos were too small for the user base. The client wanted code fixes first, so we added caching and reworked slow PostgreSQL and Neo4j queries, and in the end we also upgraded the dynos. Both were needed.
Keep it fast after the fix
Slow pages come back, one reasonable-looking change at a time, unless something checks for them. On The Merchant Engine, Lighthouse runs in CI on every change, so a slow change gets caught before it ships.
When not to bother
If your field data already passes on the pages that bring in money, more speed work moves little. Spend the time on content or conversion instead. I say that as someone who sells speed work.
Read next
Sources
- web.dev, Largest Contentful Paint, updated 4 Sep 2025
- web.dev, why lab and field data can be different, updated 18 Jul 2022
- web.dev, Web Vitals, updated 31 Oct 2024