- Cutting render-blocking CSS files from 25 to 4 brought mobile Largest Contentful Paint (LCP) on our service pages from about 4.1 s to roughly 2.6 to 2.9 s, measured in PageSpeed Insights on mobile.
- The biggest file-size win was boring: converting the blog’s images to WebP took them from 143 MB to 4.3 MB in total.
- Every change had a safety check first. Speed work breaks things quietly: unstyled flashes, missing icons, dead scripts.
- ‘Good’ LCP is 2.5 s or less, so we are close but not finished. We say so below.
Most advice on mobile speed is a list of ten tips with no numbers attached. This post is the opposite: one site (our own, built on WordPress and Elementor), four changes, the numbers we measured, the safety check for each change, and what can break. If you run a business site on WordPress and your mobile score is stuck in red, you can repeat the same order of work.
The general metrics (LCP, INP, CLS) are explained in our technical SEO checklist, so we skip the theory here. If you want a team to do this for your site, see our website speed and Core Web Vitals optimization service.
Step 0: Set up a measurement you can trust
Before touching anything, pick a handful of representative URLs, one per template: the home page, a service page, a blog post and a listing page. Run each in PageSpeed Insights on the mobile tab and write down LCP, the list of render-blocking resources and the total transfer size. You can run the same check on your own page with our free SEO audit, which uses Google PageSpeed Insights (mobile) on the page you enter.
Know which numbers you are reading. The top of a PageSpeed Insights report shows field data from real Chrome users over the previous 28 days, when your page has enough traffic to have it. The lab result below it is a single simulated mobile load. Lab numbers react the moment you deploy a fix; field numbers catch up over the following weeks. Use the lab result to judge each change, and the field data to confirm real visitors feel it.
Two habits prevent false conclusions:
- Test the same pages every time. Lab scores move a little between runs, so judge a change only when it shows up on several pages and several runs, not once.
- Bust the cache. Our host’s CDN can keep serving an old copy of a page for up to 10 minutes after a change. Add a throwaway query string such as ?test=2 to the URL when you test, or wait. Otherwise you will “measure” a fix that has not gone live yet, or miss one that has.
When you do the final measurement, test the plain URL as well, because that is what visitors and Google actually receive.
Step 1: Fewer render-blocking CSS files
What we measured: PageSpeed Insights flagged a long list of stylesheets that had to download before the browser could paint anything. Page-builder sites collect these quickly: one file per widget, per plugin, per font set. Our service pages had 25 render-blocking CSS files.
What we did: we cut that list to 4. The levers available on most page-builder sites are the same: stop loading stylesheets for widgets and plugins a page does not use, combine small files, and defer anything not needed for the first screen. This was the change that moved the headline number: mobile LCP on service pages went from about 4.1 s to roughly 2.6 to 2.9 s.
The safety check: open one page per template in a phone-width browser window with the cache disabled, and scroll the full page looking for anything that has lost its styling. Repeat on a real phone.
What can break: the classic failure is a flash of unstyled content, where the page paints plain and then jumps into its design. Deferring a stylesheet that the top of the page needs causes it. The fix is to keep the CSS for the first screen blocking (or inline) and defer only what is below the fold. A second failure is a widget that looks fine on the home page but is broken on another template, which is why we tested one URL per template, not one URL per site.
Step 2: Images, the biggest file-size win
What we measured: the blog’s images added up to 143 MB, and the blog listing grid loaded full-size images just to display small cards.
What we did: we converted the blog’s images to WebP, taking them from 143 MB to 4.3 MB in total. We also switched the blog listing grid from full-size images to 768 px wide versions. If you are unsure which sizes your theme needs, our WordPress image and layout size guide lists them.
The safety check: compare the rendered pages before and after, and filter the browser’s network tab by “Img” to spot any image still loading as a JPEG or PNG, or at its old size. Also check that social link previews on WhatsApp and LinkedIn still pick up an image.
What can break: a plugin or server rewrite that is meant to serve WebP automatically may not apply everywhere, for example when a host’s CDN serves the original file regardless. If that happens, point the page HTML at the .webp files directly. Always confirm in the network tab what actually loads, not what the plugin dashboard says it does.
Step 3: Icon fonts that carry only the icons you use
What we measured: font bytes per page. Icon fonts from page builders and their add-ons ship the whole icon set, often hundreds of glyphs, to pages that display a handful.
What we did: we subset the icon fonts, keeping only the icons the site actually uses. Font weight per page dropped by 56 to 75 percent depending on the page.
The safety check: build the subset from a full list of the icon classes in use across every template, not just the home page, then compare each template with the old version side by side.
What can break: missing icons show up as empty squares or blank spaces, usually in a menu, a footer or a rarely visited template. The safe route is to keep the original font as a fallback until you have crawled the whole site and confirmed nothing is missing.
Step 4: Removing a script nothing needed
What we did: jQuery Migrate, a compatibility layer for old jQuery code, was loading on the front end. We removed it from the front end only after the browser console showed that nothing depended on it.
The safety check: open the console (F12 in Chrome) on each template and interact with the page. When old code calls a deprecated jQuery method, jQuery Migrate logs a warning starting with “JQMIGRATE”. If you see only the line saying Migrate is installed and no deprecation warnings, it is a candidate for removal.
What can break: a slider, a popup or a form script that quietly relied on an old jQuery method will stop working. Test every interactive element: menus, forms, tabs, accordions, sliders. If you are running a theme or plugin that is no longer updated, leave this one alone.
The results at a glance
| Change | What we measured | Result | Main risk |
|---|---|---|---|
| Render-blocking CSS | PageSpeed “render-blocking resources” list | 25 files down to 4 | Flash of unstyled content |
| Blog images | Total image weight; card size on listing pages | 143 MB down to 4.3 MB; 768 px grid | Old sizes still loading |
| Icon fonts | Font bytes per page | 56 to 75 percent lighter | Missing icons |
| jQuery Migrate | Browser console warnings | Removed from the front end | Old scripts failing silently |
Overall, mobile LCP on service pages went from about 4.1 s to roughly 2.6 to 2.9 s. Google’s “good” threshold for LCP is 2.5 s or less, so some pages still sit just above it. That is the honest state of our own site.
A repeatable checklist for WordPress and Elementor owners
- Pick 4 to 6 representative URLs, one per template, and record mobile PageSpeed Insights results.
- List render-blocking CSS and JavaScript. Remove what the page does not use before you defer anything.
- Resize and convert images first. They are usually the largest transfer and the lowest risk.
- Check the actual file type and size that loads in the network tab, not what the plugin dashboard claims.
- Subset or replace icon fonts, and keep the old file until a full crawl shows no missing icons.
- Remove legacy scripts one at a time, with the console open.
- After each change, retest with a cache-busting query string, then again on the plain URL after 10 minutes.
- Open the site on a mid-range Android phone on mobile data, not just on a fast laptop connection.
- Fewer plugins help every step above. Our plugin audit guide shows how to find the ones adding weight.
- Put a recheck in your routine; speed drifts as content is added. Our website maintenance checklist has a place for it.
For the official definitions of the metrics, see Google’s Core Web Vitals overview on web.dev.
Frequently asked questions
How do I reduce website loading time on mobile?
Start with measurement in PageSpeed Insights on mobile, then fix the biggest items in this order: oversized images, render-blocking CSS and JavaScript, heavy fonts and unused scripts. Retest after every change using a cache-busting URL, and check the site on a real phone, because a fast laptop connection hides most mobile problems.
What is a good mobile load time for a business website?
Google treats a Largest Contentful Paint of 2.5 seconds or less as good for the page’s main content. Aim for that on your key landing pages first, measured in PageSpeed Insights on mobile. Perfect scores matter less than getting service and landing pages into the good range.
Why does my site score well on desktop but badly on mobile?
Mobile tests simulate a slower processor and network, so heavy images, many CSS and JavaScript files and large fonts hurt much more. Layouts that load desktop-sized images on small screens are a common cause. Fix mobile first and desktop usually follows.
Can speed changes break my WordPress site?
Yes. Deferring CSS can cause unstyled flashes, subsetting fonts can drop icons, and removing old scripts can disable sliders or forms. Make one change at a time, test every page template, keep backups and keep the original files until a full crawl shows nothing is broken.
Want your mobile scores measured and fixed safely?
We audit your key pages, make the changes one at a time and verify every template before and after.
See our speed optimization service →