Insight · WordPress & performance

Speeding up WordPress: from 38 to 96 on mobile

A slow WordPress website affects more than visitors. Editing text, saving a page or logging in can take longer than necessary too. A caching plugin helps, but it does not automatically fix every cause. For a B2B marketing agency, we investigated where the delays came from and what improving them achieved on shared hosting.

Mitchell van Honk Mitchell van Honk · founder of Mantella
Published 16 September 2026 · 8 min read · measured 16 September 2026
Mobile PageSpeed Performance
38
Before
96
After
Largest Contentful Paint12.4 s → 2.0 s
Total Blocking Time3,270 ms → 0 ms
Best Practices73 → 100
Lighthouse 13.4.1, emulated Moto G Power, Slow 4G, 16 September 2026. Lab measurements on shared hosting. View the full results table
At a glance
  • Measurement came first: server response, browser work, with and without caching, and the admin experience for logged-in users.
  • We optimised image sizes, removed unnecessary code and delayed non-critical scripts.
  • Premium hosting with reserved memory is a logical next step for heavier sites, not a guarantee of 100/100.
Short on time? Read what we improved and the checklist. The rest goes into more detail.
01

First, measure where WordPress makes you wait

This case concerns a B2B marketing agency with a WordPress website, a custom theme and ACF. The site includes services, case studies, articles and team photos. That content needed to keep working and displaying correctly. So we started by measuring.

We looked at both the server's initial response and what the browser then had to download and execute. We also compared cached and uncached pages and checked the experience for logged-in users.

That distinction matters. Caching can let a homepage return HTML quickly while visitors still wait for large images and scripts. Meanwhile, editors may face a different problem: WordPress has to do the work again for every admin action.

02

Two kinds of slowdown, two kinds of solution

The server has to generate the page

WordPress processes PHP, fetches database information and runs plugin and theme code. Page caching can skip much of that work for anonymous visitors. The admin area, logged-in visits and uncached requests usually do not benefit from the same shortcut.

The browser has to display the page

Images, stylesheets, fonts and JavaScript follow. A faster server does not make a multi-megabyte image smaller. Nor does it stop multiple trackers from asking a phone to do processing work at the same time.

We investigated both sides. That avoids simply changing caching settings while the actual bottleneck remains elsewhere.

03

What we improved on the WordPress website

Images sized for where they appear

The theme used large originals where smaller images would have been enough. We adjusted the selected image sizes and the instructions the browser uses to choose a suitable version. Images outside the initial viewport can load later, while the important first image gets priority. We also replaced a heavy PNG background with WebP.

Less unnecessary code during loading

A custom theme using ACF does not automatically need every standard WordPress stylesheet and script. We removed unused frontend components from the loading path and improved the order of the scripts the site does use. This needs checking for each website: something unnecessary here may be essential elsewhere.

Trackers no longer all start together

Analytics, marketing tags and other external scripts can create a lot of work together. Using WP Rocket's JavaScript settings, we delayed non-critical scripts. Scripts needed for immediate interaction, such as navigation and sliders, remained available when needed.

Keep checking that everything works

A high score means little if a menu only responds to the second click or a form stops working. Changes to JavaScript need checks on mobile and desktop. Consent and analytics also need to be checked again after delaying tags.

04

The results: before and after

Lighthouse mobile · 16 September 2026
MetricBeforeAfter
Mobile performance38/10096/100
First Contentful Paint1.4 s1.0 s
Largest Contentful Paint12.4 s2.0 s
Total Blocking Time3,270 ms0 ms
Speed Index7.3 s1.0 s
Cumulative Layout Shift00.104
Accessibility96/10095/100
Best Practices73/100100/100
SEO audit100/100100/100

The biggest gains are in loading the main visible content and reducing blocking JavaScript during the initial page load. LCP dropped from 12.4 to 2.0 seconds. Measured Total Blocking Time fell from 3,270 to 0 milliseconds.

These are Lighthouse lab measurements of the homepage under simulated mobile conditions, not average waiting times for all visitors. Scores can vary between runs. Delayed scripts can still create work after interaction, so we also keep testing how the site actually responds.

There are still points to address. Measured layout shift was 0.104 and the automated accessibility score was 95. A Performance score of 96 does not mean every part is perfect. Neither report includes field data to assess real visitors' experience over time.

Wondering what your score is and what causes it? Send us your website and, if available, your PageSpeed report. You will receive a brief analysis of the causes, with no obligation.
Request a speed check
05

Technical details for administrators and developers

Page cache, object cache and OPcache do different jobs

Page caching stores the generated page. A persistent object cache, such as Redis, can reduce repeated database queries and calculations. OPcache stores compiled PHP code so PHP does not have to compile it again each time. Which layer helps depends on the bottleneck. They do not replace each other.

A WebP file on the server is not necessarily a WebP delivered

Check the network panel to see what the browser actually receives. The srcset and sizes attributes must also match how the image is displayed. If a small area receives a much larger source than it needs, visitors still download unnecessary data.

Delaying scripts and their loading order go together

A script using defer waits until the HTML has been parsed. Delaying until interaction goes further. That can significantly lighten the initial page load, but dependencies and behaviour need checking. On this site, jQuery, Slick and the custom interface script had to work together correctly. Do not blindly copy an exclusion list to another website.

Check uncached requests too

If the admin area remains slow, investigate database queries, plugin work, background tasks and available PHP capacity. More memory only helps if memory availability is actually part of the problem. CPU capacity and configuration also matter.

06

When premium WordPress hosting makes a difference

The score of 96 shows how much careful optimisation can achieve on shared hosting. A migration does not automatically have to be your first step. Finding where the delay originates remains the best starting point.

Reserved memory can be a logical next step for heavier websites and online stores. Especially when WordPress processes many dynamic requests, a better-matched hosting environment can make performance more predictable.

Mantella Premium hosting €39 per month · excl. VAT
1 GB+reserved RAM per site
OPcachecompiled PHP in memory
25 GBNVMe storage
200 GBmonthly data transfer

Backups and monitoring are included. We can also arrange the migration for you, including checks of DNS, SSL and redirects.

Explore premium hosting

A score of 100/100 can also be achievable. It requires the theme, images, scripts and hosting to work well together. Extra RAM does not automatically guarantee that score. We measure which next step makes sense for your website, aiming for a fast site that works properly.

For a compact website, our basic hosting may be sufficient. If you need your own server environment, we can consider a VPS or dedicated server. The choice depends on how you use your website.

07

Checklist: check your WordPress website's speed yourself

  1. 1Test mobile and desktop and save the reports as a baseline.
  2. 2Look beyond the score at LCP, blocking time and shifting elements.
  3. 3Check how the website and admin area respond when you are logged in.
  4. 4Check the images the browser actually downloads, including dimensions and file size.
  5. 5List external scripts. Which need to start immediately and which can wait?
  6. 6Test menus, sliders, forms and consent after every script change.
  7. 7Measure again under comparable conditions and monitor real-user data where available.

Frequently asked questions about WordPress speed

Is a caching plugin enough to speed up WordPress?

A caching plugin mainly speeds up returning HTML to anonymous visitors. Large images, unnecessary scripts and trackers starting together remain. The admin area and logged-in visits usually do not benefit either. That is why we look at both server and browser.

Do I need to switch hosting first?

Not automatically. The score of 96 was measured on shared hosting. Establishing where the delay starts remains the best approach. Reserved memory is particularly useful for heavier websites and online stores with many dynamic requests.

Does the PageSpeed score affect my Google ranking?

Google uses mobile content for indexing and evaluates user experience through Core Web Vitals based on real-user data. The Lighthouse score itself is not a direct ranking factor. A faster website primarily benefits your visitors and editors.

Is 100/100 achievable?

It can be, when the theme, images, scripts and hosting fit together well. Extra RAM is not an automatic guarantee. We measure which next step makes sense for each website.

Mitchell van Honk
About the author Mitchell van Honk

Founder of Mantella, the technical partner behind businesses. Works daily on hosting, WordPress, custom software and automation for business owners, agencies and IT providers.

Sources
Get in touch