Why All WordPress Site Speed Plugins Suck

Your WordPress site is slow. You know it, Google knows it, and your customers definitely know it. So you do what every WordPress site owner does — you install a speed plugin.

WP Rocket, FlyingPress, LiteSpeed Cache, Perfmatters, Autoptimize. You toggle some switches. You minify CSS, defer JavaScript, lazy-load images. You run PageSpeed Insights. The score goes up a few points.

But your Core Web Vitals are still failing. Your Search Console still shows “Poor” URLs. Your pages still feel sluggish on mobile.

Here’s the uncomfortable truth: site speed plugins can’t fix what’s actually wrong with your site. Not because they’re bad software — some of them are genuinely well-engineered. But because the problems they solve aren’t the problems you have.

What Speed Plugins Actually Do

Let’s be fair about what these plugins are good at. Most offer some combination of:

  • Page caching — serving a static HTML copy instead of running PHP on every request
  • CSS/JS minification — stripping whitespace and comments from code files
  • Image lazy-loading — deferring offscreen images until the user scrolls
  • File combining — merging multiple CSS or JS files into fewer requests
  • Browser caching headers — telling browsers to cache static assets longer
  • Critical CSS generation — inlining the CSS needed for above-the-fold content

These are legitimate optimisations. On a clean, well-built WordPress site with a lightweight theme and a handful of carefully chosen plugins, a caching plugin might be all you need.

But that’s not your site, is it?

The Problems No Plugin Will Solve

1. Your Plugins Load Everything, Everywhere

This is the single biggest performance issue on most WordPress sites, and no speed plugin addresses it.

Here’s what we regularly find on real client sites: 40–50 CSS stylesheets and 80+ JavaScript files loaded on every single page. Not because the page needs them — because that’s how WordPress plugins work.

Install an LMS plugin like LearnDash? Its 10 CSS files now load on your homepage, your blog posts, your contact page — even though courses only exist on a handful of URLs. Add WooCommerce with a few extensions — checkout scripts, payment gateway scripts, and cart widget styles load on pages where nobody can buy anything. Add a testimonials plugin, a contact form plugin, a pricing table plugin. Each one registers its assets globally because that’s the WordPress default.

A speed plugin can minify those 50 stylesheets. It can try to combine them into fewer files. But it cannot know that LearnDash CSS has no business loading on your homepage. It cannot decide that WooCommerce Stripe gateway JavaScript should only appear on checkout. That requires a human who understands your site, your plugins, and which assets belong on which pages.

The real fix is a custom MU plugin that conditionally dequeues assets per page context — something that takes hours of code auditing and careful rule-writing. No toggle in a plugin settings panel can replace that.

2. Your Server Is the Bottleneck

Time to First Byte (TTFB) measures how long it takes your server to start sending the page. Google’s threshold is under 800ms. We routinely see WordPress sites at 1,500–3,000ms.

Speed plugins add a caching layer on top of your web server. That helps — when the cache is warm and the page is cacheable. But here’s what they can’t address:

Your database is misconfigured. We recently audited a site where MariaDB’s buffer pool was set to 20GB on a 32GB server, starving everything else of RAM and pushing the system into swap. The site had four databases totalling 71GB — two of which were abandoned copies from old migrations that nobody had cleaned up. A single audit log plugin had accumulated 4.7GB of data. No speed plugin checks your innodb_buffer_pool_size or tells you to drop orphaned databases.

Your PHP version is outdated or misconfigured. Sites still running PHP 7.4 with no OPcache, or OPcache configured with default memory limits that are too small for their plugin stack. The difference between PHP 7.4 and PHP 8.2 with OPcache and JIT enabled is not marginal — it can halve your server response time. No speed plugin upgrades your PHP runtime.

Your web server is wrong for the job. Apache with mod_php serving a WordPress site with 50 plugins is fundamentally slower than LiteSpeed with LSAPI. The web server choice determines how efficiently PHP processes are managed, how connections are handled, and whether cached pages can be served without touching PHP at all. A speed plugin running on top of a slow web server is like a fast paint job on a car with a blown engine.

WooCommerce sessions bypass your cache. This is one that catches almost everyone. WooCommerce sets session cookies for visitors — and most page cache systems (correctly) refuse to serve cached pages to users with cookies, because the page might contain personalised cart data. The result: a significant portion of your traffic hits the uncached, slow PHP path. Properly configuring cookie-aware cache bypass rules requires understanding your specific e-commerce setup — which cookies matter, which pages must be dynamic, and which can safely be cached despite the cookie. Generic speed plugin settings rarely offer you this level of config, and those that do, rely on you to know what you are doing to ensure it’s configured optimally.

3. Your Third-Party Scripts Are Killing You

Here’s a real example from a recent audit. A client’s site loaded scripts from 18 different third-party domains on every page:

  • Google Tag Manager (with a container full of tags)
  • Google Ads remarketing
  • Bing/Microsoft Advertising
  • A B2B visitor identification service (contract had been cancelled — but the script was still running, burning 1.75 seconds of execution time on every pageload)
  • A chat widget
  • A cookie consent manager
  • A translation service
  • Font loading from Adobe Typekit
  • Multiple analytics and identity resolution services that nobody in the organisation could identify or explain

Speed plugins cannot audit your Google Tag Manager container. They cannot tell you that you’re loading a cancelled Lead Forensics script that’s costing you nearly two seconds per page. They cannot identify that your Google Ads tracking is firing twice — once through GTM and once through a hardcoded script — sending data to a GA4 property that nobody is monitoring.

And here’s the dangerous part: blindly deferring or combining third-party scripts can break your conversion tracking and cost you real money. We’ve seen sites lose visibility of their paid advertising returns overnight because a speed optimisation interfered with how analytics tags fire. The difference between “defer this script” and “break your Google Ads conversion attribution” is a single incorrect configuration.

Third-party script optimisation requires understanding not just performance, but data flows — which tags talk to which platforms, which fire order matters, how consent management interacts with tag execution, and what the actual business impact is of removing or deferring each script.

The correct optimisation of scripts requires a real-world conversation with the marketing team. Like “Hey we found this script. We think it belongs to XXZ software. Do you still need this?” or “Hey, we found two GA tracking scripts. We can remove one. Which one is redundant?”. Unfortunately, site speed plugins never have these important conversations with the marketing team, so the scripts slowing down your website remain unchecked.

4. Your Fonts Are Render-Blocking

Many WordPress sites load web fonts from external services — Google Fonts, Adobe Typekit, or similar. These font requests sit directly in the critical rendering path. Your browser discovers the page needs a specific font, makes a request to a third-party server, waits for the response, and only then renders the text.

When your Largest Contentful Paint element is a heading (which it is on most content sites), the font load time is directly added to your LCP score.

Speed plugins might add font-display: swap — which shows a fallback font first, then swaps. That’s a partial fix. And it’s not a good one because it results in ‘ugly loading’ which is where fonts change before the eyes of the website visitors as the page loads. The real solution is self-hosting your fonts: downloading the woff2 files, writing proper @font-face declarations, and adding preload hints so the browser fetches them immediately. That requires manual work — extracting the right font files, understanding font licensing, editing theme templates, and adding the correct resource hints.

5. They Can’t Fix What They Can’t See

The most insidious problems are the ones that don’t show up in a lab test but devastate your field metrics.

Speed plugins optimise what they can see: the HTML output, the asset queue, the cache layer. But they’re blind to:

  • Consent management interactions — how your cookie consent banner affects when tracking scripts fire, and what happens to your conversion data when those scripts are delayed or blocked
  • Duplicate tracking implementations — the same analytics or advertising tag loading from multiple sources (hardcoded in the theme, injected by a plugin, and configured in GTM)
  • CDN misconfigurations — Cloudflare features like Rocket Loader that silently rewrite your script tags and change execution order, or a CDN edge cache that’s not coordinating purges with your WordPress cache
  • Database bloat patterns — plugin audit logs growing to gigabytes, orphaned tables from deactivated plugins, action scheduler tables filling with completed entries that are never cleaned
  • The actual regression — what specifically changed two months ago that pushed your scores from green to red? Was it a plugin update? A theme update? A new third-party tag? A hosting migration? Speed plugins can’t do forensic analysis of your git history or changelog.

6. Green Scores Don’t Mean Green Core Web Vitals

This is where we need to talk about something most site owners don’t understand — and where a lot of so-called speed experts exploit that gap.

PageSpeed Insights and Core Web Vitals are not the same thing.

PageSpeed Insights (PSI) runs a lab test. It loads your page once, in a simulated environment, on a simulated device, with a throttled connection. It measures what happens in that single, controlled run and gives you a score out of 100.

Core Web Vitals in Google Search Console are field data. They’re collected from real Chrome users visiting your site over a rolling 28-day window, measured at the 75th percentile. That means Google looks at how your site performed for real people, on those people’s real devices, on real internet connections — and asks: “Did 75% of visits meet the threshold?” If your 75th percentile LCP is 2.4 seconds, you pass. If it’s 2.6 seconds, you fail. It doesn’t matter what your lab score says.

These two numbers can (and frequently do) tell completely different stories.

Your PageSpeed Insights score can be 95 while Search Console shows every URL as “Poor.” Your lab LCP can be 1.2 seconds while your field LCP is 6 seconds. This happens because lab tests don’t account for real-world conditions: slow mobile networks, older devices, server load under real traffic, third-party scripts that behave differently under load, and the sheer variety of conditions your actual visitors experience.

Just in case there’s any doubt: when it comes to speed, PageSpeed Insights score has absolutely no bearing on how much Google likes your website, or how prominently it appears in the search results. It’s ONLY the Core Web Vitals in Search Console that moves the needle in this regard.

And this gap is exactly what bad “speed experts” exploit.

We recently examined the work of a self-proclaimed site speed professional who had been hired to fix a client’s Core Web Vitals. He delivered screenshots showing all-green PageSpeed Insights scores. The client was delighted — until months passed and Search Console still showed failing URLs. Nothing had changed in the field data. The “fix” had done absolutely nothing for real users.

When we dug into what he’d actually done, we found something remarkable. Among the usual plugin toggles and generic caching configuration, he had added CSS animation delays to above-the-fold content. Elements that would normally render immediately — headings, hero images, key text blocks — were given subtle animation delays of a few hundred milliseconds. To a real user, the page looked the same, maybe with a slight fade-in effect that seemed like a design choice.

But to Lighthouse (the engine behind PageSpeed Insights), those elements no longer counted toward Largest Contentful Paint at the moment the page loaded. By delaying when the browser considered the content “painted,” the LCP measurement in the lab dropped dramatically. The score turned green. The screenshots looked great.

The actual performance of the site hadn’t improved by a single millisecond. The HTML was still bloated. The server was still slow. The render-blocking stylesheets were still there. Real users on real devices still waited the same amount of time for the page to become usable. But the tool had been tricked into reporting a better number.

This is the performance equivalent of taking the bulb out of a car’s ‘check engine’ light instead of fixing the actual engine.

It’s worth repeating: Google ranks your site based on field data from real users, not lab scores from PageSpeed Insights. A consultant who delivers green lab scores without moving the field metrics hasn’t fixed anything. They’ve just made the dashboard look better while your search rankings continue to suffer.

When hiring someone to fix your site speed, ask one question: “Will this improve my Core Web Vitals in Search Console, and how will we measure that?” If they only talk about PageSpeed Insights scores, they’re either incompetent or planning to game the metric. Either way, find someone else.

The Real Solution

There’s no shortcut to genuine site speed improvement on a complex WordPress site. The real solution is a systematic, expert-led process:

  1. Measure properly. Not just PageSpeed Insights scores — real field data from Search Console, TTFB measurements across page templates, DOM inspection to understand what’s actually loading and why.
  2. Audit the full stack. Server configuration, database health, PHP environment, web server choice, CDN setup, WordPress plugin inventory, theme architecture, third-party tag stack. Every layer.
  3. Fix the infrastructure first. Right web server, right PHP version, right database configuration, right caching architecture. This is the foundation. No amount of frontend optimisation compensates for a misconfigured server.
  4. Surgically remove what doesn’t belong. Cancelled services still running, duplicate tracking, plugins loading assets globally when they’re needed on three pages. This requires understanding the site, not just toggling switches.
  5. Optimise the critical path. Self-hosted fonts, conditional asset loading, proper resource hints, critical CSS — implemented with knowledge of what each page actually needs.
  6. Validate everything. Not just “does the page load faster?” but “does checkout still work? Do forms still submit? Are plugins erroring due to server changes? Are conversions still being tracked? Did we break anything?”

A speed plugin can handle one step of a twenty-step process. The other nineteen require someone who understands WordPress architecture, server infrastructure, web performance, analytics data flows, and the specific business context of your site.

So Should You Uninstall Your Speed Plugin?

No. A good caching plugin is a necessary part of the stack — page caching, image optimisation, and basic asset management are table stakes. Use LiteSpeed Cache if you’re on LiteSpeed, WP Super Cache or WP Rocket if you’re not.

But stop expecting it to fix your Core Web Vitals. Stop installing a second speed plugin when the first one didn’t work. Stop buying premium “speed optimisation” plugins that promise green PageSpeed scores.

If your site has genuine performance problems — the kind that show up as failing Core Web Vitals in Search Console — you need an expert, not a plugin. Someone who will open the bonnet, look at the engine, and fix what’s actually broken.

Get Your Core Web Vitals Fixed Properly

Stop wasting time with plugins that can’t solve the real problems. We audit your full stack — server, database, plugins, scripts — and fix what’s actually broken.

Get in Touch

FREE 60-minute DIY hacks: Make AI chatbots recommend your brand

Turn ChatGPT, Gemini, Grok and Google’s AI Overviews into your Brand ambassadors in under an hour!

We’ll send you a FREE step-by-step guide (with real examples) to engineer the signals that push AI bots into thinking that your brand is the best thing since sliced bread.

You can implement most of it in under an hour and see the results within a few days without technical ability.

Where should we send the hack?

You may also like