A clinic website that scored 30 out of 100
On 28 September 2026, our client’s homepage scored 30 out of 100 on Google’s mobile PageSpeed test. Eleven days and two working sessions later, the same page scored 97. Its main content now appears in about 2 seconds on a mid-range phone, down from 6.1 seconds.
The client is a pediatric therapy centre in Canada offering ABA therapy, speech-language pathology and occupational therapy for children and teens. We’ve kept their name out of this post and blurred their web address and site previews in the screenshots. Every number is exactly as Google reported it. Their site runs on WordPress with Elementor, the same setup used by thousands of clinics, practices and small businesses.
This is the full story of what we found, what we changed, and what each change was worth. We’ve kept the dead ends in too. One of our fixes made things worse before it made them better, and that part is probably the most useful for anyone working on their own site.
We’re Webspecia, a Google Partner agency that works mostly with healthcare brands. The clinic came to us for a new website and SEO. Fixing speed was the first job, because nothing else we do for a client works well on a slow site.
Where we started: 28 September
The first audit showed a site that looked fine but loaded badly. On a phone, visitors stared at a blank white screen for more than 3 seconds. The hero headline and image took over 6 seconds to appear. Once the page was visible, it jumped around as late-loading elements pushed the text down.


| Metric (mobile) | 28 Sep | Google’s “good” limit |
|---|---|---|
| First Contentful Paint | 3.6 s | 1.8 s |
| Largest Contentful Paint | 6.1 s | 2.5 s |
| Total Blocking Time | 1,370 ms | 200 ms |
| Cumulative Layout Shift | 0.201 | 0.1 |
| Speed Index | 6.7 s | 3.4 s |
Desktop was better but not good. It scored 64, and its layout shift of 0.414 was worse than mobile’s: on a wide screen, whole sections visibly jumped as the page assembled itself.

Why a few seconds matter for a clinic
A parent looking for speech therapy for their child is usually on a phone, often between other things. If the page is blank for three seconds, many will hit back and tap the next result. That lost visit never shows up in any report. It’s just a family that booked somewhere else.
Speed also feeds everything else a clinic pays for:
- Google search. Core Web Vitals (loading, responsiveness and visual stability) are part of how Google judges page experience. A slow site starts every ranking battle at a disadvantage.
- Paid ads. Every click on a Google Ads campaign costs the same whether the page loads in 1 second or 6. A slow landing page quietly raises your real cost per enquiry.
- Trust. A page that jumps around while you read it feels broken. For a healthcare provider, that first impression counts.
- AI assistants. Tools like ChatGPT and Google’s AI answers increasingly read websites directly. A page that renders cleanly and quickly is easier for them to read and cite.
So for this clinic, speed wasn’t a vanity score. It was the foundation for the SEO and lead generation work that came next.
Session 1: fixing the foundations (30 to 65)
The first session went after the problems that affected every page, not just the homepage.
A cache plugin that wasn’t doing anything. The site had a caching plugin installed, but it had never been configured. Every visit made WordPress build the page from scratch. We replaced it with W3 Total Cache and set up page caching, browser caching and compression. The server now hands most visitors a ready-made page.
Scripts that blocked the page. Several plugins loaded JavaScript in a way that stopped the browser from drawing anything until they had finished. We deferred the non-essential scripts so the page could paint first and run them after. Mobile Total Blocking Time dropped from 1,370 ms to 420 ms.
The jumping layout. We found the elements that loaded late and pushed content around, and fixed them. Layout shift fell from 0.201 to 0.029 on mobile and from 0.414 to 0 on desktop.
We also made the page easier for browsing agents to read. Google’s Agentic Browsing check went from 2/3 to 3/3.

Desktop jumped to 95. Mobile more than doubled, to 65. But the screenshot shows the problem we hadn’t solved yet: Largest Contentful Paint was still 6.2 seconds. The page started drawing sooner, yet the hero still took just as long to appear.
Session 2: chasing the last seconds (65 to 97)
The second session, on 9 October, was detective work. The easy wins were gone. Each remaining problem needed us to work out exactly what the browser was waiting for.
Step 1: lighter images and fewer plugins
Session 2 started with housekeeping. We compressed the homepage hero image and made separate, lighter versions for desktop and mobile. We turned off Google’s Site Kit plugin, which added its own scripts to every page, and disabled several other plugins the site no longer needed. Each plugin loads code on every visit, so fewer plugins means less for a phone to download and run. The mobile score moved from 65 to 87, but Largest Contentful Paint was still 3.7 seconds.
Step 2: the hero image nobody could see coming
PageSpeed’s breakdown told us something surprising. The biggest element on screen wasn’t the hero photo at all. It was the hero section’s background image, set in the page’s CSS.
That image was only 13 KB, small enough to load in a blink. The problem was timing. The browser can’t know a CSS background exists until it has downloaded and read the stylesheet. So it sat idle for 1.6 seconds before it even started fetching the image.
The fix was one line in the homepage’s <head>: a preload hint telling the browser to fetch that image straight away, at high priority. There was one catch. The cache plugin rewrote the preload’s link with a version number, so it no longer matched the link in the CSS, and the browser would have downloaded the image twice. We rewrote the tag in a form the plugin leaves alone. The wait dropped from 1,620 ms to 750 ms, and LCP fell from 3.7 s to 2.4 s, inside Google’s green zone.
Step 3: drawing the page before the stylesheets arrive
Two combined stylesheets were still blocking the first render for about 450 ms. On every visit, the browser downloaded all of the site’s CSS before it drew a single pixel.
The standard fix is critical CSS. We pulled out only the styles the header and hero need, about 37 KB, and placed them directly inside the page. Everything else loads in the background. First Contentful Paint dropped from 1.6 s to 0.9 s.
Step 4: the fix that made it worse
Then the score fell to 78. Layout shift jumped from 0 to 0.32. The page appeared sooner, but the hero paragraph visibly moved after it appeared.
We found two causes:
- Thirty stylesheets arriving one by one. With the CSS loading in the background as separate files, the page passed through dozens of half-styled states. We combined them into a single file, so the page goes straight from “critical styles only” to “fully styled” in one step.
- The fonts. This was the subtle one. Our critical CSS included the font declarations for Poppins. When the full stylesheet arrived, it declared the same fonts again. The browser treated those as new fonts and swapped the hero text from the fallback font to Poppins mid-read, so every line reflowed. Removing the font declarations from the critical CSS fixed it.
Layout shift went back to 0. The mobile score climbed to 92, and then to 95 on the next test.
Step 5: SEO fixes found along the way
Two smaller issues turned up during testing:
- A slow robots.txt. The file that tells search engines what to crawl was being generated by an SEO plugin on every request, taking 1.5 to 3.5 seconds. PageSpeed’s own check timed out waiting for it. Replacing it with a plain static file brought it down to about 270 ms.
- “Learn More” links. Seven program cards used identical “Learn More” buttons, which tell Google and screen readers nothing about where they lead. We added hidden descriptive text (for example, “Learn More about our social skills groups”) without changing how the buttons look. We fixed a typo on one of the cards while we were there.
The SEO score went from 92 to 100.
Step 6: desktop clean-up
On desktop, the homepage was loading scripts from two Elementor add-on plugins that it didn’t use at all, including animation and filtering libraries. We switched them off for the homepage only, so every other page stays as it was. Desktop blocking time fell from 350 ms to about 210 ms on the best run.
Step 7: upgrading PHP
The last change was on the server. The site was still running PHP 7.4, a version that stopped receiving security updates at the end of 2022. We upgraded it to PHP 8.3.
PHP is the language WordPress runs on, so a newer version builds each page faster whenever the cache can’t serve it: logged-in visitors, forms, search, and the first visit after a cache clear. It also closes off the security risk of running an unsupported version. We didn’t run a separate PageSpeed test just before and after this one change, so we can’t give it its own score. It made the server faster under everything else we’d done.
The results
On mobile, the clinic’s homepage went from 30 to 97 in two working sessions. Every Core Web Vital is now in Google’s green zone.


| Metric | 28 Sep | 8 Oct | 9 Oct |
|---|---|---|---|
| Mobile: First Contentful Paint | 3.6 s | 1.5 s | 1.0 s |
| Mobile: Largest Contentful Paint | 6.1 s | 6.2 s | 2.0 s |
| Mobile: Total Blocking Time | 1,370 ms | 420 ms | 180 ms |
| Mobile: Cumulative Layout Shift | 0.201 | 0.029 | 0 |
| Mobile: Speed Index | 6.7 s | 4.0 s | 1.5 s |
| Desktop: Performance score | 64 | 95 | 86 (85 to 92 across runs) |
| Desktop: Largest Contentful Paint | 2.1 s | 0.8 s | 0.9 s |
| Desktop: Cumulative Layout Shift | 0.414 | 0 | 0.021 |
| SEO score (both) | 92 | 92 | 100 |
Desktop needs a straight answer. It hit 95 on 8 October, and our final runs scored between 85 and 92. Desktop scores are mostly decided by Total Blocking Time, which swings from run to run. The page itself loads its main content in under a second, so visitors see a fast site. What’s left is script and style processing, and that’s the next job.

What other clinic websites can learn from this
If your practice runs on WordPress and Elementor, there’s a good chance some of these problems are on your site right now.
- Check that your cache plugin is actually configured. Installing one isn’t enough. This clinic had one that did nothing at all.
- Find out what your LCP element really is. It’s often not the image you’d guess. PageSpeed’s “LCP breakdown” tells you, and the fix depends on it.
- Background images in CSS load late. If your hero uses one, preload it, and make sure your cache plugin doesn’t change the URL you preload.
- Every plugin costs something on every page. Site Kit, page-builder add-ons and widgets each add scripts. Remove what you don’t use, or switch it off on pages that don’t need it.
- Critical CSS is powerful but fragile. It made this site’s first paint almost twice as fast, but it caused layout shift until we combined the stylesheets and dealt with the fonts. Test on a phone after every change.
- Small SEO details add up. A slow robots.txt and vague “Learn More” links each cost points. Both took minutes to fix.
- Re-test more than once. Scores move a few points between runs, especially on desktop. Judge by the metrics and the trend, not a single number.
None of these changes touched the clinic’s design. Visitors see the same website. It just shows up much faster.
Is your website losing patients to slow load times?
Most clinic owners never see their site the way a parent on a phone does. Put your address into the free scanner on webspecia.in to see how fast it loads on mobile and what’s holding it back.
If you’d rather talk it through, book a strategy call. We’ll run the same audit we ran for this clinic and tell you which fixes would make the biggest difference to your site, your Google rankings and your enquiries.
Webspecia is a Google Partner agency specialising in healthcare. We handle website speed, SEO, AI search visibility, Google Ads and CRM for clinics, diagnostics centres and specialists.
All scores in this article are lab results from Google PageSpeed Insights (Lighthouse) for the clinic’s homepage. Mobile tests use an emulated Moto G Power on slow 4G. Lab scores vary by a few points between runs.





