One WordPress site, nine months, every number dated
An Australian workplace-wellbeing provider hired me in November 2025. Their site was slow for the people it was built for, and they wanted to know why.
I worked on it until August 2026. Every result below has a tool and a date next to it. The “after” numbers come from August, my last month on the site, and not from my best day. I’ve included what didn’t go up.
Where it started
On Nov 24, 2025, GTmetrix tested the homepage from Sydney. It weighed 9.16 MB across 226 requests and graded C (64). It took 28.2 seconds to finish loading. Most of the weight was images (7 MB).
What I did
- Dec 2025: replaced a robots.txt that disallowed the whole site. I also rebuilt the schema.
- Feb 5, 2026: took the unused Facebook pixel out of Tag Manager.
- Mar 2026: converted the site’s images to WebP and set up Perfmatters for script and CSS loading. On Mar 12 I added a GA4 page_view tag I’d found missing the week before (GTM version 31).
- Jun 2026: the client wanted a background video on the homepage and supplied the file. I built it as a CSS background so it wouldn’t become the element the page waits on, and kept it off phones. I also took out schema ratings I’d added in December.
- Aug 2026: my last month. GTmetrix ran the homepage every hour until Aug 13.
How it measured in August
Speed
GTmetrix, homepage, from Sydney on desktop:
| Nov 24, 2025 (one test) | Aug 1 to 13, 2026 (median of 300 hourly tests) | |
| Grade | C (64) | B (82) |
| Largest Contentful Paint | 2.8 s | 2.0 s |
| Total Blocking Time | 276 ms | 190 ms |
| Fully loaded | 28.2 s | 3.5 s |
| Page size | 9.16 MB | 4.4 MB |
| Requests | 226 | 69 |
The August page size includes the client’s background video, about 2.4 MB of it. LCP stayed at or under 2.5 seconds in 282 of the 300 August tests.
It wasn’t an A. The site hit grade A on 9 of those 300 tests. What held it at B was script work tying up the browser, a median 190 ms of blocking time.
Tracking
On Feb 4, 2026 the page took 30.6 seconds to finish loading. A day later, with the pixel out of Tag Manager, it took 3.2 seconds. The August median was 3.5 seconds.
Search
SEMrush counted the site ranking for 422 keywords in Australia in November 2025, 31 of them in the top three. On Aug 13, 2026 it was 778 keywords, 42 in the top three.
That isn’t the whole picture. On the 52 keywords I tracked in SEMrush, visibility fell over the same months, and top-three positions dropped from 35 to 29 (Dec 10, 2025 to Aug 13, 2026). The site was showing up for more searches. It wasn’t winning more of the ones that mattered most.
After August
I stopped maintaining the site in August 2026. When I re-tested on Nov 8, the homepage from Sydney on desktop measured B (82), the same as August. Eight of the twelve test setups I’d saved were slower than their June and July runs, mostly on mobile and the inner pages, from JavaScript tying up the browser.
If your site sounds like this one
Book a call and I’ll tell you what I’d look at first.
