Case Study
True Tree Service. Same company. A site eleven times lighter.
A tree service covering Humboldt and Del Norte counties came off a funnel-builder site onto a custom build. I measured both while they were still live, side by side.
The Brief
The crew was booked. The website was holding them back.
True Tree Service LLC is a native-owned tree company working Humboldt and Del Norte counties. Licensed contractor, CSLB #1132934, with ISA certified arborists on the crew. Removals, trimming, stump grinding, storm work.
Their site lived inside GoHighLevel, the same software that ran their leads. That is a common setup and it works for collecting a form. It is not built to rank. On a phone the homepage took 6.7 seconds to show its main content and more than 10 seconds before you could use it. It weighed 5.1 megabytes. Most of the pages carried the exact same title in Google.
The job was to keep everything that worked about the business and rebuild the part Google and customers actually touch.
The Result
A clean 100, on all four of Google's scores.
These are Google's own four scores, run on mobile against both homepages on the same day. Three runs each, middle result shown.
-
Performance
62 100 -
Accessibility
82 100 -
Best Practices
77 100 -
SEO
91 100
Old site: 62 / 82 / 77 / 91. New build: 100 / 100 / 100 / 100, held across three runs in a row.
Under the Hood
What the phone actually has to do.
A score is a summary. This is what sits behind it. Most tree calls come from someone standing in a yard on cell service, so every one of these is time a customer spends waiting.
| Measured on mobile | Old site | New build |
|---|---|---|
| Main content on screen Largest Contentful Paint | 6.7 s | 1.6 s |
| First thing painted First Contentful Paint | 5.3 s | 1.1 s |
| Ready to tap and scroll Time to Interactive | 10.4 s | 1.6 s |
| Page weight Everything the phone downloads | 5.1 MB | 473 KB |
| Files requested Separate downloads per visit | 137 | 17 |
| JavaScript files Code the phone has to run | 835 KB | None |
| Outside servers contacted Third-party domains | 13 | 1 |
What Google Can Read
Fast is half of it. The other half is being findable.
A site can load in a second and still not rank if Google can't tell its pages apart. These are the plain structural things I checked in the code of both sites.
-
Pages
Old site: About 20
239
Every service paired with every town and both counties.
-
Page titles
Old site: 7 of 9 pages checked shared one title
Unique on every page
The title is the blue link in Google. Seven pages all said the same three words.
-
Image descriptions
Old site: 2 of 17 images
16 of 16 images
Homepage images with alt text for screen readers and image search.
-
Sitemap
Old site: Empty file
237 URLs
The list Google reads to find every page.
-
Canonical tag
Old site: None
On every page
Tells Google which address is the real one.
-
Structured data
Old site: 2 types
25 types
Homepage schema. Services, towns served, reviews, FAQs, hours, license.
How It Was Built
Nothing on the page that doesn't earn its place.
The gap between those two columns comes from how each site is put together, not from a faster server or a trick. Four decisions account for nearly all of it.
-
Hand-coded, no page builder
The old site ran on a funnel builder that ships its whole toolkit to every visitor. The new one is written for this one company, so the phone downloads the page and nothing else.
-
No JavaScript to wait on
The page is finished HTML when it arrives. Nothing has to boot before someone can read it or tap the phone number. That is where nearly nine seconds went.
-
A page for every job in every town
Ten services across sixteen towns and two counties, each one written for that place. Someone searching stump grinding in Fortuna lands on a page about stump grinding in Fortuna.
-
Schema built in from day one
License number, service area, hours, reviews, and FAQs are all marked up so Google and AI search can read exactly what the company does and where.
How this was measured: Google Lighthouse 13.5, mobile profile, homepage of each site, October 9, 2026. Each site was run three times and the middle result is shown. Titles, image descriptions, canonical tags, sitemaps, and structured data were read directly from each site's code the same day. The new site first scored 97 on accessibility because a fade-in animation read as low contrast. I removed the fade and ran it again: 100 on all three runs.
Work with Tex
Want your site measured like this?
Send me your address and I'll run the same test on it. If a custom build would move the numbers, I'll show you which ones and by how much. I reply within 24 hours.