We measured 12 Houston roofing sites. The basics were fine.
We expected to find broken fundamentals. Instead every site had HTTPS, a viewport tag, a sitemap and a meta description. The damage was somewhere else entirely — and it was heavy.
What we measured, and what we did not.
We took twelve roofing companies serving Houston and the northeast suburbs — Humble, Kingwood, Atascocita — and fetched each homepage once on 8 September 2026, recording what a single request can establish: document weight, server response time, the size of the images the page references, how many scripts it loads, and whether the technical fundamentals are present.
We did not measure Core Web Vitals, and we are not going to claim we did. LCP, INP and CLS need a real browser and, to mean anything, field data gathered from actual visitors over weeks. Asserting them from a command-line fetch would be precisely the kind of unearned number this exercise exists to avoid.
Sites are not named. This is aggregate research into a market, not a list of businesses to embarrass, and several of them are doing more right than wrong.
We expected broken fundamentals. We did not find them.
The premise going in was the usual one: local trade sites neglect the basics. That turned out to be wrong, and it is worth saying so plainly rather than quietly reframing the finding.
Every site had a valid certificate, a viewport tag, a title, a meta description, a robots file and a sitemap. Eleven of twelve published structured data. Whoever built these sites — and in several cases it was clearly a template — got the checklist right.
That is genuinely good news, and it means the common sales pitch aimed at these businesses (your site is missing the basics) is mostly false.
The weight is in images and third-party scripts.
- 01
Three sites ship a homepage image over 1 MB
The largest single image we found was 2.2 MB, on one homepage, above the fold. Four of eleven sites had an image over 500 KB. For a homeowner on mobile data after a storm, one of those images alone can outweigh everything else on the page combined.
- 02
The median site loads 41 script tags
One loaded 65. Seven of twelve were above thirty. Each one is code that has to be fetched, parsed and executed before the page settles, and most of them are third-party — chat widgets, tracking pixels, review embeds, tag managers.
- 03
The HTML itself is not the problem
Median document weight was 47 KB and none exceeded 100 KB. These are not bloated pages in the way people usually mean. The documents are lean; what they pull in afterwards is not.
- 04
Servers were mostly quick
Median response for the HTML document was 0.38 seconds, and only one site took longer than a second. Hosting is not the bottleneck for this group.
This is the pattern worth taking away. The documents are lean, the servers are quick, the metadata is present — and then the page pulls in a two-megabyte hero photograph and forty-odd scripts, and none of the good work upstream survives contact with a phone on congested mobile data.
Both of those are also the cheapest possible things to fix. Compressing an image and auditing which third-party tags actually earn their place are afternoon jobs. Neither requires a rebuild, and neither shows up in a design review — which is probably why they persist.
What this does not tell you.
Twelve sites is a sample, not a census, and they were chosen because they surfaced for ordinary Houston roofing searches — which biases toward businesses already doing some marketing. A site nobody can find was never going to appear here.
Each site was fetched once, from one location, on one day. Response times move. Image weights change when someone swaps a hero photo. And a homepage is not a website — the interior pages, where most organic traffic actually lands, were not measured at all.
Treat it as a directional read on where the effort is worth spending in this trade, not as a verdict on any particular business.
What we would check on your site.
The same measurement, run properly against your live site and extended to the things a single fetch cannot see, is the free audit. Twelve checks, run by a person, reported in plain English.
If you already suspect the problem is weight rather than structure, targeted speed work is cheaper than a rebuild — and based on this sample, it is where most of the available gain actually sits. More on what roofing sites specifically need is on the roofing websites page.
We ran the same measurement against Houston med spa sites, which carry roughly four times the image weight and half again as many scripts — useful for separating an industry habit from how websites generally are.
Find out which group you are in.
Send your address and we will run the full twelve-check audit — including the image and script weight that this sample says is doing the damage.