Why we ran this
Homeowners look for an HVAC company the same way every time: phone in hand, on a hot day, in a hurry. The website has a few seconds to load, show a number they can tap, and prove the company is real. Miss any of those and the call goes to the next result.
So we measured it — not for the companies that rank or advertise, but for the official population of licensed contractors. We took the public register, drew a random sample, found each firm’s website, and ran the same checks on every one. No cherry-picking, no vendor’s “top 10” list. Anyone can re-derive the sample from the same public source.
The numbers above are what we found. Below is exactly how we got them, and the limits of what they can tell you.
What the numbers say — and don’t
The headline is speed. On a phone, the typical site in this sample takes several seconds to show its main content; Google’s own research puts the point where most visitors give up at around three. We are not claiming a site’s speed causes a specific number of lost jobs — we didn’t measure any single company’s bookings, and we won’t invent a dollar figure. What we can say is narrower and still damning: a large share of these sites fail the basics a mobile visitor needs before they’d ever call.
Where you see a relationship — say, sites on one platform loading slower than another — read it as an association, not proof of cause. Platform choice, budget, and who built the site all travel together. We report the pattern; we don’t pretend to know the direction.
Method
- Population: every active licensed A/C contractor in Dallas County on the Texas Department of Licensing & Regulation (TDLR) ACR register, taken from TDLR’s public bulk data. Deduplicated by business name.
- Sample: a fixed-seed random draw from that population, so the exact set is reproducible. We then searched for each firm’s own website — confirming against the licensee’s name where we could, and rejecting directory pages (Yelp, BBB, Angi and the like), which are never the company’s site.
- Three outcomes per firm: website found and audited; no website found (a data point, not a dropped row); or ambiguous/national-brand, excluded from the site audit but kept in the count. The figures above are computed on the subset we found a live website for — we don’t headline a “no website at all” rate, because our search could miss a site that exists, and we won’t overstate what we can’t verify.
- Speed and technical checks: Google PageSpeed Insights (mobile), plus a direct read of each homepage for tap-to-call, contact form, mobile viewport, HTTPS, LocalBusiness schema and platform. Most small-contractor sites don’t get enough traffic for Google’s real-user (field) data, so most speed numbers here are lab measurements — treat them as directional.
- Limits: heuristic checks misfire (a site that dials via a script instead of a plain
tel: link can read as “no tap-to-call”), sample sizes per platform get small fast, and geography is county-level. The full anonymised dataset is linked above so you can check our work.
A note on privacy
We deliberately name no company. Every row in the public dataset is an opaque ID — city, platform, and measurements only. The point of the study is the method and the aggregate, not putting any single contractor on blast.