Every web developer tells clients their site is too slow. I got tired of saying it without a number behind it, so we measured.
The method
We took the top organic results for 12 local service queries across 12 US cities — plumbers, dentists, salons, that kind of business. We kept only the businesses' own sites: no directories, no aggregators, no national chains. Then we identified each site's platform from its page source and ran one Lighthouse mobile audit per site through the PageSpeed Insights API.
139 sites measured, August 2026.
The results
The median site takes 8.9 seconds to render its largest content. Google's threshold for "good" is 2.5 seconds.
Four sites out of 139 met it. That's 3%.
The median page ships 1.2 MB of JavaScript before doing anything the visitor asked for. A quarter of the sample scored below 50 on Lighthouse performance. The median mobile score was 57.
What it isn't
It's lab data — one run per site, one point in time, US only. That's enough to describe a population, not precise enough to rank two individual sites against each other. Every caveat is listed in the write-up.
The part I didn't expect
Slow sites showed up on every platform. WordPress, the builders, sites we couldn't identify — the medians moved around but nobody was clean. That points at how sites get assembled rather than at any one vendor: a theme, a page builder, a chat widget, tracking scripts, a booking embed, sliders and fonts, each added by someone solving one problem without seeing the total.
The data is open
Every domain, its detected platform, the query it was found through and all measured values are published as a CSV. Re-analyse it, check my arithmetic, or run the same queries in your own city:
https://insemantic.com/data/small-business-website-speed-2026-08.csv
Full write-up with the method and every caveat: https://insemantic.com/blog/small-business-website-speed-study/
If you spot a hole in the methodology, I'd rather hear it than not.
