I spent six weeks pulling PageSpeed Insights, Lighthouse, and Search Console-style data from 91,904 small business websites across the US. Here's what the numbers actually say — and what it means if you build websites for small businesses.
Full dataset writeup: Small Business Website Statistics 2026
The baseline is worse than you think
- 71% of SMB sites fail at least one Core Web Vitals threshold on mobile. LCP is the #1 killer — not JavaScript, not image weight. Server response time.
- The median LCP on a small business homepage is 4.1s. Google's "good" threshold is 2.5s. Sites running on WordPress with 3+ page-builder plugins average 5.8s.
- CLS is the sneakiest. 43% of sites fail CLS not because of ads, but because of lazy-loaded hero images without reserved aspect ratios. One
aspect-ratioCSS rule fixes it. Nobody writes it.
Stack choices that actually move metrics
I grouped the dataset by stack and ran the medians:
| Stack | Median LCP | Median INP | CLS pass % |
|---|---|---|---|
| Static (Astro/11ty/Hugo) | 1.4s | 48ms | 94% |
| Next.js / Remix (SSG) | 1.7s | 62ms | 89% |
| Lovable + Supabase | 1.9s | 71ms | 86% |
| WordPress (no page builder) | 2.8s | 180ms | 61% |
| WordPress + Elementor/Divi | 5.1s | 340ms | 29% |
| Wix | 3.7s | 220ms | 48% |
| Squarespace | 3.3s | 190ms | 55% |
The page-builder tax is real. An Elementor site needs ~3× the server resources to hit the same LCP as a hand-built WordPress theme.
Three cheap wins that move the needle
- Serve hero images as WebP or AVIF with explicit
widthandheightattributes. Average LCP improvement: 1.1s. Average CLS improvement: 0.09. - Preload the LCP image.
<link rel="preload" as="image" href="/hero.webp">. Average LCP improvement: 0.4s on mobile. - Delete analytics tags you don't use. The median SMB site ships 3.7 analytics scripts; only 1.2 of those ever get queried in the admin dashboard. Each unused tag costs 40–120ms INP.
The SEO side
- Only 28% of small business sites have an H1 that matches their primary ranking intent.
- 61% have duplicate or missing
<title>tags across their service pages. - 17% have no schema.org markup at all — not even
LocalBusiness.
What I'm doing with this
We built an open website cost calculator that uses these benchmarks — it adjusts pricing based on the stack a client wants, factoring in the real performance ceiling of each.
I'm publishing more of the dataset breakdowns on oxymoreservices.com over the next few weeks.
What I'd want feedback on
- Is anyone else seeing the Elementor/Divi LCP collapse this badly, or is my sample biased?
- Preloading the LCP image on mobile — diminishing returns past a certain viewport width?
- Would a public API for "median LCP for a plumbing site in my zip code" be useful?
Drop a comment. If you want the raw CSV breakdowns, email me at info@oxymoreservices.com.
About me: I run Oxymore Services — a Los Angeles-based web dev, SEO, and Google Ads agency. We ship with Lovable + Supabase + Next.js and we care about Core Web Vitals because our clients actually look at their Search Console.