How This Site Loads in Under 100ms (Steal These Ideas for Your Store)
Quick answer: This site is fast because of what it doesn’t ship: no JavaScript framework, no web fonts, no trackers, no cookie banner, and all CSS inlined into a single small HTML file per page, served from a CDN edge. Your store can’t go that far, but the principle (every byte earns its place) is exactly how we make stores pass Core Web Vitals.
We sell speed, so this website has to be the demo. If you’re reading this on a phone, the page you’re looking at probably arrived in one round trip and rendered in well under a second (most visits, much faster).
No exotic engineering. Just subtraction. Here’s the full recipe, including the parts you can steal for a store.
The recipe
1. One HTML file per page, everything inlined
The entire stylesheet for this site is small enough that we inline it into every page. The first network response contains everything needed to render: no stylesheet round trip, no render-blocking chain, no flash of unstyled anything.
Steals for stores: inline your critical CSS (the part that renders the first screen) and defer the rest. Most themes do the opposite.
2. Zero web fonts
This page uses your operating system’s own fonts. That’s one of the larger speed decisions on the site: no 100–300KB of font files, no invisible-text flash, no layout shift when fonts swap in. System fonts are excellent now, and nobody has ever abandoned a cart over a font.
Steals for stores: if the brand needs a custom font, use one weight or two, font-display: swap,
preload it, subset it. If it doesn’t, skip it entirely.
3. JavaScript only where it does a job
There are exactly two scripts on this whole site: the little “this page loaded in X ms” counter on the homepage, and the inquiry form’s inline thank-you. Both are a few hundred bytes, inline, and run after render. There is no framework runtime, no hydration, no bundle.
A static site generator (we use Astro) does all the templating at build time and ships plain HTML.
Steals for stores: you can’t have zero JavaScript, but every script can justify itself, load deferred, and stay off the path between your customer and “Add to cart.” INP explains why this matters most.
4. No trackers, no cookie banner
We don’t run analytics scripts, pixels, or session recorders here. That’s a speed feature twice over: no script cost, and no consent banner (the most common cause of layout shift on the modern web, and the first thing your customers meet).
Steals for stores: you’ll keep some analytics, but server-side or lightweight options cover most stores’ real decisions, and every pixel you cut pays for itself in INP.
5. Static files on the edge
The site is plain files on Cloudflare’s CDN. Nothing computes when you visit; the nearest edge node just hands over the page. Time-to-first-byte stays low on every continent.
Steals for stores: your platform already has a CDN. The wins are in caching policy: long, immutable caching for assets so repeat visits and page-to-page navigation are near-instant.
What it adds up to
| Decision | What it removes |
|---|---|
| Inlined CSS, one file per page | The render-blocking chain |
| System fonts | 100–300KB + layout shift |
| No framework runtime | The hydration tax on every phone |
| No trackers | Main-thread time + the consent banner |
| Static on the edge | Server think-time |
None of this is heroic. It’s a budget, enforced. The same discipline, applied to a real store with apps, variants, and a checkout, is the entire job: every script, image, and font either earns its place on the page or leaves.
FAQ
Can a real store actually load this fast? Not quite. A store legitimately needs more: product imagery, variant logic, payments, some analytics. But stores routinely go from 5s to under 2.5s and from 500ms taps to under 200ms, which is the difference between failing and passing Core Web Vitals, and between losing and keeping half your mobile visitors.
What’s the single highest-leverage steal? The script inventory. List every third-party script, ask “would I install this today?”, remove the noes, defer the rest. It’s unglamorous and it’s the biggest number on the typical gains table.