One extra second of load time costs an average of 7% conversion. But most small-business sites are at 3-5 seconds. How to get under 1 second — and what that actually delivers.
1. Why sub-second matters now
Google's Core Web Vitals (LCP, INP, CLS) have been ranking signals since 2021, but the thresholds have tightened. 'Good' was 2.5 seconds LCP — in 2026 users on 4G/5G mobile expect under 1.2s. What was fast 5 years ago now feels slow.
2. The 4 real bottlenecks (measured, not theory)
- Too much JavaScript — WordPress themes load 80-150 requests; Astro stacks do it with 7-18
- Hero image too heavy — 800KB JPG vs 80KB WebP costs 600ms on 4G
- Third-party scripts — Google Tag Manager + 3 trackers = 800ms blocking time
- No CDN/edge — origin server in DE, visitor in IJmuiden = 200ms extra TTFB
3. The stack we use (and the numbers)
Two live examples from our own work, measured last week:
- dammie.store (eyewear shop): 473ms full load · 18 requests · 52KB HTML — Astro + Cloudflare Pages
- oceansnacks.nl (snack bar): 255ms full load · 7 requests · 23KB HTML — same stack, lighter
4. Quick wins you can do today
- WebP for all images — average 70% smaller than JPG, supported in all modern browsers
- fonts via font-display:swap — prevents 200ms of invisible text
- lazy-load everything below the fold — img loading='lazy' is free
- Google Tag Manager async — or better: server-side tracking via Cloudflare
5. When a rebuild pays off
At a 2% conversion rate and 10,000 monthly visitors, every second of load time costs ~140 conversions/year. At €100 margin per conversion = €14k. A rebuild typically costs €5-15k. Run your business case on real numbers, not gut feeling.
6. What sub-second does NOT fix
A fast site won't sell a bad product. If your conversion is 0.3% at 5 seconds, it'll be maybe 0.5% at 0.5s. Important: speed is the floor (without it you have nothing), not the ceiling. Fix speed first, then funnel/copy.