The Complete Website Performance Guide for UK SMEs (2026)
Page-speed used to be a vague "make it fast" goal. In 2026 it's three specific numbers Google measures on every site visitor: LCP (Largest Contentful Paint), INP (Interaction to Next Paint), and CLS (Cumulative Layout Shift). They directly affect search rankings, and they directly correlate with conversion rates — bounce rates double every 3 seconds of load time. This guide covers what to actually do, in priority order, for a typical UK SME website.
The 3 numbers that matter
LCP (Largest Contentful Paint) — target: under 2.5 seconds
How fast the main visible content of your page loads. Usually that's a hero image or large heading. Above 4 seconds is a clear ranking penalty; 2.5-4 is "needs improvement"; under 2.5 is good. Most slow LCPs come from heavy hero images, slow hosting, render-blocking JavaScript, or poorly-configured CDNs.
INP (Interaction to Next Paint) — target: under 200ms
Replaced FID in 2024. Measures the worst-case responsiveness of the page after a user clicks, taps, or types. Above 500ms is bad. Usually slow INP comes from heavy JavaScript blocking the main thread — third-party widgets (live chat, analytics, social embeds) are the most common culprits.
CLS (Cumulative Layout Shift) — target: under 0.1
Measures how much the page jumps around as it loads. Common causes: images without explicit width/height, ads or embeds injecting late, fonts swapping in (FOIT/FOUT). Above 0.25 is bad. Easy to fix with proper image dimensions and font-display strategy.
Where to start (priority order)
1. Measure first — actually measure
Don't guess. Use Google's PageSpeed Insights (pagespeed.web.dev) on your most-trafficked page. Look at the "Origin" data (real-user metrics from Chrome users) not just the synthetic lab data. If your origin metrics show 75% of visits passing all three Core Web Vitals, you're fine — focus elsewhere. If less than 75%, you have work to do.
2. Optimise images — biggest single win
Hero images are the most common LCP killer. The fixes:
- Convert PNG/JPG to WebP or AVIF — typically 25-50% smaller at the same visual quality.
- Set explicit width and height attributes on every <img> tag — prevents layout shift.
- Add
loading="lazy"to below-the-fold images,fetchpriority="high"to the hero. - Serve responsive images via the
srcsetattribute so phones get phone-sized images. - If using Next.js, the <Image> component does all this automatically.
3. Audit your JavaScript bundle
Heavy JS = slow INP. Common offenders for UK SME sites:
- Third-party live-chat widgets (Intercom, Drift, Crisp) — typically 100-300KB. Load on user interaction, not on page load.
- Social media embeds (Twitter, Instagram) — heavy. Replace with cached static screenshots that link out.
- Cookie consent banners — can be 50-150KB. Pick one that's lightweight (Cookiebot is heavy; Termly is lighter; rolling your own is lightest).
- jQuery or older frameworks — usually replaceable with 5-10 lines of vanilla JS.
- Unused tag manager fragments — review GTM container, remove dead tags.
4. Fix CLS with proper image dimensions and font strategy
For every image, set width and height attributes. Modern browsers use these to reserve the right space before the image loads. For fonts, use font-display: swap in your CSS so text shows immediately in a fallback font while the web font downloads — yes, it causes a visual swap, but a fast swap beats blank space.
5. Hosting matters more than people think
A static site on Vercel or Cloudflare Pages will outperform a WordPress site on shared hosting every time. For a UK SME:
- Static sites (Next.js, Astro, plain HTML): Vercel free tier or Cloudflare Pages. Sub-100ms TTFB worldwide.
- WordPress: managed hosting (Kinsta, WP Engine) costs more but is materially faster. Cheap shared hosting is the slow option.
- CDN: even if you stay on shared hosting, putting Cloudflare in front (free tier) cuts latency dramatically for international visitors.
6. Server response time
Your Time-To-First-Byte (TTFB) should be under 600ms. If it's above, the problem is server-side: slow database queries, uncached pages, or distant hosting. For brochure sites, static generation (build the HTML once, serve from CDN) eliminates this problem entirely.
The 80/20 of website performance for UK SMEs
If you have an hour and a single SME website to fix, do this in order:
- Compress and convert images to WebP. Tools: Squoosh.app (free, browser-based) for one-off; cwebp CLI for bulk; Next.js <Image> for build-time. 20-30 min, biggest impact.
- Add width/height to every image. 10 min, fixes CLS instantly.
- Audit third-party scripts. Open browser DevTools → Network tab → sort by size. Anything >100KB needs justification. 15 min.
- Move third-party scripts to
asyncordefer. Anything not critical to first paint should load after. 10 min. - Re-test PageSpeed. Should see 10-30 point improvement.
What we'd recommend for new builds
For a new UK SME website in 2026:
- Static site or Next.js for the frontend. WordPress only if there's a specific reason (heavy editorial team, existing investment).
- Vercel hosting for static / Next.js (free or £20/month). Or Cloudflare Pages (free, also excellent).
- WebP images via build pipeline (Next.js Image, Astro Image, or manual cwebp).
- Inter or system fonts with
display: swap. Avoid heavy custom fonts unless brand requires. - Minimal third-party JS. GTM/GA4 is fine. Lazy-load everything else.
- Cloudflare in front for CDN + security + analytics.
Tools we use
- PageSpeed Insights (pagespeed.web.dev) — Google's real-user data + lab tests
- WebPageTest.org — deeper waterfall analysis, multi-location testing
- Chrome DevTools → Lighthouse — quick local audits
- Squoosh.app — image compression in the browser
- Cloudflare Speed Brain — built into Cloudflare dashboard
- Vercel Analytics / Speed Insights — if you're on Vercel, free real-user CWV monitoring
When to bring in help
If your PageSpeed Origin score is below 70 for any of the three CWV metrics, the fix is usually not complicated — but it does require knowing which lever to pull first. Most SME owners spend hours fiddling with settings that move the score 5 points when the real issue is one bad third-party widget. A 1-2 hour audit from someone who's done this before usually identifies the 2-3 changes that move you from 60 to 90+.
See our care plans for ongoing performance monitoring (Silver tier and above include monthly Core Web Vitals tracking with regression alerts).
Got a workflow you want to talk through?
30 minutes, no pitch. We'll tell you honestly what we'd build — or whether automation isn't right yet.