Website Speed Optimization: Why Hosting Matters

Speed advice that starts with image compression skips the floor. If the server takes half a second to say hello, WebP will not save the bounce. Hosting is time to first byte, cache at the edge, and whether PHP is a recent version on SSD. Then you trim the theme. A2 Hosting Turbo, SiteGround, Kinsta, and a Cloudflare proxy are different attempts to raise that floor.

TTFB is the wait before the first byte of HTML. It includes DNS, TLS, and whatever PHP and MySQL did. Shared neighbors inflate it. A cold WordPress install on cheap HostGator is a different animal than Kinsta on Google Cloud with page cache. Measure uncached TTFB if you care about the server; measure cached if you care about visitors. Search Console’s Core Web Vitals mix both. If TTFB is poor when cache is on, you have a host or a cache-miss problem. If it is poor only when cache is off, you have a WordPress problem the host will still feel during logins and checkouts. Origin first. Hero images second.

A2 Hosting’s Turbo plans exist because LiteSpeed and faster disks are easier to sell than “please use fewer plugins.” NVMe and SSD beat spinning rust; almost everyone claims SSD now, so treat it as table stakes. LiteSpeed with a compatible cache plugin can outperform Apache-plus-plugin stacks on shared. SiteGround’s stack and SG Optimizer are the other common performance story on shared. These are real mechanisms. They are not a substitute for a 3 MB homepage. If a host only talks about “unlimited bandwidth” and never about PHP version or cache, they are not selling speed. Believe productized stacks more than a generic “we are fast” badge. Then re-measure after you enable the cache they shipped.

Cloudflare’s proxy can hide origin IP, cache static assets, and sometimes cache HTML if you know what you are doing. In front of Bluehost or Namecheap shared, it is often the highest-leverage speed add-on that does not require a migrate. It will not fix a 2-second PHP generate on every cart request. Set DNS correctly, use a valid origin certificate, and do not orange-cloud mail hostnames. APO or similar HTML caching is powerful and easy to misconfigure on WordPress with cookies. Start with static cache and HTTP/3. If you already pay WP Engine or Kinsta, you may already have a CDN; stacking blindly can double-cache and confuse cookies. One edge, one origin, documented.

Media libraries are where shared plans go to die. Compress, resize, and lazy-load; serve through Cloudflare or a host CDN. Then look at object cache: Redis (Cloudways, many VPS, some managed WordPress) helps WooCommerce and logged-in users where full-page cache cannot. People blame the theme because it is visible. The theme is often a page builder, three sliders, and Google Fonts loaded four ways. That is a hosting bill multiplier. Cut assets until the cached homepage is boringly small. Add Redis when the uncached paths matter. Do not buy a bigger Kinsta plan to compensate for an uncompressed hero video in the header. Hardware is not a compression plugin.

Use Search Console’s Core Web Vitals, a synthetic check from a city near your users, and your host’s own graphs if you are on Kinsta or WP Engine. CrUX is noisy on low-traffic sites; do not panic at a single weak URL. Compare before/after a host move with the same theme. Independent uptime tools are not speed tools. Lighthouse on your laptop is not your customer’s phone on LTE. If a host claims a magic percentage improvement, ask what was cached. Hosts advertise 99.9% uptime; that number is not a speed metric. Speed is distributions: p75 LCP, TTFB, INP. Change one variable at a time—CDN, then cache, then host—or you will not know what worked.

Raise the origin floor with a serious stack (A2 Turbo, SiteGround, Cloudways, Kinsta) and Cloudflare where it fits. Then diet the theme and put Redis on the paths page cache cannot touch. Measure TTFB and field LCP. A slow host with a skinny site still feels slow.