Is your site slow for customers? Five checks before you pay for 'optimization'

Site speed work has a reputation problem: it is the category where money disappears into "optimization" with nothing measurable coming back. Partly that is dishonest vendors; mostly it is skipping the diagnosis. Here are five checks that tell you what — if anything — is actually slow for actual customers.

1. Measure where your customers are

Run PageSpeed Insights (or any lab test) with the location your customers actually browse from. A site that loads in a second from Frankfurt and in five from a regional city has a geography problem, not a code problem — and CDN decisions follow from that answer. One test, one big decision clarified.

2. Look at real users, not the lab

Lab tests are synthetic; the "field data" section of PageSpeed (CrUX) shows how real visitors experienced your pages over the last month. If lab says 1.5s and field says 4s, your cache is not reaching real users — a far more useful fact than any generic advice.

3. Count what each plugin puts on the page

Open a typical product page and list the CSS and JS files with their sources. In WordPress, each file traces back to a plugin. It is common for a single page to pull dozens of files — sliders on pages without sliders, forms on pages without forms. This audit alone usually finds the top three offenders in ten minutes.

4. Unload with a safety net

The natural next step — turning off the identified bloat — is where people fear breaking the site. That fear is healthy: the safe version of this operation is trial-based. Turn an asset off, check the page for ten minutes, automatic rollback if anything looks wrong, then commit. Our Load Guard is built around exactly this pattern (per-page asset lists grouped by plugin, a safe-mode trial with auto-rollback before anything sticks), but the principle needs no plugin: change one thing, verify, keep it only if the page survived.

5. Check the tail, not the average

An average load time of two seconds can hide that one in ten visitors waits ten — usually mobile users, usually the ones about to pay. Look at the 75th and 95th percentiles. Speed work aimed at the tail (the checkout page on a phone) reliably pays; speed work aimed at the average often optimizes what was already fast.

Run the five checks and you will usually find one of three answers: geography, plugin bloat on specific pages, or an uncached backend. Each has a known fix and a measurable before/after. And if someone offers to "optimize your site" without doing these checks first — that conversation is where your money goes to disappear.