Web hosting for martial arts schools — A dojo site has three templates, and only one of them is slow

Traffic to an academy site is lumpy and predictable: a slow trickle of parents reading the junior programme page during the day, a spike after a grading or a demo, and a handful of people who found you at 21:00 and want to book a trial before they lose their nerve. Nearly all of that is cacheable. The small part that is not is the part that earns money.

That split matters because it decides where optimisation effort goes. A cached programme page is served without booting PHP at all, so its speed is essentially a network and payload question. The booking POST runs the full stack — PHP, MySQL, whatever the booking plugin does on submit — and no page cache will ever help it.

Get the split clear before you change anything. Otherwise you spend a weekend compressing images on a page that already renders in under a second, and the form that loses you enrolments is exactly as slow as it was on Friday.

A trial-class booking is a POST. It cannot be cached, it boots PHP, it writes to the database, and on many booking plugins it also sends two emails and checks an availability table before it returns anything. That is the request a parent is waiting on with a nine-year-old asking questions behind them, and it is the one nobody profiles.

Measure it the way it is actually used: submit a real test booking with DevTools open and read the timing on the POST, not the page load. If it sits above a second, the cause is almost always inside the plugin rather than under it — an availability query with no index, or a synchronous email send blocking the response. Nitro carries PHP X-Ray on CloudLinux Pro, which traces a slow request down to the function and turns an argument about hosting into a line number.

Change one thing, then retest the same three templates the same way. Lab runs wobble with test-server load and cache state, so take the median of three rather than the best of one, and treat anything you cannot reproduce twice as noise. Staging exists precisely so the change can be measured before a parent meets it.

The verdict, though, comes from the field. Search Console's Core Web Vitals report reads a rolling 28-day window of real visits, so a fix shipped on Monday will not show up by Friday — and a change that improves a lab score while the field data stays flat has improved a benchmark, not a visitor. Keep a dated note of each change and its numbers; after three months the note tells you which of your plugins is worth its weight.