How fast should a martial arts school website load on mobile?

The thresholds Google publishes

Google does not publish a load time. It publishes three metrics, measured on real visits, scored at the 75th percentile — three quarters of your traffic has to clear the bar, not your one good run on a laptop.

Metric	Good	What it is on your site
LCP — Largest Contentful Paint	2.5s or less	How long until the biggest thing on the first screen finishes rendering. On a school site that is the hero image or the hero video, every time.
INP — Interaction to Next Paint	200ms or less	How long the page takes to answer a tap. Opening the mobile menu, hitting Schedule, submitting the trial form.
CLS — Cumulative Layout Shift	0.1 or less	How much the page jumps while it loads. The reason a parent taps the 6am class when she meant the 6pm one.

INP is the one agencies get wrong. It replaced First Input Delay on March 12, 2024. If a proposal on your desk still quotes FID, whoever wrote it has not looked at this in over two years.

2.5 seconds is the published pass mark, and LCP is not the same measurement as “page load time” — it is one moment inside the load, taken across your slowest quarter of visits. Under a second is our target, not Google’s threshold. Passing is the floor. It is not the same as not losing anyone.

What the delay actually costs

53% of mobile visitors leave a page that takes longer than three seconds to load, and as load time goes from one second to seven, the probability of a bounce climbs 113%. That is Google/SOASTA research, measured across mobile sites generally, not schools.

The better-controlled figure comes from Deloitte’s study for Google, run over 30 million sessions on 37 brand sites. A 0.1 second improvement lifted retail conversions 8.4% and travel conversions 10.1%. Retail and travel, measured on paint and latency metrics rather than LCP — read it as a direction, not a rate you can multiply against your trial bookings.

None of this was measured on schools. What survives the translation is the shape of it: the cost of being slow is not a cliff at some magic second, it is a tax charged the whole way down, and a tenth of a second is enough to move it.

What makes a school site slow

It is almost never the design. It is five specific things, and a school site rarely has only one of them.

The hero video. A video element counts as your LCP element — Google times its poster image or its first frame. An autoplaying loop of someone drilling armbars is not atmosphere. It is your slowest metric, and it is the first thing you asked the browser to do.

The mat photography. Straight off the camera, 4000px wide, displayed at 800. CSS background images count toward LCP too, so the tournament shot sitting behind your headline is being timed whether you think of it as content or not.

The page builder. Elementor, Divi, a themed WordPress install. Each ships stylesheets and JavaScript for a hundred features you are not using, and you cannot remove them without breaking the editor you bought it for.

The embedded schedule. Kicksite, Zen Planner, PushPress, Gymdesk and Mindbody all hand you a schedule or booking iframe. It is convenient, and it loads a second website inside your first one. If you keep it, keep it off the homepage.

The widget layer. Chat bubble, Facebook pixel, Google Ads tag, review carousel, and a countdown timer left over from a promo two summers ago. Each is a third party you have handed your load time to, and none of them will give it back.

Lab score versus what Google reads

Test it on a phone, on cell data, standing outside the school — not on the desktop in the office. PageSpeed Insights gives you two numbers and they are not the same number.

The score out of 100 is Lighthouse: a simulation on a simulated device, which moves around when nothing about your site has. Underneath it is field data from the Chrome UX Report — actual Chrome visits to your actual site. CrUX is what Google Search uses to inform the page experience ranking factor. The lab score is a rehearsal. The field data is the performance.

If it tells you no field data is available, that is not a bug. To appear in CrUX a page has to be publicly discoverable and have a minimum number of visitors, and a single-location school often does not clear it. That leaves you staring at the lab score, which is the one that matters least — and a site too quiet to be measured is a description of the problem, not a reprieve from it.

Speed is not a ranking lever

Speed gets sold as an SEO fix. Google is plainer about it than the people selling it: “There is no single signal,” and “Google Search always seeks to show the most relevant content, even if the page experience is sub-par.” Core Web Vitals feed the ranking systems. Good scores do not buy a position.

Which is the more useful way round, because it tells you what speed is actually for. It will not put you in the map pack. It decides what happens to the parent who is already there — who searched at 9pm, tapped your listing, and is now looking at a white screen with a spinner where the kids’ class times should be. She is not comparing your load time to the school down the road. She is going back to the results and tapping the next one.