Speed is revenue: what 0.1 seconds does to a webshop, and why most sites fail the test

In 2020 Google and Deloitte ran the cleanest study on this that exists: across 37 retail, travel, luxury and lead-generation sites, a 0.1-second improvement in mobile load time lifted retail conversions by 8,4% and average order value by 9,2%. Travel conversions moved 10,1%. (Milliseconds Make Millions) Nothing has changed since except that the sites got heavier.

That is the whole business case. A webshop doing €50.000 a month that gets 0,3 seconds faster on mobile is, on those numbers, a webshop doing €60.000 a month. No new traffic, no new ads — the same visitors, fewer of them giving up.

What Google actually measures

Core Web Vitals are three numbers, measured on real visitors' devices, not in a lab:

• LCP (Largest Contentful Paint) — how long until the main thing on the page is visible. Target: under 2,5 s.
• INP (Interaction to Next Paint) — how long the page takes to react when someone taps. Target: under 200 ms. This is the one page builders fail, because every tap runs through a pile of JavaScript.
• CLS (Cumulative Layout Shift) — how much the page jumps around while loading. Target: under 0,1. Late-loading banners, fonts and cookie walls are the usual culprits.

A page "passes" when the 75th percentile of visitors hits all three. Industry-wide, only around four in ten e-commerce sites pass on mobile. Desktop is kinder; mobile is where the orders are.

How to read your own numbers in ten minutes

1. Open Search Console → Experience → Core Web Vitals. This is field data, the only data that matters. If it says "poor" on mobile, you are losing orders today.
2. Run the worst URL group through PageSpeed Insights. Ignore the score at the top; read the three field numbers and then the "diagnostics" list.
3. Look at what loads before your product image: tracking scripts, chat widgets, fonts from three different CDNs, a slider library for one slide. Every one of those is a decision someone made without seeing the bill.

What actually fixes it

We have sped up plenty of sites we did not build, and the fixes fall into a short list.

• Ship less JavaScript. Not "optimise" it — remove it. A product page needs a gallery, a cart button and variants. It does not need a 400 kB builder runtime to draw a heading.
• Render on the server. The customer should receive finished HTML, not an empty shell that assembles itself. Everything we build works this way; on our own site the entire ASCII galaxy is layered on top of a page that is already complete.
• Do not reload the page to show a product. On the butcher's ordering site we are finishing, a product opens as a sheet over the menu, with the URL updating underneath. Zero page loads between browsing and basket, and every product still has its own crawlable page.
• Images sized for the slot they sit in, in modern formats, with the first one preloaded and the rest lazy. This alone is often the difference between a red and a green LCP.
• Fonts: one family, self-hosted, swapped in. Third-party font CDNs are a round trip you pay on every visit.
• Reserve space for anything that loads late — banners, embeds, consent bars — so nothing jumps.

Fix or rebuild?

If the site is a custom build with a few bad decisions, fixing is a week's work and worth it. If it is a page-builder theme with thirty plugins, you can spend that week and land at "needs improvement" — the runtime you cannot remove is the problem. That is the honest moment to price a rebuild against the revenue you are leaving on the table, using the numbers above. We will do that maths with you for free; it is usually not close.

The part nobody puts in the pitch deck

Speed is not a project, it is a discipline. The fastest site in the world gets slow the day marketing adds a fourth tracking pixel and a chat bubble. We set a performance budget at launch — a maximum page weight and a maximum script budget — and we measure it monthly, in the same report as the rankings and the revenue. That is what "SEO as an ongoing project" means when we say it.