core web vitals a parma: what they measure and how i measure them

dev · 2026-10-06 · 3 min

every agency site in italy has the word "veloce" somewhere in it. fast, modern, optimized. i have never seen one put a number next to the word, and after measuring a few out of curiosity i understand why: the numbers would embarrass them.

so this is the number side of the pitch. what core web vitals actually are, how i measure them on this site, and what they are worth when someone quotes you a new website.

what they are

three measurements google takes of real visits:

  • LCP — how long until the main content of the page is visible. good is under 2.5 seconds
  • INP — how long until the page reacts when you tap something. good is under 200 milliseconds
  • CLS — how much the layout jumps around while loading. good is under 0.1, which in practice means: it does not jump

they are not an seo ritual. they are the difference between a page someone waits for and a page someone reads. on a phone, on a bad connection, the fast site is the one that gets the contact.

how i measure them

two tools, both free: pagespeed insights for the lab numbers, and the chrome user experience report for what real visitors actually experienced. the lab number is what you can fix today; the field number is the truth that arrives months later. anyone who shows you only the lab number is showing you the easier half.

what i did on this one

this site is the first case study, because i can publish its numbers without asking anyone's permission. measured today, in the lab, on a cold load:

  • the html document is 16 kb, and it is complete before any javascript runs: the build prerenders one document per route and language — the sitemap counts them on every build
  • first byte in 227 ms, page fully loaded in 658 ms
  • layout shifts: zero. nothing on the page moves after it arrives
  • the build itself fails if a contrast check fails: 50 cases measured on every build, not promised in a pdf

none of this is exotic. it is a handful of decisions made once, at the start: prerender instead of client-only rendering, one variable font instead of five, images sized before they are inserted. the speed of a site is decided at the architecture table, not in the "ottimizzazione" phase at the end, which is the phase that does not exist because the budget is over.

what it means if you are buying a site

when someone quotes you a website and promises it will be fast, ask for three numbers: LCP, INP, CLS, measured on the finished site, not on the demo. if the answer is a smile, you have your answer.

a fast site does not just rank a little better. it converts: every second of waiting is someone closing the tab before reading what you do. and the fix is rarely a redesign — it is how the site is built, which is why i measure it in the build and publish the result, instead of using the word "veloce" and hoping you will not ask.

the part nobody budgets for

a fast site is not a phase at the end of a project, it is a set of choices made before the first line of code: render on the server, ship one font, size images before they are inserted, do not load a script to display a paragraph. on a small site those choices take a day. on a site that has already shipped, undoing them takes a project. this is why i ask for the numbers before the design, not after the invoice.

faq

what are the core web vitals?

three field measurements google uses for real visits: LCP (how long until the main content is visible), INP (how long until the page reacts to an interaction) and CLS (how much the layout moves while loading). the thresholds are 2.5 seconds, 200 milliseconds and 0.1.

do core web vitals affect ranking?

they are a tiebreaker, not a magic factor. a slow page with the wrong answer still loses to a fast page with the right one. what they affect more reliably is whether someone stays long enough to read the answer, and that is the part that shows up in contacts.

how long does it take to improve them?

days for images, fonts and scripts that block rendering; weeks when the rendering model itself has to change. measure before and after with the same tool, on the same page, on a phone. a number you cannot reproduce is not a result.

for the service side in parma, see realizzazione siti web a parma and ux designer a parma.