Three measurements that describe whether your site feels fast, stable and responsive to a real person.
Three measurements that describe whether your site feels fast, stable and responsive to a real person.
One measures how long the main content takes to appear. One measures whether things move around while the page is loading. One measures how quickly the page reacts when someone interacts with it.
That is the whole idea. The acronyms make it sound more technical than it is.
They correlate with what users experience, which is why search engines pay attention to them. But the ranking effect is the smaller half of the argument. People leave slow sites, and they leave before they can be counted.
Images that are larger than they need to be. Fonts that block rendering. Third-party scripts that nobody has audited in two years. Layout that has no reserved space, so everything jumps when an image arrives.
Most sites get most of the way there by fixing those four things, before anyone touches the framework.
A synthetic test on a fast connection tells you the site can be fast. Field data — what actual visitors experienced on the devices they actually own — tells you whether it is. The two disagree more often than teams expect.
When they disagree, the field data is the one that matters. A perfect lab score on a site that feels slow to your customers has measured the wrong thing.
Oversized images sent at full resolution to a phone. Fonts that block rendering while they download. Third-party scripts nobody has audited in two years. Layout with no reserved space, so everything jumps as images arrive.
Fixing those four gets most sites most of the way. They are unglamorous, they require no architectural change, and they are almost always where the problem actually is.
Framework changes come last, and once the four above are handled they usually turn out to be unnecessary.
Chasing a perfect score has sharply diminishing returns. Once the page feels immediate on a mid-range phone on a normal connection, the remaining points cost more than they return.
Spend what is left on content and clarity instead. No visitor has ever converted because of the last five points on a performance score.
Numbers are abstract until you map them onto behaviour. A Largest Contentful Paint of four seconds is not a statistic; it is a person on a train looking at a blank rectangle, deciding whether your business is worth the wait. Most conclude it is not, and they do not tell you.
A poor Interaction to Next Paint feels like a phone that has stopped listening. Someone taps a menu, nothing happens, so they tap again, and now they have opened and closed it. They read that as a broken site rather than a slow one, which is a harsher judgement and a harder one to recover from.
Layout shift is the one that produces genuine anger, because it makes people act against their own intention. Tapping a link and landing on an advert is a small betrayal, and it is entirely avoidable by reserving space for anything that loads late.
Images first, almost every time. Correctly sized, in a modern format, with width and height attributes so nothing reflows when they arrive, and lazily loaded everywhere except the one image the visitor sees immediately — that one should be eager and, ideally, preloaded.
Fonts second. A web font that blocks rendering costs you the first impression it exists to make. Use font-display: swap, subset to the characters you actually use, and self-host so you are not waiting on a third party.
Third-party scripts third, and this is where the largest wins usually hide. A tag manager loading six tools, two of which nobody has used in a year, can dominate the interaction budget. Auditing that list is free and frequently halves the problem.
Only after those do server response time and framework decisions become the limiting factor. Teams routinely rewrite the architecture before deleting an unused analytics script.
A site can pass every Core Web Vital and still be unusable, because the metrics measure delivery rather than clarity. Nothing in the toolkit notices that your pricing is impossible to find or that the main call to action is the word Submit.
They also flatten the difference between pages. An overall pass can conceal a checkout that fails badly, because the homepage volume drowns it. Segment the field data by template before believing any single number.
Treat them as a floor rather than a target. Passing means you have stopped losing people for mechanical reasons, which is the beginning of the work rather than the end of it.