The Core Web Vitals work that actually moves rankings
Most performance advice is a list of 40 optimisations. Four of them do almost all the work. Here is which, and how to tell if they apply to you.
Performance audits produce long lists. Long lists produce paralysis. In practice, four fixes account for most of the improvement on most sites, and the rest is marginal.
Largest Contentful Paint is usually one image
On the majority of sites, LCP is the hero image or the headline. If it is an image, the fixes are: serve it in a modern format, size it correctly for the viewport, preload it, and never lazy load it. That is frequently the entire LCP problem.
Cumulative Layout Shift is usually missing dimensions
Layout shift comes from elements that arrive after first paint without reserved space: images without width and height, web fonts swapping, injected banners and ads. Set explicit dimensions, use font-display swap with a metric compatible fallback, and reserve space for anything injected.
Interaction to Next Paint is usually third party scripts
INP measures how long the main thread is blocked when someone interacts. The blocking is rarely your code. It is the chat widget, the analytics stack, the heat mapping tool and the consent manager all initialising at once. Defer what you can and question whether you need all of them.
Time to First Byte is usually the host
If TTFB is above 800 milliseconds, no amount of front end work will save you. That is a hosting, caching or database problem, and it needs solving first because everything else is measured from it.
How to tell what applies to you
Field data beats lab data. Lighthouse tells you what could be slow on a simulated device. The Chrome UX Report tells you what is actually slow for real visitors. Start from field data in Search Console, find the single worst metric, fix that, then measure again before touching anything else.
Frequently asked
They are a ranking signal, but a modest one compared with relevance and links. The stronger argument is commercial: faster pages convert better, which matters regardless of what Google does with the signal.
Under 2.5 seconds for 75% of real visits. Measure it in the field through Search Console rather than from a Lighthouse run on your own laptop.
Want this handled for you?
We build and run the sites we write about. Tell us the target and we will scope it.
Start a project
