The first text can appear while the main image is still loading. Speed tests need to show both how soon the host replies and when the page is ready to use.
A server can reply fast while the page still feels slow. That is why we look at more than one number when we talk about speed at HyperWP. Server timing and browser timing describe different parts of the same page load. Both are needed to understand where delays occur.
The moment the page feels ready
Largest Contentful Paint, or LCP, is a useful place to start. It tracks when the largest block of text, image or video that qualifies for the measure is drawn in view.
That may be a big heading or the main photo. It gives a good clue to when the useful part of the page has arrived. It does not mean every last thing has loaded. Google’s good target for LCP is 2.5 seconds or less.
Speed is more than the first load
A page also needs to respond when you use it. And it should stay put. A button that shifts as you tap it is a layout fault, even if the page loads quickly.
Google’s Core Web Vitals cover those three parts of the visit:
| Measure | What it tells you | Good target |
|---|---|---|
| LCP: loading | When the main content comes into view | 2.5 seconds or less |
| INP: response | How soon the page paints a response to an action | 200 milliseconds or less |
| CLS: stability | How much the layout shifts when you do not expect it | 0.1 or less |
INP stands for Interaction to Next Paint. CLS means Cumulative Layout Shift. Its score is not a length of time. These checks show when the page loads, how it responds to input and how much its layout shifts.
Each measure should meet its target at the 75th percentile. Put simply, at least three in four visits with data for that measure should fall in the good range. Phones and desktop devices get their own checks. All three measures must pass. Google explains the thresholds here.
Where the server comes in
Time to First Byte, or TTFB, is the wait for the first byte of a reply. It can include the trip across the network and the time to set up a link. It is not just work on the server. Google suggests 800 milliseconds or less as a rough guide.
Server response time affects when the browser can start work. Less delay at the start leaves more time for the browser to build the page. Our CPUs, NVMe storage and WP Accelerate cache each have a part to play.
First Contentful Paint, or FCP, marks the first visible text or image content. Its good target is 1.8 seconds or less. It marks the first content on screen. LCP tracks when the largest block that counts for that measure comes into view.
Lab tests and real visitor data
A Lighthouse score is useful. It is also one lab test with its own set of rules. It cannot show how every visitor’s device will perform. Nor does it give you INP from real visits. Scores can change between runs.
For the site as a whole, we would put LCP, INP and CLS first. TTFB and FCP help explain the result. When testing the host, TTFB helps show how long a reply takes. The page, test location and cache state should be the same on each side.
The goal is a page that loads promptly, responds to input and stays stable while people use it.