3 min read
Site speed: which number actually matters?
A PageSpeed score on its own tells you little. The three Core Web Vitals, what they mean, and which one touches your sales.
"What is my site's PageSpeed score?" is one of the questions we hear most. But that single number is of little use without knowing the three measurements behind it. And the same page, measured twice in a row on the same day, can return different scores.
The score is a summary; to make decisions you need to look at the measurements themselves. The three that Google calls Core Web Vitals are:
LCP — when does the page count as "open"?
Largest Contentful Paint measures when the biggest element on screen — usually the hero image or main heading — is painted. That is the moment the visitor calls the page "open". The target is under 2.5 seconds.
The most common causes are uncompressed large images, a slow server response, and scripts that delay the main content. There is also a less well-known cause: entrance animations that start content at zero opacity. The element is not visible until the animation finishes, so LCP is pushed out by the same amount.
CLS — does the page jump?
Cumulative Layout Shift measures how much content moves around while the page loads. Reaching for "buy" just as the button slides down and hitting an ad instead is exactly what this measures. The target is under 0.1; zero is best.
The causes are usually simple and fixable: images without declared dimensions, fonts that load late, and banners that slide in from the top. Another is a temporary placeholder during loading that is much shorter than the real content — when it is replaced, the page jumps.
INP — does it respond when you touch it?
Interaction to Next Paint measures the time between pressing a button and something changing on screen. On a phone this is the most noticeable of the three: the page looks ready, but nothing happens when you touch it. The target is under 200 milliseconds.
The cause is almost always the same: the browser's main thread is busy running JavaScript. Unnecessary plugins, third-party scripts and libraries the page never uses are what you pay for here.
Lab measurement versus real user data
This distinction settles most arguments. PageSpeed Insights shows you two different things:
- Lab: a one-off measurement from Google's server on a simulated slow phone. Repeatable but not real; it varies from run to run.
- Field data: measurements collected from your real Chrome visitors over the last 28 days. It changes slowly, but it is the number you should be making decisions on.
A new or low-traffic site will have no field data, and then the lab measurement is all you have. But if you do have traffic and the two numbers disagree, trust the field data.
Where to start?
- Measure first. Our free Site Audit shows PageSpeed scores and technical checks together.
- Look at CLS. It is usually the cheapest fix with the most direct effect.
- Shrink your images. On most sites this is the single biggest win and needs no code changes.
- Remove unused scripts. Tracking codes that were tried once and abandoned stay on sites for a surprisingly long time.
- Then measure again. An improvement made without measurement is not an improvement; it is a guess.