In 2019 I wrote a post called Score 100 on Google PageSpeed Insights. It walked through the report, explained the six numbers at the top and offered some sensible advice about not panicking.

I reread it this week. Four of those six numbers no longer exist.

So here's the update. What the score measures now, what it takes to hit 100, and whether you should bother.

What's gone

The 2019 report measured First Contentful Paint, First Meaningful Paint, Speed Index, First CPU Idle, Time to Interactive and Max Potential First Input Delay.

Only First Contentful Paint and Speed Index are still in the score. The rest were retired, mostly because they were measuring the browser's experience of the page and not yours. "First CPU Idle" was a real thing that real people were asked to care about.

I also spent a paragraph in 2019 explaining a "relatively new" image format called WebP, and warning that only four of the six most popular browsers supported it. Every current browser has for years. It's now the old, safe choice, and AVIF is the newer one people write cautious paragraphs about.

What it measures now

The performance score is built from five lab measurements, weighted like this:

  • Total Blocking Time, 30%. How long the page is too busy running JavaScript to respond to you.
  • Largest Contentful Paint, 25%. How long until the biggest thing on screen, usually the main image or headline, has appeared.
  • Cumulative Layout Shift, 25%. How much the page jumps around while it loads. This is the one where you go to tap a button and an advert shoves it down the screen.
  • First Contentful Paint, 10%. How long until anything at all shows up.
  • Speed Index, 10%. How quickly the visible part of the page fills in.

Look at the top three. Eighty per cent of the score is now about three very human complaints: it's ignoring me, I'm still waiting for the main bit, and it moved. That's a real improvement on 2019.

The colour bands haven't changed. Under 50 is red, 50 to 89 is orange, 90 and up is green.

The part most people miss

There are two sets of numbers on that report, and they don't measure the same thing.

The score out of 100 is a lab test. Google loads your page once, on a simulated mid-range phone with a throttled connection, and marks what it sees. It's useful and it's repeatable, give or take. Run it three times and you'll get three slightly different scores.

Above it are the Core Web Vitals, which come from real people using your site in Chrome. There are three: Largest Contentful Paint (good is 2.5 seconds or less), Cumulative Layout Shift (0.1 or less), and Interaction to Next Paint, which replaced First Input Delay in March 2024 and measures how quickly the page reacts when you tap or click something (200 milliseconds or less).

Those are the ones Google uses as a ranking signal. The 100 is not.

So you can score 100 in the lab and fail with real users, because your real users are on older phones than the simulated one. You can also score 70 and pass comfortably. I've seen both.

What a 100 takes

Less than you'd think, and that's the hard part. A perfect score is mostly a matter of what you leave out.

  • Very little JavaScript. With Total Blocking Time at 30%, this is the biggest lever. A page of plain HTML and CSS starts with a huge advantage.
  • Almost no third-party scripts. Analytics, chat widgets, cookie banners, embedded videos, tag managers. Each one runs on your visitor's phone, and each one is somebody else's code that you can't speed up.
  • A main image that's the right size. Modern format, sensible dimensions, loaded early and not lazily.
  • Width and height on every image and embed. That reserves the space, so nothing jumps.
  • Fonts that don't hold up the text. Fewer of them, hosted yourself, or just use what's already on the device.
  • Decent hosting. A slow server caps everything else.

A simple static page does all of that almost by default. That's why a developer's personal site scores 100 and your company's homepage doesn't. Nobody at the developer's site asked for a carousel.

Marking my own homework

It would be cowardly to write this and not test my own sites. So I ran both through PageSpeed Insights on a Saturday evening, on mobile, which is the harsher of the two tests.

My portfolio, iamjamie.co.uk, scored 97. It's a small static site with no blocking JavaScript and nothing that jumps about, so it gets most of the list above for free. The three points it dropped were on Speed Index.

This blog scored 53.

Not because of JavaScript, as it happens. Total Blocking Time was zero. The damage was a main image that took 35.7 seconds to arrive on Google's imaginary phone, and a layout shift score of 0.522, against a target of 0.1. That's exactly the "right-sized image" and "reserve the space" items from the list above. I wrote the advice and my own homepage ignored it.

Neither site has enough visitors for Google to report real-user numbers, so the lab score is all I've got. I'm going to fix the image and leave the rest. I'll let you know what it does.

Should you chase it?

My answer in 2019 was that it's about your users, not your score. That's the one part of the old post I wouldn't change.

Getting from 60 to 90 usually makes a real difference to real people, and it's nearly always worth doing. Getting from 95 to 100 usually means removing something the business needs, such as the analytics that tells you whether the site is working at all.

Every point between 90 and 100 is a trade. The chat widget is costing you four points and bringing in leads. Keep the widget.

The question to ask a client, or yourself, is what each script is earning. If nobody can answer, that script can go, and your score will thank you. If the answer is "about a third of our enquiries", that's worth the orange number.

Seven years on

The metrics changed. The format I warned you about became the boring default. The advice at the end is identical.

A perfect score proves your page is fast for a robot holding an imaginary phone. Check the real-user numbers first. Then go after the points that someone will actually feel.

Keep reading

CSS animations: how speed is everything
Nobody has ever praised your dropdown animation, no-one ever will.