Core Web Vitals: what to fix and what can wait

Google published a set of metrics today called Core Web Vitals, and named the first three: Largest Contentful Paint, First Input Delay and Cumulative Layout Shift. It also said the set will be updated every year. Read that as Google telling everyone, in advance, what it is going to start measuring sites on.

It lands in an odd week. A core update started rolling out yesterday, so rankings are moving for reasons nobody can see yet. And the website is doing more work for a local business than it has ever done, because it is now the order counter, the booking desk and the front window at the same time.

So this is worth attention. It is not worth a panic, and it is definitely not worth a rebuild. Here is how we are reading it.

The three metrics, in plain terms

• Largest Contentful Paint. How long until the biggest thing on the screen actually appears. Usually the hero photo or the headline block. Google calls under 2.5 seconds good.
• First Input Delay. Someone taps “Book now” and the page has to respond. How long before it does. Under 100 milliseconds is good.
• Cumulative Layout Shift. How much the page jumps around while it loads. This is the metric for tapping the wrong button because a banner pushed everything down half a second after you reached for it. Under 0.1 is good.

Two details matter more than the numbers. First, these are measured on real visitors in Chrome, not on your laptop, and mostly on phones on ordinary connections. The Chrome User Experience Report data behind them is what feeds the Speed report in Search Console and PageSpeed Insights. Second, Google grades at the 75th percentile, which means the bad quarter of your visits is what counts, not the average.

That second point is why sites that feel fine to the owner score badly. You load the site on office wifi on a two year old laptop with everything cached. Your patients load it in a car park on one bar.

Why this is more urgent than it was in February

Traffic to the sites we run looks different than it did three months ago. Far more of it is on phones. Far more of it arrives with intent to do one thing: check hours, order, book, call. And almost every site we touch had things bolted onto it in March in a hurry. A COVID notice bar. A chat widget somebody signed up for in a panic. A popup announcing curbside pickup. An embedded ordering menu.

Each of those is a script, and several of them shift the layout after the page has already drawn. That is exactly what Cumulative Layout Shift punishes, and more to the point, it is what loses a booking. Nobody taps twice.

What we are fixing now

In priority order, because the first two items account for most of the problem on nearly every local site we open.

1. The hero image. The single most common reason Largest Contentful Paint is bad on a small business site is a photograph exported straight from a camera or a stock site and dropped in at full size. Resize it to what it is actually displayed at, compress it, serve a smaller version to phones. This is an afternoon of work and it usually moves the number more than everything else combined.
2. Whatever was added in March. Give the notice bar and the announcement strip a fixed height so the page reserves space for them instead of pushing content down when they load. Same for any embedded menu or map.
3. Third party scripts. Open the site and count them. A chat widget, a popup tool, a heatmap trial nobody has looked at since 2018, two ad pixels and a review slider is normal, and it is too many. Cut the ones not earning their place. Load the survivors after the page is usable, not before.
4. Fonts. Two weights, not six. Preload the one used in the headline, and let text show in a fallback while the font arrives rather than sitting blank.
5. Booking and form embeds. If the booking calendar or lead form is in an iframe, give it a set height so it does not resize the page under the visitor. Do not lazy load the form itself. People came for it.
6. Hosting. A fair number of local sites sit on cheap shared hosting carrying a plugin for everything. Caching and a decent host fix more than any single code change, and they cost less than a redesign.

What can wait

A rebuild. If the site converts, do not throw it away over a score. Fix the images, cut the scripts, measure again.

A perfect score. Chasing 100 in a lab test is a hobby. Field data from real visitors is what Google says it is looking at, and it responds to much duller work: smaller images, fewer scripts, better hosting.

Removing things that earn money. The click to call bar, the booking widget and the review block can all stay. Make them load late, do not delete them.

Treating this as a ranking project. Google has not said these metrics change rankings. Our read is that it is signalling where it is going and giving everyone a year of notice, and it would not surprise us if it made that official later. Either way, we would do this work anyway, because a booking page that takes six seconds on a phone loses bookings today, independent of what Google does with it.

What to do this month

• Run your own site through PageSpeed Insights on mobile and write down the three numbers. That is your baseline.
• Open the site on your phone, off wifi, and try to book or order. Note every place it jumps or stalls.
• Fix the hero image first. Resized and compressed, nothing clever.
• List every third party script on the site and delete the ones you cannot name a reason for.
• Give the March additions, the notice bar and any embed, a reserved height so nothing shifts.
• Check the Speed report in Search Console at the end of the month and compare.

None of this is glamorous, and that is the point. Most of the speed problems we find on a local business site are four images and three scripts. If you want a hand with it, this is the boring half of how we build and fix websites, and it sits right next to the local search work because the same visit that fails to load is the one that was going to call. And if most of your traffic arrives through the map rather than the homepage, get the listing pointed at a page that loads before you spend anything on the rest.