PageSpeed Insights: speed as Google sees it
A free tool measures your site the way Google measures it: real visitor data on one side, a lab test on the other. What the scores mean, and what really counts. Tools series, part four.
A slow site loses twice: visitors who give up, and an engine that takes it into account. To find out where you stand, no need to buy anything: PageSpeed Insights is Google’s own tool, free and without an account. Paste in an address, wait thirty seconds, read.
Two very different measures on the same page
The report layers two things the documentation distinguishes carefully, and which have to be read separately.
At the top, real user data: what the people who visited your site with Chrome experienced over the last 28 days. That is the measure that counts, because it includes everything a laboratory doesn’t simulate, the average phone, the network on a moving train, the 4G of a building site. Small site, small traffic: this section can be empty, for lack of a sufficient sample. That is not an anomaly.
Below it, the lab test: a simulated visit under fixed conditions, which produces the famous score out of 100. Its value lies elsewhere: since the conditions never change, it lets you compare before and after a modification, and its list of diagnostics says precisely what to fix, almost always the same suspects, oversized images and useless scripts.
The three vital signs
In the real data, three measures carry the name Core Web Vitals, and Search Console devotes an entire report to them: LCP, the time it takes to display the largest visible element, in plain terms the moment the page looks loaded; INP, the delay between a user’s gesture and the page’s reaction; CLS, visual stability, that is, the fact that the text doesn’t jump around while you read it because an image or an ad is settling in. Each measure is rated good, needs improvement or poor, with public thresholds.
What speed weighs, honestly
It has to be said without fanfare: speed is one ranking factor among many, and a modest one. A fast page will never overtake a slow page that answers the question better. But the reverse is true too: with equal answers, slowness costs, and above all it costs before ranking even comes into play, in the visitors who abandon. So you fix speed for people, and the engine follows.
The useful move: test your three most visited pages, on the mobile tab first, since that is where everything plays out. If images are the report’s first complaint, that is the work with the best yield, and it joins the quick move already described in “Your images have a name”: images that are well named, well described and well compressed serve the same goal.
Next part, the last in the series: assembling these tools into a one-hour monthly routine, so that measuring becomes a habit rather than an event.
Who writes these notes
This journal is kept by the workshop that designs and maintains the house’s websites. Everything described here, the Search Console, internal links, the business profile, is part of the work delivered with a site: if you would rather someone took care of it, that is precisely the trade.