Confirmed: Google Site Speed Data Fed by Chrome Users

[robzilla]

Not much of a surprise, but always good to get official confirmation:

The short version is that Google is now using performance data aggregated from Chrome users who have opted in as a datapoint in the evaluation of site speed (and as a signal with regards to rankings).

[...]

John [Mueller] has confirmed this information in a Google Webmaster Central Hangout [15m30s, in German], where he explains they're using this data along with some other data sources (he doesn’t say which, though notes that it is in part because the data set does not cover all domains).

Google Confirms Chrome Usage Data Used to Measure Site Speed [moz.com]

[Travis]

This is not surprising. I always assumed they were doing so since a while.

[jambam]

Using user data like this shouldn't be a "secret" or something that eventually gets confirmed. Should be common knowledge for all chrome users that google goes as far as to track/spys on every little detail of their browsing exp and then uses that data for their own use.

[Travis]

Should be common knowledge for all chrome users that google goes as far as to track/spys on every little detail of their browsing exp and then uses that data for their own use.

This is certainly mentioned in the TOS that all users have to accept when installing Google Chrome (or a system which comes with Google Chrome). But you know, as long as regulators are not imposing to show such or such thing in a clear way, or to obtain explicit consent, big companies will do the very strict minimum in term of information.

Also, Google can argue that you can disable this from the settings. In Settings > Advanced > Privacy & Security, there are several options. Two of them "might" be related to this:

- Automatically send some system information and page content to Google to help detect dangerous apps and sites
- Automatically send usage statistics and crash reports to Google

[robzilla]

It's a pre-checked checkbox that appears under the TOS when you download Chrome. I always uncheck it, but most people probably don't.

Using user data like this shouldn't be a "secret" or something that eventually gets confirmed. Should be common knowledge for all chrome users that google goes as far as to track/spys on every little detail of their browsing exp and then uses that data for their own use.

I don't think it's really been a secret that Chrome collects the data, though I'm sure many users are not aware of it. The confirmation was just regarding the usage of that data to improve the SERPs, and site speed in particular.

[MrSavage]

Oh yes, data to improve human life. Hand it over and we can make things better than ever. The reality is the data is used to make Google's products better than anyone else. Who can compete with their data? Easy. Nobody. Our Android phones provide that same advantage. Can't compete with that data pool and with that comes products and intelligence that nobody else can even sniff. I took it for granted that this was a default of opting in rather than having to tick the box to activate it. In a sense this topic is like saying the sun is hot.

[keyplyr]

Anyone think Edge doesn't collect data for Bing?

[Travis]

Anyone think Edge doesn't collect data for Bing?

[privacy.microsoft.com...]
[privacy.microsoft.com...]

[MrSavage]

If Microsoft does it too does that mean it's okay? They do it, so we can do it. And because they do it and we do it, then it's just okay to do it. We do it because we can. I'm sure Edge and their .001% market share will just kill it with their data to create the ultimate product(s).

[Travis]

Bright minds will certainly figure out how to run Google Chrome automatically from servers :

1- very close from a site, to report very fast speed,
2- with artificial bad latency, to hurt a rival site,

[EditorialGuy]

The short version is that Google is now using performance data aggregated from Chrome users who have opted in

With or without the "opted in," the "aggregated" should remove any privacy concerns, IMO.

[robzilla]

Bright minds will certainly figure out how to run Google Chrome automatically from servers

If by "bright minds" you mean fools. Site speed is not that big of a factor, user-side latency is likely factored into the data, outliers will be averaged out in any statistically significant data set, and with the amount of data flowing in they can probably tell when something is off. Time better spent optimizing :-)

[Travis]

@robzilla do not underestimate "bright minds" :)

[keyplyr]

Headless Chrome *is* run from server level.

[analis]

PageSpeed Insights is not very precise.

what is the most accurate tool for measuring the loading speed of pages?

[Travis]

what is the most accurate tool for measuring the loading speed of pages?

What do you mean by "accurate"?

[analis]

precise

[Travis]

There are not such things as an accurate/precise tool to measure the speed of site, since it depends of a lot of external factors, especially network latency and speed. So each tools will give different results, based on the connectivity between your server (as well as the servers of third parts files you are including), and where the tools is hosted.

If by precised you mean detailed report, then you can simply use the Dev tool of Chrome or Firefox, they both have a network pan, showing all the information about each elements, with the waterfall details of the connection transactions.

you can also have more details at site like GTMetrix

[NickMNS]

Precise means that it drills down to a very fine scale, in this case a more precise tool may measure site speed in milliseconds, a less precise tool measure speed in seconds.

Accurate means how close to reflecting reality the tool is. Example:
Tool "A" : precisely measures site speed to be 2237 miliseconds
Tool "B": less precisely measures site speed to be 3 seconds

Now assume the actual load time was 3127 milliseconds. Tool "A" is precise be not accurate and Tool "B" is not precise but it is accurate.

The problem with accuracy is that it can only be determined if one has a means of determining the actual state of the thing being measured a "reference measure". In terms of page speed there are so many factors at play that it is nearly impossible to know what "actual" is. One could of course simply elect one method (eg: pagespeed insights, or any other) as the reference measure to which all tools will be compared.

Philosophically and in practice this is what happens, Radiolab explains it (much better than I could) in this podcast:
[wnycstudios.org...]

[robzilla]

Very good point. There is no such thing as "the loading speed of pages", there are too many variables involved.

Real User Monitoring (RUM) like the Site Speed report in Analytics will give you a good picture of your actual load times.

[keyplyr]

Site Speed report in Analytics will give you a good picture of your actual load times.

Well yes, kinda. There are a couple glitches in that tool though, like the inconsistencies in caching.

[analis]

I wanted to try this technique but I'm not sure of the real benefits.

<style><?php echo(file_get_contents("file.css"));?></style>

the score on pagespeed rises from 94/95 to 99

[Travis]

the score on pagespeed rises from 94/95 to 99

This is normal, external CSS files are blocking and require extra exchanges between the server and the client.

This should improve your "first paint".

In the other hand, it makes your page heavier (the size transmitted) (which also depends of the size of your CSS), but, considering that we are all gziping pages (or using brotli), the extra bytes added to the page, might not be very significant, again all depends of the size of your CSS, and page's content.

A compromise can be to use the PUSH feature of HTTP/2, which would send the CSS file in parallel with your page content. But this feature to be optimal would require a server which has a "cache-aware" mechanism like H2O for example.

[analis]

pages without CSS are 5/15 KB
file.css size 70 KB


I proceed?

[robzilla]

Please start a new thread if you want to discuss specific performance "enhancements". It's off-topic here.

[analis]

ok delete the last message

[brotherhood of LAN]

Slightly OT as I don't know whether DNS lookup time is part of this factor, considering it's unique to everyone.

Check out the uptime and response speeds of your domain's DNS provider, most of the time the domain registrar will provide a free service which is 'good enough', but often enough you can squeeze a bit extra speed if you're willing to pay for it. See [dnsperf.com...] for a list of providers (not affiliated)

[robzilla]

I don't know whether DNS lookup time is part of this factor

Anything that goes into the load time for users is probably part of the "speed factor". If users can't access your site (or leaves early) because your nameservers are too slow, that's a poor user experience.

Moving to a DNS provider with a good anycast network (a CDN for DNS, as it were) could certainly boost performance a little bit, because most web hosts host their nameservers in a single location -- and the one at the top of the list (Cloudflare) is actually free to use.

[Travis]

Yes, I am using Cloudflare, only for its DNS. (Like that the day Cloudflare is down, everybody is down :))

[mosxu]

I wouldn’t trust Google engineers with this. Simply because speed can be open to negative SEO.

Bots can make slow requests to a website and if these bots hack devices with fast connectivity Google will give credit to them and think that your site is slow.

It is really bad, I think that even one IP loading slow with fast connectivity can screw up the quality of your traffic