Definition of FID vs Aggregate FID (GSC vs Pagespeed Insights)

[elos42]

Hi, I am trying to get by head around the definition of 'Aggregate First Input Delay' as given in GSC, and how it differs from the FID number given in the 'origin summary'.

The GSC number is way too high for me (330 ms), while the origin summary says 92% of the loads happen inside 100 ms. Both are referring to FID or First Input Delay. A similar discrepancy is also seen in the FCP date, with PI giving more flattering numbers.

You can see the data here:
[drive.google.com...]

The definition given by GSC for its Agg FID is as follows: "Aggregate FID is the time it takes for 95% of the visits to a URL in this group to respond to the first user interaction on that page."

The definition given by Pagespeed Insights is 'X% of the visits to this page have a fast FID of under 100 ms'. However, PI also says it doesn't have sufficient real-world speed data for the page. So how is it able to offer this data for the page? Is it, instead, offering data for the origin as a whole? If so, how come 92% of the visits to my websites have fast FID, but nearly 50% of the pages have been flagged as having an average/aggregate FID of 330 ms? Could we be talking about different time periods?

[edited by: Robert_Charlton at 10:41 am (utc) on Nov 12, 2019]
[edit reason] Fixed broken link. [/edit]

[Robert Charlton]

elos42, I see you posted on this issue several times in the Nov 4 thread announcing the new speed report features, so I thought it might help everyone to have that context. Here's the thread...

Google Rolls Out Search Console Speed Report
https://www.webmasterworld.com/google/4971238.htm [webmasterworld.com]

The thread includes a link to the Google webmaster blog announcement, to explain Google's general approach... and I see that you posted several times about your particular issues and questions. Here's an excerpt from one of your posts which lays out some of your key issues...

I think I may be paying a penalty for the fact that 90% of my customers come over mobile connections in India, which can be quite unpredictable. So, if Google's going by load times on those mobiles, and compare those times to a 'global benchmark', the load-times for my site would be very high.

On the other hand, if load times are benchmarked against 'national averages', I wouldn't do so bad. Similarly, if mobile download times are compared with mobile download times, instead of a blended average, I wouldn't do so bad either. I don't know how they do this.

As far as I can see, sites whose consumers are located in high-speed countries such as Korea, Japan, USA and EU should automatically be marked fast, compared to an identical site on the same host targeting a country like India or Pakistan, which would be marked slow.

Google discussions about page speed have suggested that they've definitely wanted to compare apples to apples... which I'm assuming included comparisons within similar geo areas.

With insufficient "real world speed data" for your site, which is what the report notes, it's not clear, as you've observed, what user-base they are comparing you with, or how they determine where your user-base is. A Google core philosophy seems to be that they do want to keep data localized rather than "global", and therefore probably not make assumptions about international speed. A site in a high speed country, eg, might not be designed well for speed, or it might have lots of iframes, etc etc... so global comparisons become quickly meaningless.

As I remember some very early discussions, which may well have changed since Google's beta testing, this will not be like a horse race. As state-of-the-art-progresses, Google will be looking for outliers, sites which fall at the extreme low (or high) ends. I suspect that it may take them a while to get sufficient data... and except for extreme outliers, it will be a while before load speed affects rankings. I wouldn't worry too much about this at this stage.

The following Google Developers article, which goes into some detailed approaches for tracking FID, may be helpful...

First Input Delay
[developers.google.com...]

[elos42]

With insufficient "real world speed data" for your site

Thank you for your effort.
For the record, PI is complaining of "insufficient real world speed data" for that page, not my site as a whole. Earlier, it would make it clear that given that real world data about that page is not available, it is giving data about the site.
Anyway, it looks like both PI and GSC seem to be talking about two sets of data -- perhaps the period of sampling is different.

[not2easy]

I think it is a tool that can help webmasters spot instances of actionable information in the "test", not some number that is permanently assigned to the page or site that was tested. The results, good or bad are only useful to you and only for the time that you ran their analysis. I think they are clear that their tool "doesn't have sufficient real-world speed data" to seriously evaluate speed. It can help you spot things you might not otherwise see, but absent that, it is a number to ignore and go on with your work. ;)