Google Signal for Site Speed: PageSpeed ranking, Time To First Byte, or something else?

[DesignHammer]

We were having an internal discussion regarding what specific signal Google is looking for regarding Site Speed. My understanding was that Google primarily used Time To First Byte (TTFB) as its signal of Site Speed. My colleague argued that this is not part of Google's PageSpeed Insights (https://developers.google.com/speed/pagespeed/insights/) and therefore was unlikely to be the primary signal. Who is right? Is TTFB the primary signal or the score on PageSpeed Insights?

[max.favilli]

TTFB is server time, spent processing the http request. So you have more than a load of things to look after, in order to improve response time, it largely depends on the technology stack you are using. You may need to review your server side code, your sql queries, your infrastructure.

Anyway more often than not is a matter of poor backend coding.

[ninjahippo]

Massimiliano, 
You have said 'While I always religiously try to lower TTFB as much as possible.' Can you specify what you actually do which has a direct effect on your TTFB ?

[RyanPurkey]

Thumbs up to Matt on this. The over all load time of the page is what's critical, although several of the suggestions within the PageSpeed tool are aimed at making the first visible content within the page load as quickly as possible.

[max.favilli]

In GWT (Crawl->Crawl Stats) you can find "Time spent downloading a page (in milliseconds)", that's really TTFB.

I have seen few times a correlation between an improvement in TTFB and ranking. So I am convinced TTFB matters a lot.

PageSpeed is a different story, it does analyze a lot more speed factors, mostly client side. It does analyze TTFB too called "Server Response Time" in Page Speed. How these factors impact ranking I don't know. Some like "mobility issues" are officially getting you out of mobile idex if not fixed.

Some of PageSpeed suggestions are even controversial and debated in google own products forum (like suggestions to inline styles and javascripts).

Personally I try to fix as many PageSpeed suggestions as possible, but applying some common sense too. While I always religiously try to lower TTFB as much as possible.

[MattAntonino]

From what I've seen (strictly untested but across hundreds of slow sites which were then sped up) it's the overall time - so the Developer's Tool results, not TTFB. I would argue that most sites that load between 3-6 seconds would improve their overall time through a speed up in TTFB though, so they'd be related. If your site is really in pain (7+ seconds) I don't think you'd make much of an improvement by getting 1st byte faster if it still takes 10 seconds to load.

So ... your colleague would be right as far as my experience goes. We use Developer's Tool & Pingdom and if we improve the overall speed, rankings improve.