How large a ranking factor is load time?

[Tonearm]

As my website has grown, my page loading time has increased to several seconds in some cases. Could this be affecting my rankings?

[g1smd]

Possibly.

Google flags sites slower than 1.5 seconds as "slow".

I try to aim for better than 2.0 seconds.

[scooterdude]

what checker are you chaps using using

according to pingdom my site is significatly slower than the top sites in category and thats scary, they average 1.5 to 2 secs, me according to pingdom 5 to 7 secs

however, according to uri valet, mysite is faster than the others sites

I switched to vps because of this speed war and I am not pleased with the results, mind you, the other went to full dedicated server in same period :)

[deadsea]

A load time worse than 3 seconds seems to hurt rankings. Google has said they owly penalize the slowest couple percentage of sites. That would be in the 10+ seconds to load range. However, users start clicking back to the SERPs if the page takes longer than 3 seconds to load. The high click-back rates caused by slow load times can have a dramatic negative effect on rankings.

[Globetrotter]

So if your medium sized site attracts traffic from all over the world, you can hardly compete with the locals because they will probably be much faster?

And what about slow dialup connections (they still exists) how are they taken into the calculation?

On my website google doesn't seem to be able to pinpoint the slow pages, at least when I load them they are fast.

It also seems impossible to get below two seconds of you have ads, images and social media buttons (even if they are loaded in the background).

[klark0]

Well Page speed is a factor, but no one knows for sure how much it matters. If you got alot of other things going in your favor, then it probably won't make much a difference. In highly competitive spaces though, every little bit matters.

I just spent the holiday yesterday, switching over all my static objects (images, css, js, downloadable) to a CDN. My load time is now between 1.5 and 3 seconds depending on where in the world you are. There's a few other reasons to do it besides SEO, so I highly recommend it. And CDNs are surprisingly cheap. I have alot of css images, so I plan on doing a sprite or two ..that should shave some more time off.

I also took the time to 100% validate every single page on W3 for HTML5. Validation almost certainly doesn't matter, but it's nice to have ...especially since one of my competitors have gone to the trouble of doing the same.

[MarvinH]

I got obsessed with speed, too. GWT reported pages on my site loaded on average in much longer than 3 seconds. It took me many months of learning and changing things around on my site, and finally GWT reports the pages load on average in 0.8 seconds, or faster than 94%. Yet, I have not seen it having any impact on SERPs, over all traffic, etc.

Speed is over rated. If you pages load in about 3 seconds, I'd say it is just fine. Go do something more productive than obsessing yourself with speed.

[netmeg]

I get different ratings from every tool I try - in fact the one in GWT says one of my sites is slower than 75% of sites, but the site screams for me, and the page speed checker at googlelabs says it's 94/100.

What's a person to believe, and how to get a really reliable take on it? I sure don't know.

[tedster]

You can install boomerang.js [yahoo.github.com] and measure your visitor's loading speed directly. I think that's more actionable than anything a web tool can give you.

[netmeg]

Thank you, I shall jump on that.

[scooterdude]

me too thks

[jabas]

Guys you can also try cloudflare, their pro plan is awesome (on top of caching and distributing your content from multiple locations they can also optimize images, implement js hacks from their side like lazy image load, preloading the page before the media content, etc. etc.). I have a dedicated server + nginx + this for my biggest site, and I can say it's awesome, handling #*$!,#*$! visitors per day with a ridiculously low cpu usage (+ this server serves a lot of static content at the same time).

Plus it doesn't seem to affect google rankings, I use them for months and everything is fine.

PS: I'm not a cloudflare rep :D

[robzilla]

Joshua Bixby recently wrote a fairly definitive blog post on site speed and SEO: [seomoz.org ]. A good introduction if you're new to this.

[MarvinH]

Joshua Bixby recently wrote a fairly definitive blog post on site speed

Jashua Bixby owns a company which sells site acceleration products. Any article written by him will try to convey the message "Speed Is Important". His business depends on people believing it.

I'm not saying speed is not important. It is. But its importance has been overblown in the past year or two.

[robzilla]

If you actually read the article, you'll find that he is by no means overblowing the importance of site speed as a ranking factor. Rather, he answers some of the most common questions (and misconceptions) that exist on the topic, concluding that it's probably a relatively small factor (as I believe it should be).

[Jez123]

I get different ratings from every tool I try - in fact the one in GWT says one of my sites is slower than 75% of sites, but the site screams for me, and the page speed checker at googlelabs says it's 94/100.

What's a person to believe, and how to get a really reliable take on it? I sure don't know.

Me too. I had someone optimise the hell out of my site. Moving all images etc to cloud, as well as sprites and reducing all requests down to an absolute minimum and GWY reports my site as slower than before!

Does anyone know what they take into account? Is it an average all over the site over a monthly period or something else?

It's very frustrating! On pingdom it's loading under 2 seconds. And that's with a very slow first byte time. Currently in WMT it's reporting 6.6 seconds.

[MarvinH]

If you actually read the article

I actually have.

you'll find that he is by no means overblowing the importance

I pointed at his business motivations, and that the whole speed thing has been overblown in the past year or two.

[scooterdude]

@Tedster

I can't get boomerang to work

Lognormal Test pages work , but the scripts don't appear to fire once copied to other html or .aspx files

[tedster]

it's probably a relatively small factor (as I believe it should be).

I'm with you - the direct page load number should have a small weight. If there is a strong effect on the user experience, that indirect effect will show up in other ways. And depending on HOW the page loads, it certainly can.

[robzilla]

On pingdom it's loading under 2 seconds.

Pingdom is an uptime monitoring service. While its load time features are helpful, they are by no means representative of what your users are experiencing. The Pingdom probe servers are located in datacenters, where connections are much faster, and physically closer to the major internet exchanges, than what your users have at home. To illustrate: if Pingdom pings my server from the country it is located in, the RTT is a mere 2ms (a bit like pinging localhost). Not even people with a fiber-optic cable connection will see that kind of latency. If you are going cross-country, or even cross-continent, expect those figures to be much higher, especially in countries where internet connections tend to be slower. And the heavier your site is, the bigger the effect of that latency will be, as resources travel to and fro in small packets. For a more accurate simulation or tracking of page load times, have a look at [webpagetest.org ] (simulation) and [torbit.com ] (tracking).

The figures in WMT are therefore more likely to be accurate, because they represent real users' load times.

[Sgt_Kickaxe]

Make sure you test cross-browser too, I recently had an issue that caused Chrome users to experience 3x normal load times.

[netmeg]

The figures in WMT are therefore more likely to be accurate, because they represent real users' load times.

You'd think so, but they don't even agree with Google's own tool, in other iterations.

[deadsea]

Don't forget that browser extensions like AdBlock and RequestPolicy have a dramatic effect on page load time. My pages load in 800ms with AdBlock enabled and 2.5 seconds with AdBlock disabled. And when it comes to the ads slowing down your site, there doesn't seem to be much you can do short of removing the advertising completely.

[robzilla]

You'd think so, but they don't even agree with Google's own tool, in other iterations.

Which tool would that be?

[tedster]

Unless netmeg tells us otherwise, I'm assuming Google's Page Speed Insights [developers.google.com] which grew from their original browser plug-in [webmasterworld.com].

[netmeg]

Yep.

[robzilla]

That tool is useful for insight, certainly, especially now that they've added the Critical Path Explorer, but again it's not an accurate representation of what real users experience. For one, these tests appear to be performed from the U.S. and so one European site which I've just tested loads in about 900ms, whereas another U.S.-based site, which is in fact heavier than the European site, loads in less than 150ms. Yet there's no way a DSL user in the U.S. will load the site in 150ms.

The statistics in WMT, however, are based on actual real-user measurements. If only they were updated more often! Per-country statistics would be helpful too.

[scooterdude]

AFAIK

Unless netmeg tells us otherwise, I'm assuming Google's Page Speed Insights [developers.google.com] which grew from their original browser plug-in [webmasterworld.com].

That tool does not measure actual speed, it assigns score based on the elements that make up the page, so, a valid html page with a few scaled images, loading from a server behind dialup connection might well score 100 :)

[MarvinH]

The statistics in WMT, however, are based on actual real-user measurements. If only they were updated more often! Per-country statistics would be helpful too.

In my experience, these stats are updated daily most of the time. Sometimes it stopped updating for a day or a few days, but then it quickly caches up.

The accuracy of the results depends on the number of data points. For example, you will likely see a wide fluctuation of speed reported by GWT if GWT reports low accuracy due to fewer than 100 data points, but the graph is more stable when GWT reports medium accuracy (between 100 and 1000 data points).

[robzilla]

Good point, I was checking a relatively low-traffic site. Still, March seems a bit too long ago... it used to be updated more often, I believe, or for me anyway.