How Might Google Use Site Speed as a Ranking Signal?
On April 9th, 2009, many people developed an interest in speeding up their websites, after reading a post on the Google Webmaster Central Blog – Using site speed in web search ranking.
On the same day, Google’s Matt Cutts published Google incorporating site speed in search rankings on his blog. These posts introduced site speed as a ranking signal that Google would be using.
Matt Cutts told us that site speed wouldn’t be an earth-shattering signal. And that it might not have an impact within a large set of rankings. But he did stress that site speed has benefits other than just ranking, including improved user experience.
The blog posts didn’t give us a full-scale breakdown of the kinds of things that Google might be looking for when it comes to site speed. Or how Google might use site speed when ranking pages.
A Google patent granted on February 4th gives us more details.
So why did Google decide to use the site speed that a page loads as a ranking signal?
Putting it simply, the site speed patent tells us:
Given two resources that are of similar relevance to a search query, a typical user may prefer to visit the resource having the shorter load time.
Google has worked to help site owners with tools that can help them explore issues related to their site, such as their online PageSpeed Insights tool:
The PageSpeed Insights tool gives sites a score base upon how well they meet many rules (or heuristics) involving how quickly a page loads into a browser. These aren’t the factors cited in the patent, but the tool is very helpful to people attempting to improve their site speed.
There’s also a lot of information about those rules, why they are used, and how they can be implemented. Many of these are technical and you may need the help of a developer or someone who has experience optimizing the speed of sites.
Load Time Comparisons
The Google Site Speed patent is at:
Using resource load times in ranking search results
Invented by Arvind Jain and Sreeram Ramachandran
Assigned to Google
US Patent 8,645,362
Granted February 4, 2014
Filed: November 12, 2010
Abstract
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for using resource load times in ranking search results.
In one aspect, a method includes receiving a search query from a particular user device; receiving, for each of a plurality of resources responsive to the search query, a respective first score; accessing load time data that specifies, for each of the plurality of resources, a load time measure for the resource; and adjusting the first score for each of the plurality of resources based on the load time measure for the resource to generate a second score for each of the plurality of resources.
The load time of an online resource can be based on a statistical measure of a sample of load times for several different types of devices that the page or resource might be viewed upon.
The site speed patent points to these as factors that impact load time in a browser:
- The size of the resource
- The number of images the resource includes or references
- The web server that serves the resource
- The impact of the network connection on the loading of the resource
When Google measures load time to compare two different pages or resources, it might limit itself to devices that (1) are in the same country, and (2) use the same user-agent (such as the same browser).
Load time data might be collected from a web browser, a web browser add-on, or monitoring software associated with a particular user device.
Site Speed Take Aways
The site speed patent tells us that when there are two different pages or results for a query, and one load relatively quickly, while the other loads relatively slowly in comparison, the quicker result might be promoted in display order and the slower result might be demoted, so that the quicker page will appear higher in search results.
There are more details in the site speed patent, such as how it might “predict” load times for some pages. We’re also told that load time data for mobile devices might not be included, “because of the high latency of all requests for resources on such devices.”
This kind of load time information also might not be used in all cases, because “some resources may not have enough traffic from particular locations or types of devices for load time measures derived from devices sharing a particular attribute to be meaningful.” The patent gives us these examples involving site speed:
A resource in Chinese may not have enough visits from user devices located in France to generate a meaningful load time measure using solely devices from France, or a newly launched website may not have sufficient load time data associated with its resources.
If you can improve the speed of your site, it’s not a bad thing to do. It may not be as strong a factor as something like relevance or an important signal such as PageRank, but it could make a difference when two different pages are very close in both of those areas, and one loads much faster than another.
Added: 2014/02/14 – Devin Holmes, whom I work with at Go Fish Digital (he brings some great designs to our work), pointed me towards this article this afternoon – Google speeds up Chrome by compiling JavaScript in the background, which shows how serious Google is about trying to speed up the Web. Where this patent talks about comparing sites while considering things like different user agents, this is one of the reasons why it looks at user agents. If Google compared load times for two different resources they might do that while considering times to load websites on the same version of the Chrome Browser, for example, where Google’s newest tweak to the browser could help pages load faster.
Hi Bill and great post.
I wrote an article some days ago about it http://bit.ly/1g3ugjB.
It’s in Italian, but I think it’s very interesting and in-depth, maybe with gtranslate you can try to read it 😉
Hi Theo,
The patent tells us that they might not be including mobile speeds in the models they build to try to gauge how quickly or slowly a site might be, because mobile speed numbers aren’t very reliable. I don’t think that’s changed too much since this patent was originally filed.
Hi Yousaf,
I’m not sure what you are referring to with the phrase “industry averages.”
The interesting thing about this part of the algorithm is that Google is using this as a ranking factor based on industry averages.
Hi Andrea,
Very nicely done – I like all of the additional resources that you included in your article. The shopzilla example/case study itself is a really good one, but you’ve added a number of additional links to some pretty good pages.
The original statements about site speed from both the Google Webmaster Central blog and Matt Cutts blog didn’t go into a lot of detail, but the patent does. Though we don’t really know how much of what is described within the patent was used by Google. Regardless, making a site faster does have benefits that do go beyond improved search rankings.
Great read. Site speed was is and will be even more important with the raise of mobile. Thank you!
The speed of your site being a factor in the search algorithm doesn’t surprise me. In the fast lifestyles we have and the connection to the internet getting better each day no one lead is going spend time and wait for your site. Plus those millisecond cut down means that you are interested in the performance of your site and offer quality.
Great post Bill, thanks for extracting the most pertinent parts of patent & dissecting them 🙂
It’s interesting that many people speak of speed being increasingly important for mobile search – yet Google seems to state that they don’t trust their data from mobile to provide clear enough signals to implement this into a rankings – I wonder if this will change in time?
Of course I guess it’s still a good idea to have a very mobile friendly site – partly for conversions of course, but also I’m wondering if a (really) slow page load for mobile could still have an indirect (not relating to this patent) effect on mobile ranking?
My thought process here being if a page/site is really, really slow or poorly optimised for mobile from a UX perspective say, the bounce-rate from Google’s SERPs to that page/domain could be incredibly high, which may (or may not!) be a strong enough signal to warrant a lower ranking in the SERPs when the searcher is using a mobile browser… What do you think Bill? Do you think this could happen, or do you think it too noisy a signal to be used in mobile SERP?
Thanks Bill, another interesting article from you 🙂
Hard to beleive it’s been since 2009 that talks of site speed as a signal started. Considering it’s been that long, if you haven’t adopted the methodology that site speed should be important you may want to change that.
It makes sense… Google is trying to provide a good search experience. If there are sites that have viruses, spammed with ads (two things we already know they filter out of search results) or slow load times then their product suffers.
Hopefully, this added impetus will encourage businesses to look more closely at their load times. Even if it doesn’t have a huge impact on their rank, it does have a noticeable impact on conversion rates – so anything that highlights the importance of site loading times is great by me!
Site speed was never taken seriously by many big players also and they paid the price. Now it has become an important factor and is at the top . I agree user experience is directly impacted and in the long run does impact rankings.Great post thanks.
Makes sense. According to Google philosophy they’ve “always relied on user feedback to improve the quality of results” – so if users don’t like a site, bounce, or get frustrated and leave quickly, that’s essentially feedback about a resource that Google can use to improve search.
Thanks for making me discover the site speed tool, I use to use others but this one gives you a really solid break down of what you can work on to make your site faster. Can be crucial in today’s current SEO climate.
I have compared different domains with very similar metrics such as domain authority and mozrank and the real comparable different was the loading speed. The site loading faster ranked 2 to 3 positions higher.
We have definitely noticed a remarkable increase in rankings on a number sites which have been through a site speed test and improvement. Anything that improves user experience is a good thing – someone mentioned an improvement in conversions too.
Great post, I always make use of PageSpeed, GTmetix and Pingdom speed tools whilst in the development stage.
Great read. Makes sense to include this in the criteria. My only gripe is some of Pagespeed Insight’s recommendations are like W3C compliance–sometimes overly complicated and oftentimes unnecessary.