Ecommerce and Entrepreneurship Blog | About | Contact | Store

Google Site Speed Ranking Factor and SEO Dilemma

Posted on January 7, 2010 by josh

Google has recently made it obvious that site speed is an important issue. So important that they've discussed including site performance as a possible contributing metric to search rankings. Consequently, we've been paying very close attention to site speed and elevated it as a key area of focus for the development team. One area where I'd like some feedback from the SEO community or other web masters is the area of images.

Among the results from the multitude of tools that we've used to tune our site performance (another good topic for an upcoming blog post), most have recommended we serve static content from a separate, cookieless domain or, even better, use a content delivery network. For our sites, that mostly means images. Great. No problem. We know that we can make pretty significant gains. Take this page: This page contains nearly 180 images. Nothing large (the entire page's contents amount to 572 KB), there are just lots of images. {UPDATE: Since I wrote this post, we've reduced the number of elements on this page considerably.}
performance image
So, currently, we serve most of those up from on the webserver. A handful of the images are master page content that we're going to continue to serve up from the www domain, but the rest need to be separated out to another domain or taken off of the page. We know that on a typical low end DSL connection each of these images takes about 175ms to download. We also know that if we serve up all content from the same domain, then most browsers will limit the number of simultaneous downloads of resources. In fact, most older browsers will only allow 2 resources to be downloaded from the same domain at a given time. Additionally, since is not a cookieless domain, the client is transferring cookie data with each and every request! Oy! With an older browser allowing only 2 downloads at a time, assuming 180 images at 175ms each, that means the client can download all images in about 16 seconds (180 images / 2 at a time = 90; 90 * 175ms each = 15750ms = 15.75 seconds).

That is way too freakin' long.
Instead of allowing all that content to be downloaded in serial, we would be well served to parallelize resource downloads. We can do that by serving up content from alternate domains. Too many alternate domains is bad, because the DNS lookups take some time, too. But eliminating the cookies and allowing the downloads to run in parallel should improve load times considerably. When I brought this up to our manager of SEO, the STOP sign came out. We don't want to wreck whatever SEO value we've garnered from images.

According to analytics, we get a significant enough amount of traffic from images that making a change to our image domain could be detrimental. So, here's the dilemma: Which ranking factor do we chase, existing images SEO or improved site speed? We decided to tip-toe into an image subdomain by changing the way images are served on a few pages on and Images on these pages are served from, a cookieless domain. In addition, those images are still available from the subfolder so that links to existing images aren't broken.

Before we go head first into serving up all image content from the subdomain, we want to see that content get indexed. We did this a few weeks ago. No index. In looking for a solution to the index problem, I've found lots of people saying it took as long as 6 months to see an index of their images. We've been around linking to to see if we can siphon an index of the subdomain, but haven't had any luck yet. Although we really have little reason to believe there will be issues with this, we would like to make sure before making site wide changes.

I'd love to hear recommendations on how to get the new image subdomain images indexed. Or, better yet, I'd love to hear an educated opinion on the move to an image subdomain as it relates to SEO and the balance between SEO and the SEO benefits of improved site speed (aside from the other known benefits of improved site speed, like improved conversion rates).

With Google rolling out ranking factors based on page load time, it's obvious to say that if our site is slower, our pages will rank lower, and if our site is faster, our pages will rank higher.  In the most basic form, the lower our ranks, the less we sell. The longer we wait, the more we risk severe and unpredictable consequences. In all probability, of the more than 200 ranking factors (who knows?), site performance will only be a slight influence. So, it's likely that the sky is not falling, but we don't know. Barry Schwartz, in the aforelinked article, even says that "virtually no one complains that their [Adwords] quality score is low because of having a slow site." Despite this assumption, we still want to make this transition as quickly as possible in order to improve site performance.

Comments please!
blog comments powered by Disqus