Google Study Web Page Design

This has to be the most fascinating review of page design to come along in years. Google analyzed 1,000,000,000 documents crawled in December 2005. They break down the data into easy-to-read sections and drop a few confirming remarks about which elements and attributes Web site designers should and should not be using, both for search engines and compliance with future standards-strict clients (including browsers).

or the source


Thanks Gimpy - That is a

Thanks Gimpy - That is a bloody interesting study and one thing I picked up on immediately from a quick scan, that I didn't know about before is HTML5 and ping.

That's damn interesting to me :)

Did I miss something

The button class baffles us. We can't really tell what what it is used for.

I really hope that was sarcasm that I just missed entirely. I'm with Jason who knew there was a proposed HTML 5 spec.

A bit more on the html5 and

A bit more on the html5 and ping feature.

long post but some interesting points in there:

This feature does not generate ICMP messages. It simply requests that the browser perform additional HTTP requests against the "ping" URLs. All traffic is standard HTTP traffic and there should be no new firewall implications.

I no one but me having

I no one but me having problems with the "images" (or whatever they are) on those pages? I just get a black box, and am having to read from the source code to see what was written.

I seem to be missing a plugin, but am not sure which one.

I can't install the plug in

I can't install the plug in either. But I am with Graywolf some of that just reads really weird, as if they had never designed a web page before.

Lacking SVG support?

Note: You will need a browser with SVG and CSS support to view the result graphs correctly. We recommend Firefox 1.5.

firefox AND svg?

do these muppets go out of their way to make things harder?

if I cared about algo busting

there's some good data in there when you aim to *not* appear as an outlier.

Numbers would have added some clarity in some cases. For example, across 1 billion documents presumably collected in Dec 2005 (at webmaster expense, no less) rel=nofollow showed up roughly 9 times as often as rel=license. Wow. That's a helluvalot of success for rel=nofollow, eh? Considering that I have seen a webmaster use rel=license about 2 or 3 times eh-vah!, how many nofollows is that?

Such research is more reason to ban the bots down the road. From the looks of my logs, M$ must be doing a helluva lot of research these days.


there's some good data in there when you aim to *not* appear as an outlier.

shhhhh ...... ;-)

woah... that's out of sight!

woah... that's out of sight! One billion documents, that's simply extreme. An instant "must-read".

Now, how do I go about printing all this stuff so I don't have to read it off screen?

Sometimes (and those are rare) you really want a PDF.

SVG's won't print

When I try to print out the study the SVG graphics become totally unreadable. Dang.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.