Information Retrieval Based on Historical Data - Sandbox Explanation, Aging Delay?

I am surprised this hasn't made it's way over here yet, so I am posting it so we can discuss it. For some that manage large amounts of sites this may not be a huge surprise as it validates much of what we thought. However there is probably a lot we hadn't considered. A patent was released today that was filed on Decemeber 31, 2003 on the topic of "Information retrieval based on historical data". It basically goes into detail explaining the process for which Google identifies a document and obtains types of history about that document. The system then uses this data to score the document. Here is excerpt from the document:



It also talks about how the degradation of the results can be offset with "fresher" documents. Stale documents can degrade the index so to say. So its important to keep your pages fresh as you can.



There is information that validates what happens when you site initially ranks highly for a few weeks and then suddenly drops.

It also talks about the use of domains in the determining the history of a site in order to help assign a score to the page. Thus looking at the whois data, which they do.

Too much to list here. Check it out for yourself. Interestingly enough we can't be for sure this is exactly what is in use in Google right now. BUT, if they go to the trouble of putting a patent on the technology there is darn good reason they should be using it.

Google's Patent on Historical Data

This was originally released on SEW forums.

Thank you Phoenix, it's a very interesting and amazing document.

regards

Thanks Ben, very telling information. This is better than a lot of opinions.

Dave

Yeah this is a stellar document! There is a ton of information in there. There is even parts about how the toolbar obtains information for scoring, and I think a reference made to favicons in the mix. Plus a host of other topics.

Another interesting fact about the document is that Matt Cutts is mentioned as one of the inventors. He regularly attends SES conferences and is a very nice guy open to talking with SEO's and webmasters.

I've been too busy all day analyzing the document to come here and post about it. It is an incredible read - and validates sooooo much of what we all have been saying for the past year. There is so much information here that I intend to read, reread, analyze, pick apart, and think about. If this isn't the holy grail of google SEO, it is a close cousin. This is a must read. Everyone...read each section carefully, and think about how that section applies to your site. Go...now...do it.

From what I remember reading about New York he was extremely evasive in discussing the sandbox. Guess some of this puts it in a different light.

Dave

This is a hard read. Can any of you pull out the diagrams?

Dave

Noticed references to traffic (claim 34). Makes that urchin purchase more telling.

Just had a wierd thought. This was filed 12/31/03 after Florida, and many changes have occurred since the filing date including sandboxing. More changes occurred 2/2/05. Wonder how all that plays into the document.

Dave

There's a lot in there that makes sense.

Google is now an ICANN registrar. This validates what a lot of us thought about that: that Google now has direct access to domain registration dates which we have known for some time is related to how a site is ranked (ie "sandbox" filters).

Basically here's what it appears to be, in a nutshell (if someone reads differently please feel free to correct):

Google has filed a patent on how their SE works. Specifically they want to patent using site age (based on domain registration date, which Google has access to as an ICANN registrar) and BLs obtained as a method for
1. Determining site rank in SERPs
2. Filtering out "spam"

In short, they basically say site age is not only a factor in SERPs (which we've known for a long time) but is actually a function of how Google views your content and your relationship to the Internet. This is what some people have said the "sandbox" is: an aged filter.

Yes this shines a new light on SEO. After Florida many SEOs were discouraged (including myself) because the algo seemed to not "make sense" and too many discrepancies occured from one site to the next which didnt make any significant rule solid. But now there are reasons as to why anomalies occur. Scary and yet fascinating stuff.

I'll be reading this on my way home today so I'll have plenty to discuss tomorrow

I've been reading it too. I should have a document in plain English with explanantions and hopefully some diagrams (I can't see theirs either, so I'll have to make my own). With some luck, I'll have it finished by tomorrow morning. Sorry I didn't post on it earlier, I haven't done anything else all day but work on this - there's too much good stuff.

BTW - Did anyone notice this is from Dec. of 2003, yet only released today. Is that how patents typically operate - is 1.5 years of secrecy standard?

Wow, this indeed is interesting. Thanks for sharing!!! Is there a way to see those diagrams?

This is the evidence that those of us on the "yes of course there's a sandbox" needed.

While there’s still no evidence that Google actually employs these methods, we now have the fact that they intended to employ them at the time of filing, and that the effects of these tactics are quite obvious today.

Thanks Phoenix. We haven’t all gone mad.

/*tom*/

Frickin' amazing! I can't wait for the geeks to put this in plain english for me/us. I can only grasp about 5% of it with my little pee brain.

Pheonix, excellent find! ;)