Posted by on Feb 8, 2012 in Google, Innovation, On-page, SEO, SEO Industry, social media, Social Search | 0 comments

Part of our internal training here at Catalyst Search Marketing is a monthly webinar I host called, “Patent Watch”.

Last month we looked a Google patent called  ”Social Search“, and yesterday we dug into another Google patent about, “Personalized Search”.

Patent Details:

  • Filed: March 9th, 2010
  • Pub. Date: July 1st, 2010
  • Pub. No: US 2010/0169297 A1
  • Assignee: Google Inc.
  • Link to patent: here

 Overview:

Search engines are pretty good at returning relevant results for a query, but they aren’t very good at tailoring those results to reflect the searchers interests. This patent explains how Google might solve this.

Patent Breakdown:

First we need to tell Google what topics we are interested in. This would be done through Google+ these days, but back in 2010, before G+ was public, the patent showed a “directory” like category drop-down where users could select which categories match their interest. Behold:

Notice that I’ve highlighted the category of: Computers.

If you like computers too, why not subscribe to my RSS Feed so you can get more great info on them computers.

Now:

Let’s say a user searches for [stanford] without an personalization on place. Their results may look like this:

What you didn’t see is that Google actually performed two queries at the same time. The first query returned results for Stanford based only on their Information Retrival (IR) score. The second query, the one you don’t see yet, returned the results with full personalization. As you move the slider towards max, you’ll see results from the second query start to affect the SERPS.

Let’s see what happens

Now I’m going to move the slider half way between min-personalization and max-personalization. Because I had earlier chosen the category of, “Computers” as one of my interests (2 images above) results like “The Register” are now making their way into my top 10.

Now why did “The Register” start to move up? It’s because, “The Register” is a page that is listed in Google’s internal directory as being related to “Computers”. Just as if, “The Register”  was listed in DMOZ under the computers category.

Now What Happens

What happens when we move the slider all the way to max-personalization?

Now, “The Register” has moved all the way to the top? How did this happen?

Boost Scores

Imagine that Google has an private version of DMOZ where it categorizes the most authoritative websites. Each website has a score from 1-10 for the category. So for example, “The Register” may have a boost score of 8 in the category of “Computers”.

Now when I do a search for “Stanford”, Google orders my results based on an Information Retrival (IR) score. That score is based on the standard relevancy and link algorithms Google uses for it’s rankings. Now if I enable personalization, if one of the sites in my initial results is also listed in the same category within Google’s directory as one of my interests, the boost score of that site is multiplied times the initial IR score to create a new ranking order – personalized results!

How to calculate the Boost Score

The boost score is calculated much like PageRank. Google starts with an initial seed list of websites that it knows is very relevant to the category. For example, the NIH.gov would be listed in the Health category. From that initial seed list, Google looks at websites that those seeds link to most. That second set of sites receives a score based on that link popularity. Now Google does another pass, and looks at that 2nd seed list to see what sites they link to most often. That 3rd set of sites receives a score based on those inbound links. That score for the 3rd set is the boost score for that site relating to the category.

So again, if one of the sites in the initial results is bucketed in a category that matches the searchers interest, the boost score for that site is multiplied times the initial IR score to reorder the results. Max personalization uses the full boost score as a multiplier, while partial personalization uses a fraction of the boost score.

Take Aways

  • Links from seed sites will give your site a higher boost score which improves personalized results
  • Follows and shares from influential people within Google+ may act in the same was as seed links
  • Google+ will store your interests
  • Google uses 70+ personalization signals for non-logged in searchers
  • You don’t need to be logged in to Google for personalization to take affect. Your interests can be saved in cookies

Want to attend the next Patent Watch?

We’re considering opening this webinar series up to SEOs outside of Catalyst. If you would like to attend next month’s webinar, leave a comment below. If there is enough interest, we’ll open it up and I’ll send you an invite. Note: You’ll only get an invite if you leave a comment, because that’s the contact info I’ll use to let you know the date/time.

If you’re not already following me on Twitter, my mother highly recommends that you do.