Discussion: Google Ranking Patent 8,682,892

10 replies
  • SEO
  • |
Google was just awarded a search ranking patent some have said may be related to Panda. Read it here - but also viewing the attached diagrams on the USPTO site will help.

United States Patent: 8682892

You may want to skip down to the "Detailed Description" on the USPTO page to get an "English" version of what is happening:

- Search query is made and list of possible results is identified in the search index.

- The results have an Initial Score which is based on relevance to the search query and quality of the resource.

- A modification is made to the initial score based on Independent Links and Reference Queries.

- "Independent Links" are counted. These can be hyperlinks or references to a site without an html link.

- To determine what is an independent link, the system tries to determine if links are related to the site being considered to be listed in the search results. Factors include:

- Owned by the same entity,
- Hosted by the same entity,
- Created by the same entity.
- Identical or similar content,
- Identical or similar images,
- Identical or similar formatting,
- Identical or similar Cascading Style Sheets (CSS)
- and so on

- "Reference Queries" are searches deemed related to the website. A search is only counted once for a person using cookies or a login identifier to ID the person.

- The Modification Factor is IL/RQ (The number of Independent Links divided by the number of Searches)

- If the Initial Score is not above a certain threshold level no modification to the score is made.

- If the Initial Score is not above a second threshold level the Modification Factor is adjusted so that it decreases the higher the Initial Score.

- If the Initial Score is above the second threshold level a second modification is applied to the initial score.

---------------------
If you think I have recapped any of the above wrong from the patent let me know
---------------------

Hypothetical : search for "Warrior Forum"

warriorforum.com has an initial score based on relevance and quality of 100. There are 2000 independent backlinks and 5000 searches associated with warriorforum.com, or a modification factor of .4 Assume 100 is above the unknown 1st threshhold, and we will assume above the unknown 2nd threshold so the .4 is also modified. Maybe to .44. Perhaps the final ranking score is 144.

acme.com has an initial score based on relevance and quality of 5 because it has a page about the Warrior Forum but is a bit spammy. There are 2 independent backlinks and 100 searches associated with acme.com, or a modification factor of .02. Assume 5 is below the unknown 1st threshold so the modification factor is ignored. Final ranking score is 5.

------------------------------

Some initial thoughts:

- Quality of website impacts the initial score. If this is Panda, or whatever you want to call it, your site gets no modifications to rank better no matter how many backlinks or brand type searches it gets.

- Website references count as links even if not hyperlinked.

- The number of backlinks is important to increasing the modification factor.

- The more popular a search phrase the more backlinks are needed to maintain a high modification factor.

- The dozens (hundreds) of SEO factors Google claims can be put in the categories of (1) website relevance to a search, (2) overall website quality, and (3) which links are connected to each other. This may help simplify SEO analysis.

.
#682 #892 #discussion #google #patent #ranking
  • Profile picture of the author yukon
    Banned
    Interesting read towards the bottom of that page/link in OP, implied link AKA citation & the mention of same CSS files across multiple domains.
    Signature
    Hi
    {{ DiscussionBoard.errors[9099266].message }}
    • Profile picture of the author nik0
      Banned
      Thank you so much for summarizing all this as it makes my head spinning when I read through those 45 claims. It confirmed some thoughts I had lately about the independance of the links.

      They specifically mention CSS, which should not bring us into trouble using a certain CMS like Wordpress as long as unique themes/css are being used.

      However it can devalue the links due to formatting, which often follows the same pattern when using Wordpress, sure not each Theme uses the same pattern, but when you have 100 themes you can be pretty sure that there are only like 10-20 different formats being used, which again can cause the link to be less valued.

      Interesting stuff, I knew uniqueness of a link has a certain count (that's why SAPE works so awesomelly well), and I also always had the feeling that amount of links also weights heavily.

      So if you can buy a domain with 20 independent links or 2000 independent links, both PR3 for example, then I'm pretty sure that the one with 2000 independent links will be more effective, as it influences the overall strength/quality of a site.

      Yesterday I posted that I had the feeling that placing blog posts after blog posts has a diminishing effect after some time and all this kind of confirms that.

      Time to make some changes!
      {{ DiscussionBoard.errors[9099333].message }}
      • Profile picture of the author yukon
        Banned
        Originally Posted by nik0 View Post

        Time to make some changes!
        Keep in mind this is a patent. Just because someone owns a patent doesn't mean it's being used.

        I'll still be testing a few things from that info.
        Signature
        Hi
        {{ DiscussionBoard.errors[9099434].message }}
        • Profile picture of the author nik0
          Banned
          Originally Posted by yukon View Post

          Keep in mind this is a patent. Just because someone owns a patent doesn't mean it's being used.

          I'll still be testing a few things from that info.
          Yes will keep that in mind.

          A good test for me would be:

          - Launch two near identical sites
          - Site 1: Receives 30 links from 5 unique shared hosting accounts
          - Site 2: Receives 30 links from 30 unique shared hosting accounts

          Second test:

          - Launch two near identical sites again
          - Site 1: Receives 30 links from Wordpress sites
          - Site 2: Receives 30 links from a mix of WP/Joomla/B2/Lifelite/Nucleus/Drupal

          Just have to make sure that the PR and other metrics match up a bit so that there are no huge differences based on that.

          Perhaps test it a couple of times.

          Nonetheless, it's not a bad idea anyway to spread it better and have more uniqueness, decreases footprints either way, especially automatic detectable ones.
          {{ DiscussionBoard.errors[9099508].message }}
  • Profile picture of the author yukon
    Banned
    Here's the 2nd guy listed on the OP patent:
    Signature
    Hi
    {{ DiscussionBoard.errors[9099533].message }}
    • Profile picture of the author nik0
      Banned
      Originally Posted by yukon View Post

      Here's the 2nd guy listed on the OP patent:
      There's for sure a lot to learn from these patents.

      Bookmarked it for when I have some extra time to go through all of it!
      {{ DiscussionBoard.errors[9099688].message }}
      • Profile picture of the author yukon
        Banned
        Originally Posted by nik0 View Post

        There's for sure a lot to learn from these patents.

        Bookmarked it for when I have some extra time to go through all of it!
        Some of those patents are from when the guy worked for Yahoo. Still interesting reads & I'm sure the guy brought some of his work over to Google. Google hired him for a reason.
        Signature
        Hi
        {{ DiscussionBoard.errors[9100096].message }}
  • Profile picture of the author kindsvater
    The use of "reference queries" or counting the number of searches done, to impact search rankings, is not something I have previously seen discussed.

    Using a search count as the denominator of a fraction seems odd to me, when the fraction may be used to bump a ranking.

    If I am reading the patent right, and this stuff does make your head spin, the count is not the number of searches done for the keyword being searched, it is the count of the number of searches associated with the website being ranked.

    I think this means the more popular a phrase for a site the more backlinks Google expects to see.

    Examples:

    1 backlink and 1 search = 1.0 modification factor to site's initial score

    1 backlink and 10 searches = 0.1 modification factor to site's original score

    5 backlinks and 10 searches = 0.5 modification factor to site's original score

    0 backlinks = no score modification

    Example: the phrases "warrior forum" and "wso" are associated with warriorforum.com. As these search phrases increase or decrease in popularity, there should be a corresponding increase / decrease in backlinks.


    The patent also refers to normalization of scores. Potentially, 1000 backlinks to a site connected with 1 search, creating a 1000.00 modification factor, is so out of whack the score is normalized down to something reasonable. Maybe the site is flagged for a possible penalty per another algorithm not part of this patent.

    .
    {{ DiscussionBoard.errors[9099644].message }}
    • Profile picture of the author nik0
      Banned
      Originally Posted by kindsvater View Post

      The use of "reference queries" or counting the number of searches done, to impact search rankings, is not something I have previously seen discussed..
      You probably know of these search bots that look for your site in Google, like click bots in an attempt to increase rankings, they even went as far as page 20 I think.

      Some people claimed it worked but that the rankings dropped as soon as they quit using the bot, so it seems to be a real time thing that immidiately gets downgraded when the queries stop.


      Originally Posted by kindsvater View Post

      The patent also refers to normalization of scores. Potentially, 1000 backlinks to a site connected with 1 search, creating a 1000.00 modification factor, is so out of whack the score is normalized down to something reasonable. Maybe the site is flagged for a possible penalty per another algorithm not part of this patent.
      This can be interpretated as when you spam a ton of back links make sure you also buy a ton of traffic to manipulate things in your favor?
      {{ DiscussionBoard.errors[9099682].message }}
  • Profile picture of the author jadog
    Thanks for sharing man. This is the kind of thing you want to be trying to stay ahead of.

    Time to Start coding up a Random CSS/HTML Generator to add some Extra CSS/HTML to my sites on the network all running the same theme I'm guessing they will use this for more than just finding link networks.
    {{ DiscussionBoard.errors[9101349].message }}

Trending Topics