GOOGLE PATENT: Determining Site Quality Score

[aakk9999]

Thanks to Bill Slawski and his trawling of Google patents, we are now aware of a recently granted patent which discusses signals that attempt to define the quality of a website to Google.

The patent tells us explicitly what features it was looking for in a site that might seem to indicate that the site was a quality site.
(...)
This patent is about a search system that includes a site scoring engine that generates site quality scores for sites.

How Google May Calculate Site Quality Scores (from Navneet Panda) [seobythesea.com]

If this patent is in use, then Google is looking at the search queries to determine the quality of the site. Searcher connecting the query with the site would indicate a higher quality of the site for this query. Bill lists several things which are being looked at, such as
- using "site:" operator with the query
- typing the query and adding the name of the site,
- navigational queries to the site
and so on.

In the patent, when counting these instances, the multiple queries that are determined to come from the same user would count as 1, on the other hand, the same query, if determined that it was sent by two different users, will count as 2 (this and similar was no doubt done in attempt to avoid quality score manipulation).

Later on, within the comments to the article, Bill also says:

So, the patent seems to place a lot of value on queries that somehow mention and refer to the site when they ask for information.

Read Bill's take on the patent here - there are lots of very interesting details in his article:

How Google May Calculate Site Quality Scores (from Navneet Panda) [seobythesea.com]

[superclown2]

Interesting. I assume that this would benefit well known (ie 'brand') sites and EMDs.

[seoskunk]

I think it benefits sites with sitelinks for sure and I am pretty sure sitelinks indicate a basic form of trust rank from google

[aristotle]

We"ve discussed this particular signal here numerous times over the years. It's so obvious that I don't know why Google would be awarded a patent for it. Except that the U.S. patent office will give a big corporation almost any patent they apply for.

[toidi]

One site label identifying a particular site can be specified using an operator, e.g., a “site:” operator, followed by a name, e.g., a domain name, for the particular site.

FUD! Nobody in the real world knows what this is or how to use it. If the goog has to resort to this to determine site quality then the algo is pretty much useless.

tomorrow we will be able hire a service that uses their network of zombie machines to run 'site' queries for seo.

[martinibuster]

Makes perfect sense to detect site popularity and perhaps quality by how many times the site is searched for by unique visitors. It's a positive signal.

[francie brady]

I'm sort of concerned about the EMD's getting extra love - hopefully they have something in the algo to deal with that too :-)

[aakk9999]

With regards to EMD, it is not just enough to search for the query, the resulting EMD has to be clicked upon too. From the patent abstract (emphasis mine):

...wherein a query is associated with the particular site when the query is followed by a user selection of a search result that (a) was presented, by the search engine, in response to the query...

What I have seen with a few EMDs I have is that ones that do not get much traffic have declined in ranking over the last 2 years or so. I am also following some other EMDs that are competitors and I can see them sliding down too.

From what I am seeing, having EMD with a great and relevant content and a good UX seem to perform the best at least for a medium to low competitive queries.

[samwest]

So let's all run and chase this metric........NOT!

[aristotle]

toidi -- I agree with your point about the use of the site: operator being a poor signal because the typical searcher never uses it. But most likely Google realizes this too, but just threw it into the patent application to try to beef it up and make it look more impressive to the patent office. In other words, they don't really intend to try to use it, but only added it to the patent application to improve the chances for approval.

I would also like to say again that it's ridiculous to be awarded a patent for such an obvious idea as this, which we've discussed here many times in the past, and which I have no doubt that Bing and other search engines have also used for years in their ranking algorithms.

[Selen]

That explains all these ads posted in 3rd world countries asking users (and often giving them proxy server access) to enter marketed domain and some search term and visit the site.. it is prone to spam.

[seoskunk]

That explains all these ads posted in 3rd world countries asking users (and often giving them proxy server access) to enter marketed domain and some search term and visit the site.. it is prone to spam.

Well I suppose at least their not posting links or clicking competitors adverts.... Now their browsing websites through proxies, too funny. Why would they not just automate it? Anyway it doesn't work unless you have a massive private proxy network, even then there are too many giveaways that any engineer at Google would spot on close inspection, screen size, referer, proxy address, browser behaviour, search term usuage, proxy headers, too many problems going down that path. I think Google has a pretty good map of most public proxies as well.

[whatson]

This really is how the large corporations can simply beat any smaller business trying to be successful online.
As far as EMDs are concerned, yes, they will likely rank high for their domain key terms, but when other keywords are introduced it might no rank anywhere, e.g. key1key2.com, may rank #1 for "key1 key2", but if you search "key1 key2 key3" then it might not even rank anywhere in the top 30.
It is therefore indirectly not a site wide "algo", but per page.

[MikeNoLastName]

Heh, looks like the cat is finally out of the bag. We've been using this and related queries for years once we noticed it working. One of the problems though is that they also apparently now take into account the region of the IPs, So if you "train" it in one country (or even city) the results will not necessarily apply to ALL countries/cities, only those searching near the original IP "cluster".

[supercyberbob]

Time to write a bot (or hire a couple thousand 6 year olds) and spam the Bebus out of this.

Thanks Navneet Panda bear.

[Whitey]

Interesting. I assume that this would benefit well known (ie 'brand') sites and EMDs.

Google suggest based on the probability of the intended phrase and brand bias .... these have been in play for many years in producing search results. Maybe 4 - 6 years.

From what I'm seeing a site can be a clone with thin affiliate content, yet be a strong brand with strong brand queries and see absolutely no effects of Panda. Which fly's in the face of the quality guidelines being peddled by Google at the inception of Panda. Nothing new here, except this patent may reinforce the understanding's behind Google's intent with the disclosed methods.

If you want to rank, users have to like your brand.

[JS_Harris]

Every coin has two sides.

- using "site:" operator with the query
- typing the query and adding the name of the site,
- navigational queries to the site
and so on.

These could be used as negatives against a site too. Google has repeatedly told us to build for people, not for rankings. Webmasters constantly looking up their own pages in Google would indicate they are altering their site with the intent of testing impact on Google rankings. That's not a good thing in Google's eyes.

I doubt they would be looking for signs of manipulation to this degree, but it's Google so who knows.