What if Google used Analytics Data in its Search Algorithm!

[mccreath]

Going past the argument on if Google should be allowed to use Analytics data in its search algorithms, I'd like to consider the question, what if they did?

Google could refine its organic search results by using analytics data such as:

• * Visits to a page
  * Time on a page
  * Popularity of links to a page
  * Goals triggered after a visit to a page
  * Bounce rate of a page
  * Exit rate of a page

However, they would have to be careful in how they use the data:

• * They cannot cause preferential treatment to websites using analytics
  * Spammers should not be able to improve a websites performance. E.g. by generating fake traffic.

With this in mind, whatever strategy Google used, it would have to be based on the fact that it does not increase/decrease a websites performance in its search results.

So what can Google do?

They could adjust the individual page ranks within a website to relate more to the popularity of a page. Pages that get viewed frequently and for a long time get boosted, while pages that get ignored are dropped.

The important thing is that the overall page rank for the whole website stays the same. In this way, a traffic spammer could only increase its results in one area at the expense of another.

If Google does this, its organic search results will be skewed towards pages that tend to cause positive results (based on viewing time, goals, not bouncing etc.). This means the user should have a higher chance of being provided appropriate results.

For the website owner, emphasis moves towards impressing the user and not just Google. Which is what we should be doing anyhow!

Say there is a website that has 10,000 products and 20 are very popular among the users. This system would result in those products doing better in organics searches. Without analytics data all 10,000 product pages would all be rated equally and at a low level.

We have to make sure the algorithm does not get into a positive feedback loop! Pages should not get boosted because they are already getting good traffic (especially from Google searches or paid links)

The "formula" should concentrate on the quality of experience the user has and not the visitor frequency:

• * How long does a visitor stay?
  * Does the visitor leave the website afterwards?
  * Does it lead to goals being accomplished?
  * Does the visitor return to the search page?

Maybe call it the PageJoy(tm) index ;-)

The original search terms may also be included in the "formula" to further refine pages related to those terms.

The next step would be to correlate search terms used with the above rating for page. This info could be used to skew the key phrases a page relates to in order to match pages up with phrases that create the most "joy".

I believe this would be a step towards the semantic web.

[collusion]

Why not give preferred treatment to those who use analytics? I would if I were the developer on the algorithm.

Should they not do that because a bunch of webmasters will cry "not fair!"?

If you have nothing to hide then there is no reason to complain about that. On the other hand, if you are spamming then that would seriously hinder your abilities.

By giving preferred treatment to those who use analytics Google will seriously increase the relevance in their search results. That would be one factor. Then they would go on to check for spam elements and quality.

[whipnet]

Most sites have Adsense ads on them now. Why do they need Analytics to accomplish this?

*

[mjtaylor]

Originally Posted by mccreath

Going past the argument on if Google should be allowed to use Analytics data in its search algorithms, I'd like to consider the question, what if they did?

Allowed?

And who says they don't use the information gathered by Analytics?

Originally Posted by mccreath

Google could refine its organic search results by using analytics data such as:

• * Visits to a page
  * Time on a page
  * Popularity of links to a page
  * Goals triggered after a visit to a page
  * Bounce rate of a page
  * Exit rate of a page

It seems clear to me that they already use this sort of information to determine organic results. I had a client site recently change a large part of its site so that visitors could use an automated reservations system to search, view and book vacation rentals.

We kept all the old pages on the site as they were, but added a way for the visitor to use the reservation system. Almost all the visitors stopped using the on domain pages to view rentals - and guess what happened to our search query placment on Google. Slipped from page 1 to page 2. Nothing else changed on the site.

My conclusion: Google tracks the traffic to those old pages and seeing far far less on the site for vacation rentals, pushes the site down in the SERPs. Traffic on the real estate side of the site remained constant, and so did the SERPs for that term.

We know Google already returns results to individuals based on their search behavior. Sites I have visited before appear higher in the SERPs when they are relevant to a search.

More personalization is coming ....

MJ

[thorfjalar]

I think taking visitors into consideration (the algo) could improve the quality of search results. Especially measuring time spent onsite by someone searching for particular keywords, that could indicate how useful the information found was. Number of visitors however would be easier to "game" resulting in every ad campaign gaining a bonus from Googles organic search results (the loop you refer to).

My guess is that Google has had this idea ever since they introduced their toolbar and that they are already taking analytics into the equation. In my oppinion they should give it higher value to thwart link selling/buying strategies, but as before said, place most emphasis on trying to measure how useful the individual searcher finds the site, how much time s/he spends reading/viewing the material, how extensively s/he explores the site by clicking on menulinks etc, if the visitor follows a recomended link out or does a type in etc.

[fivetone]

* They cannot cause preferential treatment to websites using analytics

Why not?

* Spammers should not be able to improve a websites performance. E.g. by generating fake traffic.

Would competitors then be able to decrease a websites performance. E.g. by generating fake bounce rate, exit rate, or time on page?

[nexternal]

If the goal is to improve the relevance of the search rankings, then this question is irrelevant.

In this example, the results would be skewed because google only had partial data (not everyone uses GA). For instance, what if a page were wildly popular, but not utilizing GA. What then? Is that page less relevant, more relevant? The relevance is now indistinguishable because you no longer have an objective algorithm or objective perspective.

P.S. If you think that google doesn't use analytics data to make changes to the overall algo, you'll need to explain why G purchased Urchin and why GA is free....

[crazyblue]

I think it's obvious that showing preference to sites using Analytics would damage the integrity of Google's search results.

The best use of Analytics information is to double-check the effectiveness of their search algorithms.

[Webnauts]

Originally Posted by mccreath

With this in mind, whatever strategy Google used, it would have to be based on the fact that it does not increase/decrease a websites performance in its search results.

More about this: http://appft1.uspto.gov/netacgi/nph-...RS=20070088693

Originally Posted by mccreath

I believe this would be a step towards the semantic web.

If that happens, I will be very thankful!!! :)

Some more interesting info: http://www.readwriteweb.com/archives...ion_search.php

[sirius77]

I think that Google already uses many of the techniques you mention.

Using tracking data is almost certainly a centerpiece in Google's efforts to combat click fraud and detect spam.

We have some log analysis data that suggest that all the major search engines use total site traffic as a ranking factor. I will post more on this later.

The idea that Google would not do something that was perfectly legal and benefited Google because it caused "preferential treatment to web sites" is naive. See: http://www.webproworld.com/viewtopic.php?t=73591

It is pretty obvious that Google and maybe all the majors maintain databases of sites as well as databases of pages. Tracking data (visits, time on page, backouts to SERPs, etc.), total site traffic data, and other site-unique data can be used to adjust the site's rank factor. Because Google has such a tracking advantage, they have a big advantage in this area.

Sites running Google advertising give Google more tracking data on themselves. This could be a benefit to the site or not depending on the situation.

[nexternal]

Originally Posted by sirius77

The idea that Google would not do something that was perfectly legal and benefited Google because it caused "preferential treatment to web sites" is naive. See: http://www.webproworld.com/viewtopic.php?t=73591

Frankly, I think the idea that G would cannibalize their patented technology to achieve less relevant results seems naive.

I'm also not sure how a post by you would support your point. For example...

Originally Posted by sirius77

Is Google more sleazy and dishonest regarding their practices despite their "do no evil" motto? In my opinion, yes.

Is that an admission of your own subjectivity? Looks like it.

However, if you could provide some examples of how using partial data could lead to better results, or more google success, I'd be excited to hear about it.

[sirius77]

nexternal: I guess I don't understand what you mean by "partial data?" They are using more data not less. And why would their results be "less relevant?" It seems to me that their results would be better not worse by using tracking data and other information that is relevant to the quality of a web site. And why would using more data be "cannibalizing?" I don't think Google is doing something wrong in connection with their use of tracking data and other analytic data in ranking.

The completely different practices I consider "sleazy" in your quote involve double standards in the way Google treats large-business web sites vs small-business web sites. I also consider false and misleading statements about the "honest" nature of PageRank to be sleazy. Google is the best search service in the world on their own merit -- they don't need to be sleazy to win. So why do they do it? It's a mystery.

[Peter (IMC)]

You probably have to think beyond the idea that using Analytics on a site has a direct influence on the site it self.

Analitics provides Google with (algoritmical) knowledge of keywords and they can extrapolate information about things like trends and preferences of visitors. But in no way can the data of 1 site be used directly in the rankings of that 1 site.

[Webnauts]

Originally Posted by Peter (IMC)

But in no way can the data of 1 site be used directly in the rankings of that 1 site.

Peter I must agree with you. If that would be otherwise, Google PR would turn out to another Alexa ranking system. :)

[mjtaylor]

Originally Posted by Peter (IMC)

But in no way can the data of 1 site be used directly in the rankings of that 1 site.

Please tell us why you believe (or know) that it couldn't.

Thanks,

MJ

[Webnauts]

Originally Posted by mjtaylor

Please tell us why you believe (or know) that it couldn't.

Thanks,

MJ

Because that data can never be objective/accurate.
If Google will require every site owner to use Analytics, that will be another story.

[Peter (IMC)]

Originally Posted by mjtaylor

Please tell us why you believe (or know) that it couldn't.

Thanks,

MJ

It´s mostly common sense... :)

In order to use the Analytics data of 1 site you need to compare it to something. But to what can you compare it? Google doesn't have a set of standards that they compare to.

All their standards are algoritmically determined, which basically means that there is no standard. The standards are different for each keyword. For example, for 1 keyword the results show an average age of backlinks of 8 months, but for another keyword it could be just 1 month. The algorithms figure this all out on their own and what´s even better, these "standards" are completely flexible. They change over time for each keyword.

So back to Analytics data, they have to compare it to "something" and that "something" they'd have to algorithmically take from all the Analytics data they have. But then you run into problems. Because not all sites use analytics. So you can't use it for all sites. In fact you'd only be able to use it for the sites that actually use Analytics.

So instead of creating an algorithm just for those few sites that use Analytics, they´re much better of to use the Analytics data to better understand sites and what visitors like. Then they have to "translate" this information into algorithms that can be applied to all sites in their index.

The good thing is that once you have enough sites to collect data from, you can use that data on all sites. It´s like the poles they do for elections. They don't have to call 200 million americans to get a good idea of what the results of the elections will be, they only need to call a couple of thousand of them and they already can make a very close guess.

[mjtaylor]

Well, thank you both for trying to answer me ... but I am not convinced. It seems totally logical (if not common sensical ;D) that any data collected on a site can be applied to the same site.

If it can be observed (from any of the data available to a search engine, whether from a toolbar or from Analytics) that when someone types in a given query there is a certain likelihood that they will visit a particular site, I don't see why that information couldn't be applied. Why can't it be compared to the data on other related sites? Because the data is partial? Can't it be extrapolated based on generalized surfer behavior?

I don't mean to get contentious here; likely I am in over my head when it comes to understanding how an algorithm is constructed. A reverse engineer I am not. ;D

They call what I do "organic" SEO for a reason - it's not science, it's an art based on experience.

But, again, guys, I do appreciate your attempt to educate me. I learn much from these forums - and I learn the most when I am wrong.

MJ

[Webnauts]

Mjtaylor did you read the page I posted above? http://appft1.uspto.gov/netacgi/nph-...RS=20070088693

[Peter (IMC)]

well, reverse engineering is a waste of time if you ask me. However, understanding search engine logic is something I like. But for "organic SEO" you don't need to understand the algorithms. All you have to do is make sure you use the correct HTML code so the search engines can read your pages as they´re presented to your visitors. Then add some user experience logic on top of that and you´re pretty much set for high rankings. If that doesn't give you enough results, some link building may help. That is in a nutshell, SEO.. :) How good you´re at SEO is more related to how well you understand your market and your visitors than to how well you understand the algorithms.

Adding Analytics to your site can't impact (negatively or possitively) your rankings. That wouldn't make sense at all. Just think of it,.. are you going to advise your clients to add Analytics because it will get them higher rankings? If it'd be that easy,.... :)

It´s the same thing with AdSense,... it has been suggested that adding AdSense to your site gives you higher rankings.... As if Google is going to sacrifice their SERP's to make a couple of Dollars more.

[katropa]

According to Google's patent application (see Webnauts link) they do. Look at their own CONCLUSION:

[0120] Systems and methods consistent with the principles of the invention may use history data to score documents and form high quality search results.

Great!

[billybw]

Thanks for the interesting discussion everyone. I really don't know enough to contribute. But I would like to ask about the Google Toolbar. Doesn't it collect data about user behavior, and why couldn't Google use that data as well as analytics data for ranking purposes?
Thank you.

[mjtaylor]

Originally Posted by Webnauts

Mjtaylor did you read the page I posted above? http://appft1.uspto.gov/netacgi/nph-...RS=20070088693

Not thoroughly, no. I don't have the time or inclination. I skimmed it, though, and while I see the relevance of the topic, it is rather dense. My hat's off to you if you can read text like that through once and understand it.

I did look at the conclusion, though, (always a good speed reading habit) and noted that it says:

>>Systems and methods consistent with the principles of the invention may use history data to score documents and form high quality search results. << (emphasis mine)

Now, I admit I am not absolutely positive about what that means, but it sounds to me as though historic data could, indeed, be applied to a site and produce relevant results.

Do you read that differently, Webnauts?

MJ

[mjtaylor]

Originally Posted by billybw

Thanks for the interesting discussion everyone. I really don't know enough to contribute. But I would like to ask about the Google Toolbar. Doesn't it collect data about user behavior, and why couldn't Google use that data as well as analytics data for ranking purposes?
Thank you.

I would venture it does.

That toolbar is very important to Google - if it weren't they would stop reporting PR to the public through it and not let link traders see the PR value of a potential link trade.

And apologies to katropa who had already noted the patent app conclusion - I just didn't read that far in the forum before responding to Webnauts. M'bad.

Cheers, MJ

[jackit_chick]

Originally Posted by katropa

According to Google's patent application (see Webnauts link) they do. Look at their own CONCLUSION:

Great!

Are you saying that they DO use adsense because of this statement? If you are, I'm going to have to disagree. All this is saying that they use history data...which in my opinion means that if they see a site that has had tons of visitors to it for a long time then suddenly doesn't, it will still maybe stay up high for awhile because of the history data they have obtained. Just an example! Point is, I don't see how the statement says that they use adsense if that's what you're meaning.

[mjtaylor]

Originally Posted by jackit_chick

Are you saying that they DO use adsense because of this statement? If you are, I'm going to have to disagree.

The thread is not about Adsense; the topic is whether Google could use data from Google Analytics to influence search results.

The text was referenced to speak to whether historical data from a site could be used by the search engine to postion that site.

MJ

[Webnauts]

I am still convinced that Google Analytics have no influence on rankings. How many sites rank and often high with Google, and do not use Analytics?

So here I want to add a summary of some information I picked up, which I believe makes more sence:

DOCUMENT SCORING BASED ON QUERY ANALYSIS

[0010] Systems and methods consistent with the principles of the invention may score documents based, at least in part, on history data associated with the documents. This scoring may be used to improve search results generated in connection with a search query.

[0011] According to one aspect, a method may include determining an extent to which a document is selected when the document is included in a set of search results; generating a score for the document based, at least in part, on the extent to which the document is selected when the document is included in a set of search results; and ranking the document with regard to at least one other document based, at least in part, on the score.

[0012] According to another aspect, a system may include means for determining an amount of time one or more users spent accessing a document; means for generating a score for the document based, at least in part, on the amount of time the one or more users spent accessing the document; and means for ranking the document with regard to at least one other document based, at least in part, on the score.

[0013] According to yet another aspect, a method may include determining a set of search terms relating to a particular topic or news item; identifying a first document that is associated with the set of search terms and a second document that is not associated with the set of search terms; generating a first score for the first document and a second score for the second document, where the first score is higher than the second score; and ranking the first document with regard to at least one other document based, at least in part, on the first score.

[0014] According to a further aspect, a method may include receiving a search query; performing a search based, at least in part, on the search query to identify a group of search result documents; determining a staleness of a search result document in the group of search result documents; determining whether a stale document is preferred for the search query; generating a score for the search result document based, at least in part, on the staleness of the search result document and whether a stale document is preferred for the search query; and ranking the search result document with regard to at least one other one of the search result documents based, at least in part, on the score.

[0015] According to another aspect, a method may include determining an extent that a document moves positions in search result rankings; determining a score for the document based, at least in part, on the extent to which the document moves in search result rankings; and ranking the document with regard to at least one other document based, at least in part, on the score.

[0016] According to yet another aspect, a method may include determining an extent that a rank of a document changes over time; determining or adjusting a score for the document based, at least in part, on the extent that the rank of the document changes over time; and ranking the document with regard to at least one other document based, at least in part, on the score.

[0017] According to a further aspect, a system may include means for identifying a document that appears as a search result document for a group of discordant search queries; means for determining a score for the document; means for negatively adjusting the score for the document; and means for ranking the document with regard to at least one other document based, at least in part, on the negatively-adjusted score.

Am I going off-topic? I hope not. :)

[Peter (IMC)]

Webnauts,

How do you see that Pattent application being related to Analytics? That pattent is about query analisys, isn't it? What does that have to do with Analytics?

[Webnauts]

Originally Posted by Peter (IMC)

Webnauts,

How do you see that Pattent application being related to Analytics? That pattent is about query analisys, isn't it? What does that have to do with Analytics?

Sorry Peter, but I guess I was thinking out loud.

Well my believe is that possibly Google looks at the number of clicks per impression to add that to the formula of where to rank your pages. The more clicks per impression, the higher your rankings go.

[Peter (IMC)]

Originally Posted by Webnauts

Sorry Peter, but I guess I was thinking out loud.

Well my believe is that possibly Google looks at the number of clicks per impression to add that to the formula of where to rank your pages. The more clicks per impression, the higher your rankings go.

Ah you mean the number of clicks a site gets in the SERP's. I believe they had a similar pattent which looked at local PageRank, based on the number of links you get from other sites ranking high for a keyword, but I don't believe they actually are using it.

The issue with looking at how many clicks you get compared to position in the SERP has the problem that it kind of helps sites that already rank high and no help at all with lower ranking sites, making it even more difficult to get in the top 10.

However, they may use the logic that if a factor isn't available for a site, it simply isn't incorporated in the overal score. So the overal score wouldn't be the sum of its factors, but more like the average of its factors. In that case you could argue that the more data sources available the better the over all score will be. But that also implies that SEO would get the best results if you make sure Google knows as little as possible about your site and at the same time making sure that the few things they do know are extremely good. Somehow that seems to simple and I can't imagine Google applying such a simple logic.

[Webnauts]

Originally Posted by Peter (IMC)

But that also implies that SEO would get the best results if you make sure Google knows as little as possible about your site and at the same time making sure that the few things they do know are extremely good.

I did not use before Google Analytics, since some time ago. Does that mean that it would be better to get rid of them, or did I misunderstand you?

[Peter (IMC)]

Originally Posted by Webnauts

I did not use before Google Analytics, since some time ago. Does that mean that it would be better to get rid of them, or did I misunderstand you?

No I didn't say that. I followed that line by saying that I don't believe Google works that simple.

In my opinion, if you do things right,.. it will show up everywhere, so it doesn't matter whether or not Analytics data is used. In any way, it doesn't make much sense to use analytics data of a site as part of determining the over all score of that site.

Much better to use that data in general to improve search results.