Google bounce factor research data is in

in Seo Articles, Google News, Web Analytics | Friday, June 8th, 2007 | Trackback

I have been promising this article for a while but was waiting till I had more data in. I have done extensive research and I feel I have now done enough research to atleast provide a basic understanding of this new technology and its effects on your site and mine, how we can combat any problems this may cause our site and how we can actually benefit from it.

The data I will be presenting will be based on my own research, and research done by other participants at SeoChat and 100 anonymous participants who were kind enough to donate their time.

I hope this article helps you better understand what effects Googles behavior factoring has on you site. Please be sure to post your comments and thoughts.

Okay first of all lets go over to SeoChat and notice a experiment which was started by a distinguised member gazzahk and the goal of this experiment was to determine if clicks to a site in the Google results can have a positive or negative effect on a Google ranking. Here is the official thread for the experiment.

Experiment #1 - Do clicks effect Google rankings?

To experiment this theory we asked the audience for particpation and got about 65 who stated they were participating. We took a site and page which was not currently being used for anything and which we were sure would asas not get any new backlinks.

We then found a phrase for which our test site ranked around the bottom of page 1(#10,11) for. We asked users to do this search and click on the result for this particular site. The site in the experiment was not using Google Analtyics so Google could not(I don’t think they could :P) determine what the users did upon clicking into the site UNLESS they returned to Google which all participants were forbidden to do.

It took about 2 weeks to see a significant change, also not all participants entered at the same time which I believe helped keep the experiment looking natural.

At first we only saw a change of one or two positions so the site stuck around positions #10, 11, 9 and 8. But after about two weeks the site started improving much more considerably, moving upto position #4 and even reportedly #2.

Keep in mind two important things for this experiment. First this was a small scale experiment produced on a phrase which was not competitive at all. Secondly probably aside from the participants there were no other searches for this phrase so in Googles eyes 99.99% probably even 100% of the searches for this result ended in that exact site. This told Google that the other results were probably not what he searchers wanted so Google began moving it up.

I attempted to re-produce this experiment on a much larger scale for a medium competition level keyword with about 100 participants and I could see absolutely no change in positioning so my conclusion is that this particular factor, that is the Quality-Click-Analyis factor has a very small value and you will likely only see results if the majority of searchers are clicking that exact site or if the majority of clicks to other sites result in bounces.

So for those of you who were thinking about hiring Indians to click on your sites positions in Google you can forget that idea, I don’t see anything much less than 5-10K indians changing much j/k

So the results of that experiment are the following:

For a company keyword search you will likely see your rankings rise based on the fact that most searchers will click your listing as that is what they want.
For a medium competition level keyword 100 participants all over the world clicking on one listing did not make a noticable change in ranking, probably because there were far more searchers clicking the other listings.
Likely the majority of clicks going to one site for a particular search would help google determine that site deserved a higher position.
Hiring Indians to click your listings probably won’t help you much

———————
Experiment #2 - Google bounce factor… does it exist?
—————————–

Okay this experiment was slightly different. On top of asking participants to click a specific listing we told them to spend time at the site, browse around a bit, click links, pretend to be reading etc - Basically act as though you love the site in a natural manner.

On this site unlike experiment #1 Google Analytics was installed on all pages. Like experiment #1 the site/page was not being used and no links were gained before(about 1 year), during the process or even afterward. Never were the sites linked to for any of these experiments. We used three sites for this experiment as well as the data from three sites Google Analtyics accounts which were kindly offered by three webmasters at SeoChat.

For this experiment I had another 100 participants who helped make this experiment possible. I had hoped for something like 500 people and did not expect a significant negative or positive change in ranking based on just 100 participants.

Okay let me explain what this experiment was trying to prove. We were trying to prove that the Google Analtyics data was/is being used to effect sites rankings in Google. We wanted to prove that the Google Bounce factor actually exists and that visitor length on a site plays a part in ranking or the loss of a ranking in Google. So what did the data say? Read on…

I had the participants do their part of the experiment over a weeks time, similiar to the first experiment. They would click the listing, spend on average 7-10 minutes at the page and then would proceed to click related links(inside the site. No outbound links) on the site and eventually spend anywhere from 30 minutes to 1 hour on the site. They would then exit to a page on the site that was not watched by Google analytics but was still on the site. This way Googles last notice of the movement of that user was still on the site.

For the first week we saw no change and were beginning to think the phrase was too competitive and that more users were needed. After about a week we started seeing some changes. After a week and a half the site jumped 47 positions from about 80 to about 33. We believe that with continued experimentation we could have ranked it on the first page.

During this time, while the experiment was being concluded the bounce rate was down to 13% and the average visit length went from 1 1/2 minutes to 12 minutes.

After the experiment ended and users stopped clicking into the site we watched as the sites bounce rate began to go back upto about 75% and the visit length to return to about 3 minutes. After about 9 days we saw the site return to a little better than it was before the experiment(around 78) - Again nothing was done with this site/page except for the experiment. It was basically a dead site.

So based on this site and my own research on my own sites I have determined the following.

The Google Bounce Factor does indeed exist. Too many searches resulting in a click to your site which then result in a bounce could negatively effect your site.
Somehow lowering this bounce rate, by providing what the users want most likely can indeed have a very positive effect on your site.
Visitor length likely has a small effect on your rankings as well however in all experiments we could not accurately determine this as a fact. So it still remains speculation. However we are now positive the bounce factor does exist.
Google Analytics data is INFACT used by Google to manipulate results in their search engine. By using Google analytics your data could have either a negative or positive effect on your site.
The overall bounce rate for the site and each individual bounce rates for each of your keywords plays a role.

Many of you are probably thinking you should remove the Google Analytics code from your sites and stop using their service. This is debatable. It really depends on your sites data, is it that bad? If your bounce rate is much higher than 65% and closer to 75% and your average visitor length is very low I would recommend removing the code for now. Maybe use Clicky Web Analtyics 2.0 for a while until you can get your bounce rate lower. Also if you have a high bounce rate I would try and figure out why, try and fix the issue and then maybe you can use Google Analtyics again.

I am not sure on this, I have no evidence yet to support it but I also believe the data from your Google Analytics account may be used for Google Adwords positioning too. I don’t do enough PPC to test this however. If a Google Adwords guy would like to test that theory be sure to let me know the results.

Many webmasters have expressed their concern that this type of ranking factor could influence webmasters to hire users to click their listings just to raise their rankings. Well we have already determined that it is highly unlikely you will gain any ranking even remotely competitive based only on clicks however taking into effect whether or not the majority of users bounce and how long they stay on the site can have a effect on your rankings.

However the data also indicated that this isn’t a one-time thing. The data overall and overtime can continue to have an effect so you may lose your rankings if the bounce rate gets high again.

This research is based on the SeoChat experiment, my own sites and about 5 other sites who were kindly donated for use in this experiment. I awant to personally thank all those who participated their time, their sites and their analytics info. It was a great help in making this experiment possible.

If you have comments/questions please let me know.

If you have recently lost a ranking in Google I would be happy to take a look. Send an email to randy[@]1st-rankings.com

44 Responses for “Google bounce factor research data is in”

Jacob F. Says:
June 8th, 2007 at 6:58 pm
I followed your link from SEO Chat, and your article is very useful. Have you considered disabling Analytics on the test sites after the last phase of the experiment - when their rank drops? It would be interesting to see if disabling GA would bring them up, and to what level they would rise.
visio Says:
June 8th, 2007 at 7:11 pm
No on the test sites for this experiment or my own. I haven’t yet had enough data on that to determine what would happen.
I speculate that upon removal the ranking would be solely based on links and other seo factors and not analytics data so your position would return to that which it would have been prior to using Google analytics.

Again no data on that yet. But it is a good idea. Maybe I will try it and see what happens. Thanks for the comments!
Gary Wells Says:
June 8th, 2007 at 7:14 pm
I thought I had seen some changes after I added Google analytics. Now I know why. Thanks for the explanation.

Gary
WebGeek Says:
June 8th, 2007 at 8:48 pm
Randy, that is a great experiment you’ve conducted! Great work! Possibly the first one out there like that - at least the first one I’ve heard about. This could open the door to a whole slew of even more extensive experiments. This is valuable SEO data! I updated THE post to include a link to your findings. Thanks for sharing this fantastic info!
visio Says:
June 8th, 2007 at 10:54 pm
Yeah I did alot of searching and couldn’t find many articles on it. I do hope this will get the wheel rolling so others do more research and figure out even more.

Thanks for the link by the way!
Barry Says:
June 9th, 2007 at 4:37 pm
Always GOOD to see new info and experiments in the works so good job. I still do’nt think we can make huge conclusions from your results though. a) 100 users is quite small, b) regular google flux is always happening that that you cannot control c) competitors are doing their thing all the while as well. those are some intangibles that may have affected results. overall good job though, now do it x1000 the scale!
Sean Says:
June 10th, 2007 at 2:01 am
Great article. Very interesting
WebGeek Says:
June 10th, 2007 at 12:24 pm
RANDY - RIGHT ON, and you’re welcome!
Prairie Says:
June 10th, 2007 at 9:28 pm
awesome research, visio, thanks for your time & dedication.
Dylan Says:
June 12th, 2007 at 7:54 pm
Very Interesting Experiment but I do have a question about the reported rankings. The write-up mentions that The Page in the First experiment moved “upto position #4 and even reportedly #2″. Were these rankings being reported by the people participating in the experiment? If So, were they clearing their cookies and being careful not to be signed into a google account? If not, then the findings could be tainted by personalized results and not an actual improvement in rankings.
visio Says:
June 12th, 2007 at 8:07 pm
DylaN that is a good point. I say reportedly for that specific reason, I am not positive it was accurate. The #4 position was positive increase and is proven. I participated in the experiment with one network and logged onto another to see the results so no correlation was made with the participation and the results.

Thanks for the question though. Research has to be thought out, one little flaw and it is all messed up.

I once did a test like that and because of one lousy point the whole thing was ruined.
Frederick Townes Says:
June 13th, 2007 at 12:12 am
Fantastic work! I’m thrilled someone took time to prove the convjecture that we have had concerns about for some time. Thanks!
STELLA rOY Says:
June 13th, 2007 at 1:22 pm
thANKS FOR THE INFO. i AM ALSO USING GOOGLE ANALYTICS FOR SOMETIME. i ALWAYS USE TO WONDER WHY GOOGLE IS GIVING SUCH A POWERFUL TOOL FOR FREE. NOW I KNOW WHY AFTER READING THIS. WHAT NEXT CAN BE DONE TO MAKE GOOGLE STOP DOING THIS TO MILLION OF PEOPLE AROUND THE WORLD?

-sTELLA rOY
Ajeet Says:
June 13th, 2007 at 5:02 pm
Very interesting analysis. LOved it. Gives food for thought.
Embuck Says:
June 14th, 2007 at 9:55 am
This is really great case study. thanks for sharing such a interesting findings.
Chris Mcgiffen Says:
June 14th, 2007 at 12:54 pm
does this however conclusively prove that google analytics data is being used to calcualte the serp’s? there still seems like there could be other factors at work here, such as query popularity changing weightings of terms, or bounce rate as measured by the time users are away from the serp’s. neither of these actually needs data from analytics to have an effect.
Best_optimized Says:
June 15th, 2007 at 11:28 am
I have done some testing one of my sites also and have found that on all of the terms that visitors don’t spend much time of the site that the Google rankings fall to 50 to 500 in the serps. Although this don’t prove it was
analytics that did that. I have taken analytics off that site and will see what happens now. I don’t think “bounce rate as measured by the time users are away from the serp’s.” would work for Google because users could open the site in another window or tab.
Robbert Says:
June 19th, 2007 at 4:00 pm
Very interesting to say the least, Using analytics for many customers, so it’s interesting to see these kind of experiments.
Peter van der Graaf Says:
June 20th, 2007 at 9:20 am
Google doesn’t use Analytics as much for bounce information and other interesting stuff they can find there. They mainly use it to teach their algorithm, but I’ve seen proof of testing the use of analytics data in ranking, but so far no proof of real integration into the algorithm.

For bounce rate info they use: Visitors that quickly return to the search results to find another website!
visio Says:
June 20th, 2007 at 1:12 pm
So basically your saying that Google only tracks visitors returning to the search and does not use any data from google analytics.

I have heard that before however recent testing has proven different. I will have another article out on that within a week.
Jason Says:
June 20th, 2007 at 1:17 pm
I own a good many sites and I have been closely watching the analytics data upon hearing this and surprisingly your right. High bounce rate keywords drop like stones but the low bounce rate keywords keep rising.

I took your advice and took analytics off one of my sites because the bounce rate was just way too high.

Thanks for doing the experiments. It has greatly helped improve my understanding of how Google works.

Jason W.
Peter van der Graaf Says:
June 20th, 2007 at 9:17 pm
I will ask my reputable sources within Google and people close to Google developers. This is something they probably will be very frank about.

It is very hard to do any real research on this subject just by activating and de-activating your Analytics accounts. Although I’ve seen proof of Google testing with analytics data, Google toolbar data and such, so far no proof of that bounce rate being used over the normal return rate.

Please prove me wrong, because I know many ways to influence Google’s measurements. If this is becomming an important factor, I can be #1 on everything.
Las Vegas Guy Says:
June 21st, 2007 at 7:58 pm
That is very, very interesting. From a google standpoint, it would make perfect sense because they set up the “voting’ system via links in the first place. Now they are taking votes from surfers….perhaps. Interesting about the bounce rate. Now if they could just find a way to get rid of the garbage in the index.
Peter (IMC) Says:
July 4th, 2007 at 2:23 am
Randy,

Interesting research, but there are some points to consider. I have access to more than 50 google analytics accounts and decided to do some comparisons. One of the things I looked at is the amount of traffic in relation to the bounce rate. According to your research, when I see the bounce rate go down, traffic should go up. But there is an issue here. In my experience, bounce rates go up when traffic goes up. Especially if most of that traffic comes from search engines. But that´s the overal bounce rate.

Bounce rate per keyword might be different, but then the logic should be that the higher ranking phrases have lower bounce rates. based on what I see in all the accounts I have access to, that relation doesn’t exisT.

But I did a quick overview so I might be wrong. It’s interesting enough though. One thing I am sure of is that what ever you´re looking for in your experiments, is right there in your analytics data. All you have to do is extrapolate the date from it. (i.e. put it all in excel and start comparing )
Mehmet Buyukozer Says:
July 5th, 2007 at 6:23 pm
Very good job! I had these questions for quite some time. Do you know any experiment similar to this to check google vote buttons effect on rankings?

thanks
visio Says:
July 5th, 2007 at 6:38 pm
There are a few problems with this factor, I was actually beginning an article on the problems with the Google bounce factor. I agree there, but I do believe these experiments have proven it does exist.

Not always will the highest ranking sites have the lowest bounce rates, like you I have seen this to be true(BTW glad your doing your own research, keep it up!!!)
I am trying to test it on really competitive keywords like “web hosting”.

I am still researching it as there are many aspects I have no idea on yet, And if you happen to find any interesting stuff please let me know.
visio Says:
July 5th, 2007 at 6:41 pm
Mehmet: NO i haven;’t however Matt wrote an article at his blog which mentioned those so I am assuming they have some effect, I might try that sometime. I think they are much less valuable then the bounce rate though.
Josh Says:
July 24th, 2007 at 6:57 pm
aH, gaming google continues to be fun…
Katy Jo Says:
July 25th, 2007 at 6:36 pm
I guess the outsourcers will just have to keep clicking on competitors websites to drain their cpc accounts then…
Garry Says:
August 5th, 2007 at 5:49 pm
Very useful article. Thanks.
Julio Otimização de Sites Says:
August 21st, 2007 at 8:49 pm
iNTERISTING ARTICLE. iT´S HARD TO RUMINATE ABOUT ALL THE POSSIBLE THINGS THAT CAN AFFECT YOUR PLACEMENT SO I DECIDED TO STICK WITH WHAT HAS BENN WORKING FOR ME AND MY COMPANY.
CD Rates Says:
September 7th, 2007 at 4:45 pm
The one potential problem I see, is “what is a true bounce?”. Some discussion has been going on at WPW, Bounce Rates

Google defines a bounce as a visit that doesn’t go anywhere else on the site, and doesn’t mention anything about time spent on that page.

If the page the visitor comes to is highly relavent to what they are looking for, and have no cause to go elsewhere, you would expect a high bounce rate, correct?

How can google punish you if the page is working as designed?

Do adsense clicks count as a bounce?

ChrisCD
Jumbo CD Investments, inc.
Aurélien Says:
September 8th, 2007 at 11:45 pm
GreaT analysis thank YOU !

Your Article In French : http://www.outil-referencement.com/blog/index.php/403-google-bounce-factor
Andy Fletcher Says:
September 10th, 2007 at 10:14 am
Vision, i’d love to get involved next time you need help with this kind of testing - an hour of my time out of a week would be yours if you want it. thanks for a great article
visio Says:
September 10th, 2007 at 11:44 am
CD: If the page is highly relevant and they have no cause to go elsewhere then they would stay… that is not bouncing, bouncing is when they leave your site.

I am not positive on adsense bounces… I assume they would be similiar.
visio Says:
September 10th, 2007 at 11:46 am
Andy we’d love to have you, I have been extremly busy with a massive project lately so haven’t gotten too many experiments going, we only got a few right now but in January 2008 we are planning some good ones so check back with me then. There is also a reason for this wait till Jan, we expect an interesting turn of events about then.
CD Rates BLog Says:
September 10th, 2007 at 6:21 pm
Visio,

In my mind I would agree. If they stay they aren’t just “boucing” Away. But the Google definition of a bounce, makes no mention of time spent on the page. Which leads me to wonder.

The google definition is –”Bounce Rate: Bounce raTe is the percentage of single-page visits (i.e. visits in which the person left your site from the entrance page). Bounce rate is a measure of visit quality and a high bounce rate generally indicates that site entrance (landing) pages are not relevant to your visitors. You can minimise Bounce Rates by tailoring landing pages to each keyword and ad that you run. Landing pages should provide the information and services that were promised in the ad copy. ”

So for me I guess I need to look at time-spent and bounce rate to determine if any one page is effective or not.

But I wonder if google is doing the same thing so as to not penalize a page that is working as intended.
visio Says:
September 13th, 2007 at 9:36 am
CD believe me, in our tests time does play a role. Bouncing BACK to Google seems to be the biggest thing, bouncing onto another site never seemed to be treated like bouncing back to Google was.

As i said before this process if very new so there will be bugs so I hope they get it right
Experienced SEO Daryl Says:
September 25th, 2007 at 4:44 pm
Great article, and research. This is why I wrote my own stats counter, last thing I want is anyone having my search data.
visio Says:
September 25th, 2007 at 5:28 pm
Well I would be interested in taking a look at it… if it is good I may even review and recommend it and help make it popular
Ben Says:
October 9th, 2007 at 8:54 pm
I hate to be a party pooper, but I just heard from my direct contact at Google Analytics, and he laughed at me when I mentioned the idea suggested by this so-called ‘test’.
unless you folks are all into consiracy theory, this test has no validity whatsoever and is directly being refuted by Google. By the way, our google contact says that Analytics is “a completely separate system” from that of Google search. No data flows between the two, but the google toolbar does report relevance data for search. Data usage is always explicitly disclosed when you sign-up or agree to use any of google’s products.
visio Says:
October 10th, 2007 at 9:32 pm
HAHAHA thanks for the laugh!!!

First of all never ever goto google to learn how goolge works.

Lets put this into a logical equation…

First we have google who owns google analytics, now of course they want you to use their software correct? So why would they admit to something such as analytics effecting google search when it could potentially cause a massive drop of users using their analytics services.

Secondly many of the reps do not know how advanced google search actually is and they may not be aware of the use of google analytics as a factor…

Now I believe my studies have proven it is a factor and not only that but many webmasters/seo have challenged me and did their own research and so far all of them have come back with “Hey your right, google does manipulate their search with our analytics data.”

If that is not enough for you go read Google’s privacy statements and their patents, you’ll see it is covered there - That doesn’t prove they use it but it does disprove your illogical idea that just because a google rep told you it couldn’t be done than it must be so.

Now if someone can come to me with data and say “HERE, this does not add up, I researched what you said and something is missing, I am not getting the same results” - Come to me with that and you have a case, we will tackle that together but don’t come to me telling me what google says, I could care less what they have to say. I research my theories and I base my articles and seo on that and that alone not on what google has to say on the matter.

Come back when you have a real argument…

Ps. I removed your links, I didn’t see where we needed 10 links in the url field so I figured you didn’t need any, was that something that same google rep recommended? lol
Ben Says:
October 11th, 2007 at 7:37 pm
I want to correct my statement about the “links.” I believe I listed 4 websites, not 10 as VISIO STATES.
Dave Says:
April 17th, 2008 at 6:59 pm
That’s amazing, it makes perfect sense now.
We’ve seen SERP drop for ‘apparently no reason’ and we hadn’t made the correlation before.
I’m going to check the rest of your experiments now and see if there’s anything else that will help.

Thank You!
Dave

Google bounce factor research data is in

44 Responses for “Google bounce factor research data is in”

Submit your comment!

Categories

Navigation

RSS

Archives

Meta

Seo Experts Forums !