Ranking based on Traffic flow

[Jazajay]

Hi all
I've been slated in the past, not on this forum, for saying that google looks at your traffic and applys a slight + or - ranking boost.

I have seen an online test that proves it does play a part. Also you should check out the science of pagerank a good bed time book as it is boring as hell, not for any one not into algorithums, but it does document that traffic flow does play a role in PR. Use a well known site that has a view inside this book section. Scroll in 3 pages section 12.? at the top, for the table of contents for that book.

I mean all PR is is a way to caculate the chance of you coming accross a page by chance. IE a page with lots of links like an auction site has higher PR due to more links are pointing to it, so if some one was clicking randomly around the internet there is a high chance they will come accross a well known auction site than a site with only 2 links.

Any way I saw the other day in the SERPS that G has a new "you have visited this page 3 times in the last 10 days" text added to the results for a page to that well known auction site.

This in my eyes proves that traffic has a slight ranking boost. Other wise why record the data. I'm not saying it's massive or you would never get movement in the SERPS as sites at the top would get the most traffic by default. Maybe it's traffic that does not come from the SERPS that is used who knows.

Any way as normal I digress.

I was just wondering if any one else has noticed anything similar and their thoughts on it?

I haven't noticed it main stream as of yet only that link and three others, maybe it is still in beta.

Jaza

Edit: Just relized my error can a mod move this, I hang around this forum too much sorry.

[cryptblade]

Is that all the proof you have? I disagree with you but I'm willing to listen to someone who has a good argument. But you don't give any good proof or evidence. So disagree with you I do.

[jimbeetle]

Any way I saw the other day in the SERPS that G has a new "you have visited this page 3 times in the last 10 days" text added to the results for a page to that well known auction site.
This as been around for quite a while as part of G's personalized search and you only see it if you are logged in.

Now, I agree with cryptblade that it is not proof of traffic being a factor, even if it did appear for all searches. But...if you take a look at the source code for one of G's results pages you'll notice that the good folks at Google do (usually) log clicks -- and have been doing so for quite some time.

That's also not proof that it is using click data in the algo, just that it has oceans full of click data and can apply it whenever it feels it would be beneficial for the user.

[Jazajay]

To be honest I use the net a lot and it's the first and only time I've come accross it.
Is that all the proof you have?

Actually no.
The test I mentioned took a unquie domain name -aghre dot com or something similar- with very little competetion with a few minor articles spread over several pages, on the other few sites competing keyword. It then had 60 differnet people all with googles toolbar installed to hit the site 2ice a day for 15 mins every day for one month. They went to 5 different pages.

The test ran for 30 days around day 14,15 it jummped to the top of the results. Again no serious competetion. The test ended after day 30 and not soon long after it fell down the SERPS. to position 4 or 5 or something.

varaibles used -
1. The site had no links to start with and gained none during the test.
2. No content or onpage optimization was added during the time period.
3. The only changing variable was traffic.
4. Competing sites were few with little content.

The "competing" keyword was a rare long tail that was not heavily optimized with no direct external and few internal links pointing to them on the competing sites. This test proved to me it plays a slight part. However the experimant is at least 2 years old now.

2. I take it you dont know the mathamtically formula to pagerank very well
http://www.amazon.co.uk/gp/reader/0691122024/ref=sib_rdr_toc?ie=UTF8&p=S00B&j=0#reader-page
you should check it out page 3 table of contents 12.3
rankings based on traffic flow, nice little header.
The book is very matamatical and quite dry, not a recommended buy if you are not into algorithums.
Date of publication 2006 so written probably 2005.

3. I still don't see how that is very personalizing my expericence I know how many times I visited a site in the last month, I don't really need G to tell me. Personalization would be to recommend similar sites, something simlar to the site linked above uses, but they would need to track what sites I and other people who also visit that site visit, and amount of times visited for this to work.

The data alone can be used in the algo surly. As in if 80% of people click 3 of more pages and stay for 40mins the site has more relevant information than if 90% of the people click the back button after 5secs.

4. Do you read the disclaimer when you download the G toolbar - will we track your usage statisics, or something similar, a nice spin of coarse to make it sound less intrusive than we monitor your activities.

I personally don't see the improvement it would bring and would just mean more wasted cpu time to G' servers to input the data anyway.

To be honest the added bottleneck in cpu time alone is not justified - so many sites, so many requests, would put so much unesercary strain on the servers for a little useless gimic.

The only logical expernation for the added cpu wastage, which bare in mind will probably be quite high due to the amount of requests G has to handle every day, would be to bring the data in for apart of the algo as the book I refeneced says.

They then can release you have visited this site .... as they all ready have the nesecary data. It would not make sense to waste that many cpu cycles and risk a bottleneck just to bring in a useless gimic.

There is only two possible reasons why they want to track that variable in my eyes and it's not only for personalization.
1. as I mentioned above proper user statistics of how relavant a site is. duration of stay links clicked ='s more users found the information thay wanted.
2. A possible - 12% of people who viewed this site also found this other site of use.

Now that would be an algo.

[Marcia]

Actually, there was a technology in use for several years call Direct Hit that some search engines used, but they're history now.

>>A possible - 12% of people who viewed this site also found this other site of use.

Like Amazon, but that's based on actual sales figures that Amazon has, and they sell all of the ones listed on their site. So it's "in-house" so to speak.

Not so with the SERPs, and it would be very prone to abuse and deck-stacking, just like when Yahoo Directory listings were displaying by a ranking algorithm and the denizens rolled their clickbots out to improve their traffic scores for the click metric in the mix.

[Jazajay]

Ok I didn't know that.

Still I reckon it does play a slight role. Obviously not a major part. I still think the disclaimer on the tool bar suggests more than they are letting on.

If it wasn't Google and you downloaded it it would sound like spyware.

Maybe the test was during the HITS era time frame, it would fit, in which case I have no practical examples to fall back on - dam you Marcia :D

>>Like Amazon, but that's based on actual sales figures that Amazon has, and they sell all of the ones listed on their site. So it's "in-house" so to speak.

True but google doesn't have sales so they would have to use another metric.

>>would be very prone to abuse and deck-stacking

Yeah so is link buying and on page optimization to name 2.

The second people know of a ranking method they will always find away to abuse it regardless of what it is.

I still think that it would still be madness to collect that amount of data for personalization alone that doesn't make sense.

I've just learnt about CPU usage and bottlenecking and how to work around it, or optimize for it when you have enough evidence. So I can increase CPU cycles and speed from my servers. My sources are from people who optimize sites such as flicker, massive amounts of code 50,000 lines, 10,000 lines of JS etc..

If you take the amount of CPU cycles need to add the data use it etc.. X by the amount of sites out there it would not look pretty and all for a gimic. hum to be honest that doesn't make sense to me. G would need to use the data for more for it to be of any benefit for them.

The extra work load, server and financial, would not = the payoff.
That is the first thing CPU optimization teaches you. I like the new compression by the way :-)

I still think it could be used for more for that reason alone.